JP2005276797A - Fuel cell power generation system and control method thereof - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a fuel cell power generation system capable of efficiently controlling generated power by predicting different power used for each home and realizing energy saving and a control method therefor.
An electric power measurement unit 230 measures electric power used by a home appliance, and an electric power use prediction unit 200 determines a predicted value of future electric power usage from a predetermined time based on the electric power measured by the electric power measurement unit 230. The power generation command unit 220 determines whether to start and stop the fuel cell power generation apparatus 100 based on the predicted value of the used power of the used power prediction unit 200.
[Selection] Figure 2

Description

  The present invention relates to a fuel cell power generation system including a fuel cell power generation device that generates electric power to be supplied to an electrical device, and a control method thereof.

  2. Description of the Related Art Conventionally, a fuel cell power generation device that generates power from a fuel cell that supplies electric power and hot water using fuel gas is known. It is desirable that the power generated by the fuel cell is equal to the power used in home appliances in consideration of power generation efficiency. Therefore, it is necessary to measure and manage the power used by the home appliances and to control the generated power so as to follow the value. However, there is a possibility that the fuel cell cannot follow a sudden change in electric power used due to the characteristic of utilizing a chemical reaction between hydrogen and oxygen obtained by reforming city gas or the like. For this reason, if the power generation output of the fuel cell power generation device is controlled based on only a simple change in power consumption, for example, a decrease in the utilization rate of the generated power at the start-up of the power usage, The surplus of the generated power at the time of going down will reduce the energy saving performance. Therefore, in order to ensure energy saving, it is necessary to avoid a rapid change in power consumption and frequent start / stop.

  Therefore, in the conventional fuel cell power generation device, it is assumed that the characteristics of changes in power consumption are assumed for each time zone, and the control of the power generation output of the fuel cell is specified and held for each time zone by the change rate, dead time and offset. Thus, even if the power consumption changes sharply, frequent start / stop is not performed without following (for example, see Patent Document 1).

In another conventional fuel cell power generator, the power used is predicted in advance by simulating, and the power generation output of the fuel cell is efficiently controlled according to the predicted value (for example, Patent Document 2). reference).
JP 2002-291116 A JP 2003-61245 A

  However, it has been difficult for the above-described conventional fuel cell power generation apparatus to adapt to the power consumption load for each household. That is, in the case of the technique disclosed in Patent Document 1, since the change rate, dead time, and offset that should be set in advance are originally different for each home, it is difficult to calculate an optimum value in each home. Furthermore, it is difficult to cope with cases where the usage status of the device changes due to seasonal changes or lifestyle changes, and the power consumption changes greatly.

  In the case of the technique of Patent Document 2, the power usage load is predicted by simulation in the load prediction step, but it is difficult to generate a predicted value if there is no past data that matches the equivalent environmental conditions. In order to perform the simulation, the characteristics of the fuel cell are required. However, since the characteristics are assumed to be different for each household, a predicted value cannot be generated substantially. However, it is difficult to predict different power consumption.

  The present invention has been made to solve the above-described problems, and is capable of efficiently controlling generated power by predicting different power used for each home, and can realize energy saving. It is an object of the present invention to provide a system and a control method thereof.

  A fuel cell power generation system according to the present invention is a fuel cell power generation system including a fuel cell power generation device that generates electric power to be supplied to an electrical device, the power measurement means for measuring the power used by the electrical device, Based on the used power measured by the power measuring means, a used power predicting means for predicting a future used power for a predetermined time from a predetermined time, and a predicted value of the used power predicted by the used power predicting means. And an operation control means for controlling the operation of the fuel cell power generator.

  According to this configuration, in the fuel cell power generation system including the fuel cell power generation device that generates the generated power to be supplied to the electric device, the used power used by the electric device is measured, and based on the measured used power, a predetermined power is used. Future power usage is predicted for a predetermined time from the time, and the operation of the fuel cell power generation device is controlled based on the predicted value of the power usage predicted.

  Further, in the fuel cell power generation system, the power consumption predicting means includes a prediction data holding means for holding power used before the predetermined time as prediction data, and a prediction held by the prediction data holding means. Neuro-model prediction means for predicting future power use for a predetermined time from a predetermined time by using a neural network model with input data as input values, and the use in the same time zone as the time zone predicted by the neuro-model prediction means Based on learning data holding means for holding actual measured power values as learning data, learning data held by the learning data holding means, and predicted values of power used predicted by the neuro model prediction means To learn the neural network model of the neuro model prediction means Preferably includes a neuro model learning means.

  According to this configuration, the power consumption before a predetermined time is stored as prediction data, and the future power consumption is predicted for a predetermined time from the predetermined time by the neural network model using the prediction data as an input value. The actual measured value of the power used in the same time zone as the time zone to be used is stored as learning data, and the neural network model is learned based on the stored learning data and the predicted value of the predicted power usage. .

  Further, in the above fuel cell power generation system, the operation control unit is configured such that when the predicted value of the used power obtained from the used power predicting unit exceeds the predetermined lower limit value continuously for a predetermined time, the predicted value is It is preferable that the start-up operation of the fuel cell power generator is performed from a time obtained by subtracting a time required for starting the fuel cell power generator from a time when the predetermined lower limit value is exceeded.

  According to this configuration, when the predicted value of the used power continuously exceeds the predetermined lower limit value for a predetermined time, the time required for starting the fuel cell power generator from the time when the predicted value exceeds the predetermined lower limit value The starting operation of the fuel cell power generation device is performed from the time of subtracting.

  Further, in the above fuel cell power generation system, the operation control unit is configured such that when the predicted value of the used power obtained from the used power prediction unit is continuously lower than a predetermined lower limit value for a predetermined time, the predicted value is It is preferable to perform the stop operation of the fuel cell power generation device from the time when it falls below a predetermined lower limit value.

  According to this configuration, when the predicted value of the used power is continuously lower than the predetermined lower limit value for a predetermined time, the fuel cell power generator is stopped from the time when the predicted value falls below the predetermined lower limit value. .

  In the above fuel cell power generation system, it is preferable that the power measuring unit measures the upper limit value as the used power when the used power of the electric device exceeds the upper limit value of the generated power.

  According to this configuration, when the power used by the electrical device exceeds the upper limit value of the generated power, the upper limit value is measured as the power used.

  Further, in the fuel cell power generation system described above, based on the hot water supply measuring means for measuring the amount of hot water supplied by a hot water supply device that supplies hot water using the fuel cell power generator, and the amount of hot water measured by the hot water supply measurement means. And a hot water supply amount prediction means for predicting a future hot water supply amount for a predetermined time from a predetermined time, wherein the operation control means predicts the predicted value of the used power and the hot water supply amount predicted by the used power prediction means. It is preferable to control the operation of the fuel cell power generation device based on the predicted value of the hot water supply amount predicted by the prediction means.

  According to this configuration, in the fuel cell power generation system including the fuel cell power generation device that generates the generated power to be supplied to the electric device, the used power used by the electric device is measured, and based on the measured used power, a predetermined power is used. Future power consumption is predicted for a predetermined time from the time, and the amount of hot water supplied by the hot water supply device is measured. Based on the measured amount of hot water, the amount of future hot water supply for the predetermined time from the predetermined time The operation of the fuel cell power generation device is controlled based on the predicted power consumption and the predicted value of the hot water supply amount.

  Further, in the fuel cell power generation system, the hot water supply amount prediction means includes a prediction data holding means for holding the hot water supply amount before the predetermined time as prediction data, and a prediction held by the prediction data holding means. A neuro model predicting means for predicting a future power consumption for a predetermined time from a predetermined time by using a neural network model with the input data as an input value, and the hot water supply in the same time zone as the time zone predicted by the neuro model predicting means Based on learning data holding means for holding actual measurement values as learning data, learning data held by the learning data holding means, and a predicted value of hot water supply predicted by the neuro model prediction means A new neural network model learning means for learning the neural network model Preferably includes a model learning unit.

  According to this configuration, the hot water supply amount before a predetermined time is stored as prediction data, and the future power consumption and hot water supply amount are predicted for a predetermined time from the predetermined time by the neural network model using the prediction data as an input value. The measured value of the hot water supply amount in the same time zone as the predicted time zone is held as learning data, and the neural network model is trained based on the stored learning data and the predicted hot water amount prediction value. Done.

  In the fuel cell power generation system, the operation control unit may be configured to use a current hot water storage system based on a predicted value of used power obtained from the used power prediction unit and a predicted value of hot water supply obtained from the hot water supply amount predicting unit. It is preferable to predict the accumulated hot water amount accumulated from the amount, and adjust the generated power so that the predicted accumulated hot water amount does not exceed a predetermined maximum hot water storage capacity.

  According to this configuration, the accumulated hot water amount that is accumulated from the current hot water storage amount is predicted based on the predicted value of the electric power used and the predicted value of the hot water supply amount, and the predicted hot water storage amount is determined in advance. The generated power is adjusted so as not to exceed the possible amount.

  In the fuel cell power generation system, the operation control unit may be configured to use a current hot water storage system based on a predicted value of used power obtained from the used power prediction unit and a predicted value of hot water supply obtained from the hot water supply amount predicting unit. The accumulated hot water amount accumulated from the amount is predicted, and when the predicted accumulated hot water amount exceeds the predetermined maximum hot water storage amount, a correction is made to slightly reduce the predicted value of the used power, and the corrected used power It is preferable to further predict the accumulated hot water storage amount integrated from the current hot water storage amount based on the predicted value of the hot water supply amount and the predicted value of the hot water supply amount obtained from the hot water supply amount prediction means.

  According to this configuration, the accumulated hot water amount that is accumulated from the current hot water storage amount is predicted based on the predicted value of the electric power used and the predicted value of the hot water supply amount, and the predicted hot water storage amount is determined in advance. When the amount exceeds the possible amount, a correction is made to slightly reduce the predicted value of power usage, and the accumulated hot water amount accumulated from the current hot water storage amount based on the corrected predicted power usage value and predicted hot water supply amount is calculated. Further predicted. When the predicted accumulated hot water storage amount is less than a predetermined maximum hot water storage amount, the operation of the fuel cell power generation device is controlled based on the predicted value of the used electric power finally obtained.

  In the fuel cell power generation system, the operation control unit may be configured to use a current hot water storage system based on a predicted value of used power obtained from the used power prediction unit and a predicted value of hot water supply obtained from the hot water supply amount predicting unit. The amount of accumulated hot water accumulated is estimated from the amount, the predicted accumulated hot water amount exceeds the predetermined maximum hot water storage amount, and the predicted value of the used power obtained from the used power prediction means has a predetermined lower limit value. It is preferable that the fuel cell power generation device is stopped from the time when the maximum hot water storage capacity is exceeded when it continuously falls below a predetermined time.

  According to this configuration, the accumulated hot water amount that is accumulated from the current hot water storage amount is predicted based on the predicted value of the electric power used and the predicted value of the hot water supply amount, and the predicted hot water storage amount is determined in advance. The fuel cell power generation device is stopped from the time when it exceeds the maximum allowable hot water storage amount when the allowable amount is exceeded and the predicted value of the power consumption is continuously below the predetermined lower limit value for a predetermined time.

  In the fuel cell power generation system, the operation control unit may be configured to use a current hot water storage system based on a predicted value of used power obtained from the used power prediction unit and a predicted value of hot water supply obtained from the hot water supply amount predicting unit. The amount of accumulated hot water accumulated is estimated from the amount, the predicted accumulated hot water amount exceeds the predetermined maximum hot water storage amount, and the predicted value of the used power obtained from the used power prediction means has a predetermined lower limit value. When the estimated value exceeds a predetermined lower limit value continuously for a predetermined time, the start operation of the fuel cell power generation device is performed from the time obtained by subtracting the time required for starting the fuel cell power generation device from the time when the predicted value exceeds a predetermined lower limit value. It is preferable.

  According to this configuration, the accumulated hot water amount that is accumulated from the current hot water storage amount is predicted based on the predicted value of the electric power used and the predicted value of the hot water supply amount, and the predicted hot water storage amount is determined in advance. The time required to start the fuel cell power generator from the time when the predicted value exceeds the specified lower limit when the possible amount is exceeded and the predicted value of power used exceeds the specified lower limit continuously for a specified time. The starting operation of the fuel cell power generation device is performed from the time of subtracting.

  A control method for a fuel cell power generation system according to the present invention is a control method for a fuel cell power generation system including a fuel cell power generation device that generates generated power to be supplied to an electrical device, and measures the power used by the electrical device. Based on the power consumption measured in the power measurement step, a power usage prediction step for predicting future power usage for a predetermined time from a predetermined time, and a prediction in the power usage prediction step. And an operation control step for controlling the operation of the fuel cell power generator based on the predicted value of the used power.

  According to this configuration, in the control method of the fuel cell power generation system including the fuel cell power generation device that generates the generated power to be supplied to the electrical equipment, the power usage used by the electrical equipment is measured, and based on the measured power usage Future power consumption is predicted for a predetermined time from a predetermined time, and the operation of the fuel cell power generator is controlled based on the predicted value of the predicted power usage.

  Further, in the control method of the fuel cell power generation system, a hot water supply measuring step for measuring a hot water supply amount supplied by a hot water supply device that supplies hot water using the fuel cell power generation device, and a hot water supply amount measured in the hot water supply measurement step And a hot water supply amount prediction step of predicting a future hot water supply amount for a predetermined time from a predetermined time, and the operation control step includes a predicted value of the used power predicted in the used power prediction step, and It is preferable to include a step of controlling the operation of the fuel cell power generation device based on the predicted value of the hot water supply amount predicted in the hot water supply amount prediction step.

  According to this configuration, in the control method of the fuel cell power generation system including the fuel cell power generation device that generates the generated power to be supplied to the electric device, the amount of hot water supplied by the hot water supply device is measured, and the measured amount of hot water is based on the measured amount of hot water. A future hot water supply amount is predicted for a predetermined time from a predetermined time, and the operation of the fuel cell power generation device is controlled based on the predicted use power and the predicted value of the hot water supply amount.

  According to the fuel cell power generation system and the control method thereof of the present invention, the power used by the electrical equipment is measured, and the future power consumption is predicted for a predetermined time from a predetermined time based on the measured power used. Since the operation of the fuel cell power generation device is controlled based on the predicted value of the predicted power consumption, it is possible to efficiently control the power generation power of the fuel cell power generation device by predicting different power usage for each home. And energy saving can be realized.

  Embodiments according to the present invention will be described below with reference to the drawings. In addition, about the same structure in each figure, the same code | symbol is attached | subjected and the description is abbreviate | omitted.

(First embodiment)
FIG. 1 is a diagram showing an overall configuration of a fuel cell power generation system according to a first embodiment of the present invention. A fuel cell power generation system 1 shown in FIG. 1 includes a fuel cell power generation device 100, a power generation control unit (controller) 101, an inverter 102, and a wattmeter 103, and is connected to a home appliance 104 and a commercial power source 105.

  The fuel cell power generation device 100, the inverter 102, the home electric appliance 104, which is an example of an electric device, and the commercial power source 105 are connected to a home power system. The wattmeter 103 measures the power used by the home appliance 104. The wattmeter 103 is connected to the controller 101, and power used by the home appliance 104 is sent from the wattmeter 103 to the controller 101. The home appliance 104 is an electric device used at home, such as a refrigerator or a washing machine. Note that the number of home appliances connected to the wattmeter 103 may be not only one but also a plurality.

  The controller 101 is connected to the fuel cell power generation apparatus 100, and a start / stop command or the like is output from the controller 101 to the fuel cell power generation apparatus 100 to control the operation of the fuel cell power generation apparatus 100.

  The fuel cell power generation apparatus 100 generates hydrogen by converting chemical energy into electrical energy by reacting hydrogen obtained from fuel such as city gas with oxygen in the air. The power generated by the fuel cell power generation apparatus 100 is supplied to the inverter 102, and is supplied from the inverter 102 to the home appliance 104. When the power used by the home appliance 104 is greater than the generated power, the inverter 102 purchases (purchases) power from the commercial power source 105 to compensate for the shortage. Conversely, when the generated power is greater than the used power, the inverter 102 sells (sells) the surplus generated power to the commercial power source 105. When the commercial power source 105 does not permit power sale, the surplus is processed by the main body of the fuel cell power generation apparatus 100. In order to improve the efficiency of the fuel cell power generation apparatus 100, it is better not to purchase and sell power as much as possible. Therefore, the controller 101 needs to send a start / stop instruction appropriately so as not to purchase and sell power as much as possible from the power used by the home appliance 104 measured by the wattmeter 103.

  FIG. 2 is a block diagram showing a configuration of the controller 101 shown in FIG. The controller 101 shown in FIG. 2 includes a power usage prediction unit 200, a power generation command unit (corresponding to an example of an operation control unit) 220, and a power measurement unit 230.

  The power measurement unit 230 measures the power used by the home appliance 104. The power usage prediction unit 200 predicts future power usage for a predetermined time from a predetermined time based on the power usage measured by the power measurement unit 230. The neuro model prediction unit 202, the neuro model learning A unit 203, a prediction data holding unit 204, and a learning data holding unit 205.

  The neuro model prediction unit 202 holds a hierarchical neural network model, and predicts future power usage for a predetermined time from a predetermined time by the neural network model. Details of the features and learning methods of the neural network model are disclosed in “Shunichi Amari, New Development of Neural Networks, pp. 73-86, Science Co., Ltd., 1994”. Is omitted. In the present embodiment, the power usage prediction unit 200 predicts the power usage on the day of prediction up to 24 hours ahead in units of one hour.

  FIG. 3 is a diagram for explaining a configuration of a neural network model used in the neuro model prediction unit 202 of FIG. The neural network model 300 is a hierarchical neural network and has three layers: an input layer, an intermediate layer, and an output layer. As a configuration of the neural network model 300, it is necessary to improve prediction accuracy by using a predicted value as an output parameter and configuring data having a strong causal relationship with the predicted value as an input parameter. For this reason, the output parameter is the predicted power usage value for the day, and the input parameter is the power consumption for the previous day, which is considered to have a strong causal relationship with the predicted value.

  In the present embodiment, since prediction is performed in units of one hour until 24 hours ahead, the output parameters of the neural network model 300 are “predicted power usage value of 0:00 on the current day” and “predicted power consumption value of the current time of 1:00”. ”, ..., 24 data of“ predicted power consumption at 23 o'clock for the day ”and input parameters are“ usage of power at 0 o'clock the previous day ”,“ power use at 1 o'clock the previous day ” ,..., 24 data of “power used at 23 o'clock the previous day” are used. The power used in the midnight range is an average of the power used from 0:00 to 1:00. By configuring the neural network model in this way, when the past day of 0:00 to be predicted is input, by inputting the previous day's power consumption, the power usage for the day is predicted in units of one hour (24 hours worth). Prediction).

  In the neural network model 300, the weighting coefficient of the neural network model 300 is corrected by learning the difference (error) between prediction and actual measurement by the reverse error propagation method (back propagation) in order to increase the accuracy of prediction.

  The prediction data holding unit 204 acquires the used power from the power measuring unit 230 and holds the used power for one hour unit for the past 24 hours. Then, when the date is updated, it is sent to the neuro model prediction unit 202 as the power used on the previous day, which becomes the input of the neural network model 300. The neuro model prediction unit 202 can output the power usage predicted value for 24 hours on the current day by inputting the sent power usage of the previous day to the neural network model 300, and the power usage predicted value is the power generation command unit. 220, the operation of starting / stopping the fuel cell power generation apparatus 100 can be determined.

  In the neural network model 300, in order to improve the prediction accuracy, the difference (error) between the predicted power consumption and the actual power consumption for the time corresponding to the predicted power consumption (hereinafter referred to as actual power consumption) is reversed. It is necessary to train by the error propagation method. In order to perform learning for actually improving accuracy, the power consumption prediction value and the power consumption measurement value for the past several days are required. Therefore, the learning data holding unit 205 holds the power usage predicted values for 24 hours output from the neuro model prediction unit 202 and the power usage measured values corresponding to the respective time zones for the past several days, They are used as learning data. The learning data holding unit 205 holds, from the power measuring unit 230, the used power actual value that will be obtained at the time when one day that is the target of the used power predicted value has passed, as the used power in units of one hour. When the measured power usage values are collected for 24 hours, they are output to the neuro model learning unit 203. By performing such an operation for several days, the neuro model learning unit 203 can secure a pair of data of predicted power usage and actual power usage for several days. The weight coefficient of the neural network model 300 of the model prediction unit 202 can be corrected, and as a result, prediction for each household can be performed with high accuracy.

  In order to improve accuracy, it is better to divide the data used for learning. For example, in the case of predicting weekdays, it is more effective to use weekday data as data used for learning.

  In an initial state in which the neural network model 300 is not learned at all, it is necessary to make predictions after at least one learning has been performed and data necessary for learning can be secured for several days.

  The predicted power usage values for 24 hours from the 0 o'clock range to the 23 o'clock range output from the neuro model prediction unit 202 of the power usage prediction unit 200 are output to the power generation command unit 220. The power generation command unit 220 determines start / stop so that the fuel cell power generation apparatus 100 can be efficiently operated from the predicted value, and outputs a start instruction or a stop instruction to the fuel cell power generation apparatus 100 based on the determination result.

  Next, the operation determination of the fuel cell power generation device 100 based on the predicted used power value output from the used power prediction unit 200 will be described.

  In general, the variable range of the generated power of the fuel cell power generation apparatus 100 is set to an upper limit value and a lower limit value from the viewpoint of performance and efficiency. For example, in the fuel cell power generation device 100 used at home, the upper limit value is set to 1 kW, the lower limit value is set to 0.5 kW, and the generated power follows the used power if the used power is within the upper and lower limit values. In other words, electric main operation is possible.

  However, if the start / stop determination is simply made when the power used falls within the range of the upper and lower limit values, there is a very problem in terms of efficiency. This is because, when the fuel cell power generation apparatus 100 is started up, a rise loss such as pump power or residual heat occurs while the generated power gradually rises. Therefore, frequent starting / stopping is not preferable in terms of efficiency due to frequent loss. Generally, it is said that it is necessary to operate for 2 to 3 hours once it is started. From the above, the power generation command unit 220 determines start / stop from a predicted value of power used for a predetermined time, for example, 3 hours.

  FIG. 4 is a flowchart for explaining the operation of the power generation command unit 220 of FIG. When a predicted power usage value for 24 hours is input from the power usage prediction unit 200 at 0:00 on the day of operation (step S1), the power generation command unit 220 uses the power usage prediction value for 24 hours to make a start determination. The first time Tstart when the preset lower limit value is continuously exceeded for 3 hours is searched (step S2). Assuming that the time required to start up the fuel cell power generation apparatus 100 is ΔT, the power generation command unit 220 outputs a command for starting up at (Tstart−ΔT) (step S3). In addition, in order to perform the stop determination, the power generation command unit 220 searches for the first time Tstop when the predicted lower limit value for 24 hours falls below the preset lower limit value for 3 hours continuously (step S4). . Then, the power generation command unit 220 outputs a stop command to the fuel cell power generation apparatus 100 at the time of Tstop (step S5).

  FIG. 5 is an explanatory diagram showing an example of the relationship between the power generated by the fuel cell power generation apparatus 100 and the predicted power usage when the series of operations described above is performed. In FIG. 5, the vertical axis represents power usage, and the horizontal axis represents time. In addition, the predicted power usage from 8:00 to 16:00 and the power generated by the fuel cell power generation apparatus 100 are omitted.

  In the activation determination, a predicted power usage value from 0 to 23:00 is searched, and a time zone that exceeds the lower limit value of the generated power for three hours is searched. In this example shown in FIG. 5, since the time zone exceeding the lower limit value of the generated power for 3 consecutive hours starts at 5 o'clock, Tstart = 5 o'clock, and the start command by the power generation command unit 220 is (5-ΔT ) Is output. The fuel cell power generation device 100 that has received the start command starts and starts generating power. At this time, since the fuel cell power generation apparatus 100 is started earlier by the time ΔT required for starting the fuel cell power generation apparatus 100, it is possible to supply generated power equivalent to the predicted power consumption at 5 o'clock. In addition, since it is obtained in advance from the predicted power consumption that power consumption has been generated for at least 3 hours after Tstart = 5 o'clock, there is no surplus in generated power and no need to stop for at least 3 hours after startup. . Therefore, more efficient start-up operation is possible.

  Similarly, in the stop determination, a predicted power usage value from 0 to 23:00 is searched, and a time zone that is lower than the lower limit value of the generated power is searched continuously for 3 hours. In this example shown in FIG. 5, Tstop = 20 o'clock because the time zone that falls below the lower limit of the generated power for 3 hours continuously starts at 20:00, and the stop command by the power generation command unit 220 is output at 20:00. The The fuel cell power generation device 100 that has received the stop command is stopped, but since Tstop = 20 o'clock, the power consumption is below the lower limit for at least 3 hours, because it is obtained in advance from the predicted power consumption value. It is no longer necessary to start up again for at least 3 hours later. Therefore, more efficient stop operation is possible.

  FIG. 6 is a diagram for explaining the power consumption measured by the power measurement unit 230 of FIG. Note that in FIG. 6, the vertical axis represents power used by home appliances, and the horizontal axis represents time. In FIG. 6, the change of the used electric power for 1 hour between 0 o'clock and 24:00 is represented.

  As described above, the power consumption of the home appliance 104 is detected by the wattmeter 103, and the detected value is acquired by the power measurement unit 230 and input to the power consumption prediction unit 200. In this case, as shown in FIG. 6, the power captured by the power generated by the fuel cell power generation apparatus 100 is the power up to the upper limit value of the generated power. Therefore, the electric power actually required for the generated electric power is indicated by the hatched portion in FIG. 6, and the electric power exceeding the upper limit is not relevant. Therefore, as shown in FIG. 6, the power measurement unit 230 acquires a portion that does not exceed the upper limit value of the generated power (shaded portion in FIG. 6) as the used power, and outputs it to the used power prediction unit 200.

  In this way, when the power used by the home appliance 104 exceeds the upper limit value of the generated power, the upper limit value is measured as the used power, so that the power rises sharply in a short time, such as a hair dryer or a microwave oven. Thus, the power used that is not actually captured by the fuel cell power generation device 100 can be excluded, and as a result, the power used by the actual fuel cell power generation device 100 can be predicted. A stop determination can be made.

  Here, the operation of the control system of the fuel cell power generation apparatus 100 that predicts the power consumption of the home appliance 104 and determines the operation of the fuel cell power generation apparatus 100 will be described.

  When it is determined by a timer (not shown) built in the controller 101 that the prediction start time has come, the prediction data holding unit 204 outputs the prediction data to the neuro model prediction unit 202. In this embodiment, power consumption for 24 hours from 0:00 to 23:00 is predicted in units of one hour. Therefore, when 0:00 is reached, the prediction data holding unit 204 sets each time from 0:00 to 23:00. The power used in the band is output to the neuro model prediction unit 202 as prediction data.

  The neuro model prediction unit 202 predicts the future power usage of the fuel cell power generation apparatus 100 for 24 hours from 0:00 to 23:00 using a neural network model. The predicted value of the power used in each time zone from 0 o'clock to 23 o'clock predicted by the neuro model prediction unit 202 is output to the power generation command unit 220.

  The power generation command unit 220 determines the operation of the fuel cell power generation apparatus 100. In other words, the power generation command unit 220 indicates that the predicted value of power used in each time zone from the 0 o'clock range to the 23 o'clock range output from the neuro model prediction unit 202 exceeds the lower limit set in advance for 3 hours. The first time Tstart is searched.

  In addition, the power generation command unit 220 indicates that the predicted value of power used in each time zone from the 0 o'clock range to the 23 o'clock range output from the neuro model prediction unit 202 falls below a preset lower limit value for 3 hours continuously. The first time Tstop is searched.

  The time is counted by a timer (not shown) built in the controller 101, and the power generation command unit 220 subtracts the time ΔT necessary for starting the fuel cell power generation device 100 from the time Tstart, and then the fuel cell power generation Instruct the device 100 to start up. Further, the power generation command unit 220 instructs the fuel cell power generation device 100 to stop when the time Tstop is reached.

  As described above, according to the present invention, the used power prediction unit 200 can predict the used power different for each home for 24 hours by learning the neural network model 300 from the measured used power value of the home. Since the start and stop determinations can be made so that frequent start / stop is not performed based on the power used predicted in step 1, the fuel cell power generation apparatus 100 can be operated efficiently.

  In the present invention, the 24 hours of the day is predicted in units of one hour as power consumption prediction, but the number of hours and the number of units are not limited to 24 hours or one hour, and the neural network model 300 is not limited. Depending on how the input parameters and output parameters are taken, it is possible to easily configure, for example, prediction of power consumption for 6 hours in 30 minute units.

  Further, in the present invention, the electric power used is an object of prediction, but if the amount of hot water used is detected, the hot water supply can be predicted with the same configuration. This example will be described in detail in the second embodiment.

  In this way, the power usage used by the home appliance 104 is measured, and based on the measured power usage, the future power usage is predicted for a predetermined time from a predetermined time, and the predicted value of the predicted power usage is predicted. Therefore, it is possible to efficiently control the power generated by the fuel cell power generation apparatus 100 by predicting the power used differently for each household, thereby reducing energy consumption. Can be realized.

  In addition, the power consumption before a predetermined time is stored as prediction data, and the future power consumption is predicted for a predetermined time from the predetermined time by the neural network model 300 using the prediction data as an input value. The actual measured value of power used in the same time zone as the band is stored as learning data, and the neural network model 300 is learned based on the stored learning data and the predicted value of power consumption to be predicted. Therefore, it is possible to more accurately predict the power usage that varies from home to home.

(Second Embodiment)
Next, a fuel cell power generation system according to a second embodiment of the present invention will be described. In the fuel cell power generation system according to the first embodiment, based on the used power used by the home appliance 104, future used power is predicted for a predetermined time from a predetermined time, and the predicted value of the used power is calculated. Based on this, the start / stop of the fuel cell power generation apparatus 100 is determined. On the other hand, the fuel cell power generation system according to the second embodiment measures the amount of hot water supplied by a hot water supply device that supplies hot water using the fuel cell power generation apparatus 100, and determines a predetermined amount based on the measured amount of hot water. The amount of hot water supply in the future is predicted for a predetermined time from the time, and the operation of the fuel cell power generation apparatus 100 is controlled based on the predicted value of power consumption and the predicted value of the hot water supply amount.

  FIG. 7 is a diagram showing an overall configuration of a fuel cell power generation system according to the second embodiment of the present invention. The fuel cell power generation system 2 in the second embodiment shown in FIG. 7 includes a fuel cell power generation device 100, a power generation control unit (controller) 101, an inverter 102, a power meter 103, a hot water storage tank 106, a backup burner 107, and a hot water supply load meter 109. And connected to the home appliance 104, the commercial power source 105, the hot water supply device 108, and the city water supply unit 110.

  The fuel cell power generation device 100, the inverter 102, the home appliance 104, and the commercial power source 105 are connected to a home power system. The wattmeter 103 measures the power used by the home appliance 104. The hot water storage tank 106, the backup burner 107, the hot water supply device 108, and the city water supply unit 110 are connected to a hot water supply system in the home. The hot water supply device 108 is a device used when a resident uses hot water such as a bath, a shower, and a bathroom. The hot water supply load meter 109 measures the amount of hot water supplied to the hot water supply device 108, the hot water supply temperature, and the temperature of city water supplied from the city water supply unit 110, and measures the amount of heat of the hot water supply load used by the hot water supply device 108. Hereinafter, the amount of heat of the hot water supply load is referred to as used hot water supply heat amount.

  The wattmeter 103 and the hot water supply load meter 109 are connected to the controller 101, and the electric power used by the home appliance 104 is sent from the wattmeter 103 and the used hot water supply amount of the hot water supply device 108 is sent from the hot water supply load meter 109. The controller 101 is connected to the fuel cell power generation device 100, and a power generation command is output from the controller 101 to the fuel cell power generation device 100. The fuel cell power generation device 100 generates power according to the power generation command.

  The fuel cell power generation apparatus 100 generates hydrogen by converting chemical energy into electrical energy by reacting hydrogen obtained from fuel such as city gas with oxygen in the air. The power generated by the fuel cell power generation apparatus 100 is supplied to the inverter 102, and is supplied from the inverter 102 to the home appliance 104. When the power used by the home appliance 104 is greater than the generated power, the inverter 102 purchases (purchases) power from the commercial power source 105 to compensate for the shortage. Conversely, when the generated power is larger than the used power, the inverter 102 sells (sells) the surplus power generated to the commercial power source 105. When the commercial power source 105 does not permit power sale, the surplus is processed by the main body of the fuel cell power generation apparatus 100.

  The fuel cell power generation apparatus 100 generates heat simultaneously with power generation. The generated heat is stored as hot water in the hot water storage tank 106 as power generation hot water supply heat. Hot water stored in the hot water storage tank 106 is discharged from the hot water supply device 108 in response to the user's request. At this time, if there is no desired hot water in the hot water storage tank 106, the backup burner 107 generates hot water and provides it to the hot water supply device 108. Reasons for running out of hot water in the hot water storage tank 106 include a case where the electric power used by the home appliance 104 is small and the electric power generated by the fuel cell power generation apparatus 100 is small, or the amount of hot water used by the hot water supply device 108 is very large. Conceivable. On the contrary, when the electric power used by the home appliance 104 is very large and the generated power of the fuel cell power generation apparatus 100 is large, or when the amount of hot water used by the hot water supply device 108 is very small, the generated hot water is relatively The hot water storage tank 106 may become full. In this case, the fuel cell power generation device 100 must be completely stopped so that the generated hot water supply heat is dissipated to the outside or no further generated hot water supply heat is generated. Will do.

  Therefore, in order to increase the efficiency of the fuel cell power generation apparatus 100, it is necessary to control the power supply hot water supply heat so that the hot water in the hot water storage tank 106 is not filled as much as possible, that is, to send a power generation command for appropriately adjusting the generated power. is there.

  FIG. 8 is a block diagram showing a configuration of the controller 101 shown in FIG. The controller 101 shown in FIG. 8 includes a power usage prediction unit 200, a hot water supply amount prediction unit 240, a power generation command unit 220, a power measurement unit 230, and a hot water supply amount measurement unit 250.

  First, the power used will be described. The power measurement unit 230 measures the power used by the home appliance 104. The power usage prediction unit 200 predicts future power usage for a predetermined time from a predetermined time based on the power usage measured by the power measurement unit 230. The neuro model prediction unit 202, the neuro model learning A unit 203, a prediction data holding unit 204, and a learning data holding unit 205.

  The neuro model prediction unit 202 holds a hierarchical neural network model, and predicts future power usage for a predetermined time from a predetermined time by the neural network model. Details of the features and learning methods of the neural network model are disclosed in “Shunichi Amari, New Development of Neural Networks, pp. 73-86, Science Co., Ltd., 1994”. Is omitted. In the present embodiment, the power usage predicting unit 200 predicts power usage up to 24 hours after the predicted time in units of one hour.

  FIG. 9 is a diagram for explaining a configuration of a neural network model used in the neuro model prediction unit 202 of FIG. The neural network model 310 is a hierarchical neural network and has three layers: an input layer, an intermediate layer, and an output layer. As the configuration of the neural network model 310, it is necessary to improve the prediction accuracy by configuring the predicted value as an output parameter and configuring data having a strong causal relationship with the predicted value as an input parameter. For this reason, the output parameter is the predicted power usage value for the day, and the input parameter is the power consumption for the previous day, which is considered to have a strong causal relationship with the predicted value.

  In this embodiment, since prediction is performed in units of one hour from the predicted time to 24 hours ahead, the output parameters of the neural network model 310 are “predicted power usage value of predicted time base”, “(predicted time + 1)”. 24 data of “predicted power usage value”,..., “(Predicted time + 23) predicted power usage value”, and the input parameter is “the same power usage the previous day as the predicted time base” , “Used power of the same day as (predicted time + 1) units”,..., “Used power of the same day as (predicted time + 23) units” is used. Note that the predicted power consumption of the time base is an average of the power used from 0:00 to 1:00 when the predicted time is 0:00. By configuring the neural network model in this way, when the past day of 0:00 to be predicted is input, the previous day's used power is input, so that the used power after the predicted time of the day is predicted in units of one hour ( Prediction for 24 hours).

  In the neural network model 310, the weighting coefficient of the neural network model 310 is corrected by learning the difference (error) between the prediction and the actual measurement by the reverse error propagation method (back propagation) in order to increase the accuracy of the prediction.

  The prediction data holding unit 204 acquires the used power from the power measuring unit 230 and holds the used power for one hour unit for the past 24 hours. Then, when the date is updated, it is sent to the neuro model prediction unit 202 as the power used on the previous day, which becomes the input of the neural network model 310. The neuro model prediction unit 202 can output the power consumption predicted value for 24 hours after the predicted time by inputting the transmitted power consumption of the previous day to the neural network model 310. It is sent to the command unit 220.

  In the neural network model 310, in order to improve the prediction accuracy, the difference (error) between the predicted power usage value and the actual power usage for the time corresponding to the power usage prediction value (hereinafter referred to as power consumption actual measurement value) is reversed. It is necessary to train by the error propagation method. In order to perform learning for actually improving accuracy, the power consumption prediction value and the power consumption measurement value for the past several days are required. Therefore, the learning data holding unit 205 holds the power usage predicted values for 24 hours output from the neuro model prediction unit 202 and the power usage measured values corresponding to the respective time zones for the past several days, They are used as learning data. The learning data holding unit 205 uses the power measurement unit 230 to calculate the power usage value that is obtained for the past 24 hours when the predicted time that is the target of the power usage prediction value has passed. And is output to the neuro model learning unit 203 at the time when the measured power usage values are aligned for 24 hours. By performing such an operation for several days, the neuro model learning unit 203 can secure data in which the predicted power usage value and the actual power usage for several days are paired. The weighting coefficient of the neural network model 310 of the prediction unit 202 can be corrected, and as a result, prediction for each home can be performed with high accuracy.

  In order to improve accuracy, it is better to divide the data used for learning. For example, in the case of predicting weekdays, it is more effective to use weekday data as data used for learning.

  In the initial state in which the neural network model 310 is not learned at all, it is necessary to make prediction after at least one learning is performed since data necessary for learning can be secured for several days.

  The predicted power usage value for the previous 24 hours from the prediction time output from the neuro model prediction unit 202 of the power usage prediction unit 200 is output to the power generation command unit 220.

  Next, hot water supply will be described. The hot water supply measuring unit 250 measures the amount of hot water used by the hot water supply device 108. The hot water supply amount prediction unit 240 predicts a future hot water supply amount for a predetermined time from a predetermined time based on the hot water supply amount measured by the hot water supply amount measurement unit 250. The neuro model prediction unit 242 and the neuro model A learning unit 243, a prediction data holding unit 244, and a learning data holding unit 245 are provided.

  The operation of the hot water supply amount prediction unit 240 is basically the same as that of the power consumption prediction unit 200, but the neural network model used in the neuro model prediction unit 242 is as shown in FIG. The neural network model 320 is a hierarchical neural network and has three layers: an input layer, an intermediate layer, and an output layer. As the configuration of the neural network model 320, since prediction is performed in units of one hour from the predicted time to 24 hours ahead, the output parameters of the neural network model 320 are “predicted value of hot water supply amount for predicted time base”, “(predicted). +1) Predicted hot water supply amount of units ", ...," Predicted time +23) Predicted hot water supply amount of units ", and the input parameters were the same day as the predicted time base 24 data of “the amount of hot water supply of the same day as the (predicted time + 1) stand”,..., “The amount of hot water supply of the same day as the (predicted time + 23) stand” are used. Note that the amount of hot water supply in the time base to be predicted is the average of the amount of hot water used from 0 o'clock to 1 o'clock when the predicted time is 0 o'clock. By configuring the neural network model in such a manner, when the past day of 0:00 to be predicted is passed, the amount of hot water supply after the predicted time of the day is predicted in units of one hour by inputting the amount of hot water supply of the previous day ( Prediction for 24 hours).

  In this way, the hot water supply amount before a predetermined time is stored as prediction data, and the future hot water supply amount is predicted and predicted for a predetermined time from the predetermined time by the neural network model 320 using the prediction data as an input value. The actual measured value of the hot water supply amount in the same time zone as the learning time zone is stored as learning data, and the neural network model 320 is learned based on the stored learning data and the predicted predicted hot water supply amount. . Therefore, it is possible to predict the amount of hot water supply that varies from home to home more accurately.

  When the operations of the neuro model prediction unit 242, the neuro model learning unit 243, the prediction data holding unit 244, and the learning data holding unit 245 are performed in the same manner as the power usage prediction unit 200 with the above configuration, prediction is performed from the neuro model prediction unit 242. The predicted hot water supply amount for the previous 24 hours from the time is output and output to the power generation command unit 220.

  The power generation command unit 220 adjusts the generated power so that the fuel cell power generation apparatus 100 can be efficiently operated from the predicted power usage value and the predicted hot water supply amount.

  Next, the operation of the power generation command unit 220 will be described. FIG. 11 is a flowchart for explaining the operation of the power generation command unit 220 of FIG. Since the predicted time is in units of one hour, first, when the time is updated, it is determined whether or not the predicted value can be updated (step S11). If it is determined that the predicted value cannot be updated (NO in step S11), the process in step S11 is repeated. If it is determined that the predicted value can be updated (YES in step S11), first, the current hot water storage amount Qnow [kWh] of the hot water storage tank 106 is acquired from the hot water storage tank 106 (step S12). A general method for calculating the current amount of stored hot water Qnow [kWh] can be obtained by attaching a temperature sensor or the like to the hot water storage tank 106 and measuring the temperature distribution of the hot water remaining in the tank.

Next, a predicted power usage value Pgene (i) [kWh] (i = 1 to 24) and a predicted hot water supply amount Photo (i) [kWh] (i = 1) from the power usage prediction unit 200 and the hot water supply amount prediction unit 240, respectively. To 24) are acquired in increments of one hour until 24 hours ahead (step S13). Next, the process enters a step of calculating the generated power command value Pprof (i) [kWh]. First, the power usage predicted value Pgene (i) [kWh] is set as the initial value of the generated power command value Pprof (i) [kWh] (step S14). This corresponds to the case where the main operation is performed completely up to 24 hours ahead. Next, the generated hot water supply load Qgene (i) [kWh], which is the amount of heat additionally generated when the fuel cell power generation apparatus 100 is generated by the generated power command value Pprof (i) [kWh], is calculated. This is calculated using the following equation (1) (step S15).
Qgene (i) [kWh] = Pprof (i) [kWh] × (hot-water supply efficiency [%] / power generation efficiency [%]) (1)

Next, the predicted integrated hot water storage amount Qadd (i) [kWh], which is the amount of heat that is added to or subtracted from the hot water storage tank up to 24 hours ahead of the current time, can be used as the power generation hot water supply load Qgene (i) [kWh] and the predicted hot water supply amount Phot. (I) From [kWh], the following equation (2) is used for calculation (step S16).
Qadd (i) = Qgene (i) −Photo (i) + Qadd (i−1) (2)

Next, as shown in the following equation (3), it is determined whether or not the value obtained by adding the current hot water storage amount Qnow [kWh] and the predicted integrated hot water storage amount Qadd (i) [kWh] is smaller than 0 ( Step S17). This is because there is an error in the predicted electric power usage value Pgene (i) [kWh] and the predicted hot water supply amount Photo (i) [kWh], and the predicted hot water storage heat quantity Qnow + Qadd (i) of the hot water storage tank is negative in terms of the numerical value. This is to prevent the case. If it is determined that the value obtained by adding the current hot water storage amount Qnow [kWh] and the predicted cumulative hot water storage amount Qadd (i) [kWh] is 0 or more (NO in step S17), the process proceeds to step S19. To do. On the other hand, if it is determined that the value obtained by adding the current hot water storage amount Qnow [kWh] and the predicted integrated hot water storage amount Qadd (i) [kWh] is smaller than 0 (YES in step S17), the predicted integrated hot water storage amount (Qadd ( i) [kWh]) is defined as the following equation (4) (step S18).
Qnow + Qadd (i) <0 (3)
Qadd (i) = − Qnow (4)

  Next, the power generation command unit 220 determines whether or not the value obtained by adding the current hot water storage heat amount Qnow [kWh] and the predicted integrated hot water storage heat amount Qadd (i) [kWh] is smaller than 0 until i = 24. Repeat (steps S19 and S20).

  Next, a time i at which the predicted integrated hot water storage heat amount Qadd (i) [kWh] exceeds the hot water storage heat amount Qmax−Qnow is calculated (steps S21, 22, 23). Here, Qmax [kWh] is the maximum amount of heat that can be stored in hot water, and is a fixed value that depends on the size of the hot water storage tank 106. If the predicted integrated hot water storage heat amount Qadd (i) [kWh] exceeds the hot water storage possible heat amount Qmax-Qnow, it means that the hot water storage tank 106 becomes full at time i. That is, the power generation command unit 220 determines whether or not the predicted integrated hot water storage amount Qadd (i) is larger than a value obtained by subtracting the current hot water storage heat amount Qnow from the maximum hot water storage heat amount Qmax (hot water storage heat amount). If the predicted integrated hot water storage amount Qadd (i) is larger than the hot water storage possible heat amount Qmax-Qnow (YES in step S21), the process proceeds to step S24, and the predicted integrated hot water storage heat amount Qadd (i) is stored in the hot water storage possible heat amount Qmax- If it is equal to or less than Qnow (NO in step S21), it is determined whether i = 24. If i is not 24 (NO in step S22), i is incremented by 1 (step S23), and the process returns to step S21.

Here, when there is a time i at which the predicted accumulated hot water storage amount Qadd (i) exceeds the hot water storage possible heat amount Qmax-Qnow, the generated power command value Pprof (i) [kWh] is corrected using the following equation (5). (Steps S24, 25, 26). After the correction, the process returns to step S15. If there is no time i at which the predicted accumulated hot water storage amount Qadd (i) exceeds the hot water storage possible heat amount Qmax-Qnow (YES in step S22), the process returns to waiting for the predicted time update (step S11), and the next update time is set. wait.
Pprof (j) = Pprof (j) −ΔPprof (5)

  In the above equation (5), j = i, and ΔPprof is a change step for correcting the generated power command value Pprof (i) [kWh], and is generally set to a sufficiently small value (fixed value). Keep it.

  Through the series of operations described above, the generated power command value Pprof (i) [kWh] is corrected so as not to fill the hot water storage tank 106 and is finally sent to the fuel cell power generator 100. The fuel cell power generation apparatus 100 generates power so that the generated power command value Pprof (i) [kWh] is output from the power generation command unit 220.

  FIG. 12 is an explanatory diagram showing an example of the relationship between the generated power of the fuel cell power generation system and the amount of stored hot water. FIG. 12 shows the predicted power usage value Pgene (i) [kWh], the hot water supply amount prediction value Phot (i) [kWh], the generated power command value Pprof (i) [kWh], and the generated hot water supply load Qgene (i). ) [KWh], predicted accumulated hot water storage amount Qadd (i) [kWh] Current hot water storage heat amount Qnow (i) [kWh], maximum hot water storage possible heat amount Qmax [kWh].

  If an operation in which the generated power command value Pprof (i) [kWh] is always equal to the predicted power usage value Pgene (i) [kWh], that is, a so-called main operation is performed, the predicted integrated hot water storage amount Qadd (i) [kWh] As indicated by “Before correction”, for example, in Qadd (10) after 10 hours, the maximum hot water storage heat capacity Qmax [kWh] of the hot water storage tank 106 is exceeded, and the fuel cell power generation device 100 is stopped or the amount of heat of the tank is increased. It is necessary to dissipate part of the heat, resulting in loss. Therefore, in this case, by repeating steps S11 to S27, the generated power command value Pprof (i) [kWh] is corrected, and the predicted integrated hot water storage heat amount Qadd (i) [kWh] is finally set to “after correction”. It can be modified as shown so as not to exceed the maximum heat storage capacity Qmax [kWh]. Further, by performing the operations of steps S11 to S27 every hour, it is possible to always operate until the predicted accumulated hot water storage heat amount Qadd (i) [kWh] does not exceed the maximum hot water storage heat amount Qmax [kWh] until 24 hours ahead. Therefore, it is possible to stop the fuel cell power generation apparatus 100, or to suppress the generation of loss due to heat dissipation for a part of the tank heat amount, and to perform efficient operation.

  As described above, according to the present invention, the power generation command unit 220 generates power so that the hot water in the hot water storage tank 106 does not become full based on the predicted power usage value from the power usage prediction unit 200 and the hot water supply amount prediction value from the hot water supply amount prediction unit 240. Since the command power value can be obtained, it is possible to operate with energy saving even in each home in different usage environments where power and hot water supply are different.

  In addition, the power used by the home appliance 104 and the amount of hot water supplied by the hot water supply device 108 are measured, and based on the measured used power and hot water supply, the future used power and hot water for a predetermined time from a predetermined time. The amount of power used is predicted, and the power generation power of the fuel cell power generation apparatus 100 is adjusted based on the predicted value of the predicted power usage and the predicted value of the hot water supply amount. The power generated by the fuel cell power generation apparatus 100 can be controlled efficiently, and the hot water fills up the hot water storage tank 106 to dissipate the heat to the outside or the fuel cell power generation apparatus 100 must be completely stopped and restarted. As a result, there will be no rise loss due to, and energy saving can be realized.

  Furthermore, the accumulated hot water amount accumulated from the current hot water storage amount is predicted based on the predicted value of the electric power used and the predicted value of the hot water supply amount, and the predicted accumulated hot water amount exceeds the predetermined maximum hot water storage amount. In this case, a correction for slightly reducing the predicted value of the used power is performed, and an accumulated hot water amount accumulated from the current hot water storage amount is further predicted based on the corrected predicted value of the used power and the predicted value of the hot water supply amount. . When the predicted accumulated hot water storage amount is below a predetermined maximum hot water storage amount, the operation of the fuel cell power generation apparatus 100 is controlled based on the predicted value of the used power finally obtained. Therefore, it is possible to realize energy saving by adjusting the generated power without completely stopping the fuel cell power generation apparatus 100.

  Although the fuel cell power generation system 2 of the present invention includes the power meter 103 and the hot water supply load meter 109, the present invention is not particularly limited to this, and the power meter 103 and / or the hot water supply load meter 109 is used as the fuel. Needless to say, a desired effect can be obtained even if the battery power generation system 2 is provided outside the battery power generation system 2 to acquire only data.

  In the present embodiment, hot water is prevented from filling the hot water storage tank by adjusting the generated power command value while starting the fuel cell power generation device 100 without stopping the fuel cell power generation device 100. doing. However, the present invention is not particularly limited to this, and the power generation command unit 220 is based on the predicted value of the used power obtained from the used power prediction unit 200 and the predicted value of the hot water amount obtained from the hot water supply amount prediction unit 240. The accumulated hot water amount accumulated from the current hot water storage amount is predicted, the predicted accumulated hot water amount exceeds the predetermined maximum hot water storage amount, and the predicted value of the used power obtained from the used power predicting means is a predetermined value. When the lower limit value is continuously decreased for a predetermined time, the fuel cell power generation device 100 may be stopped from the time when the maximum hot water storage capacity is exceeded.

  Moreover, the power generation command unit 220 integrates the current hot water storage amount based on the predicted value of the used power obtained from the used power prediction unit 200 and the predicted value of the hot water supply amount obtained from the hot water supply amount prediction unit 240. Predicting the amount, the predicted accumulated hot water storage amount exceeds a predetermined maximum hot water storage amount, and the predicted value of the power usage obtained from the power usage prediction unit 200 continues a predetermined lower limit value for a predetermined time continuously. When exceeding, the starting operation of the fuel cell power generation device 100 is performed from the time obtained by subtracting the time required for starting the fuel cell power generation device 100 from the time when the predicted value exceeds the predetermined lower limit value. In these cases, the fuel cell power generation apparatus 100 is stopped, but a rise loss due to restart occurs. However, since unnecessary power generation is not performed, energy saving can be realized.

  The fuel cell power generation system and the control method thereof according to the present invention can also be used for a power generation device having other starting means such as an engine. It can also be applied to control the amount of hot water used instead of the power used.

1 is a diagram illustrating an overall configuration of a fuel cell power generation system according to a first embodiment of the present invention. It is a block diagram which shows the structure of the controller shown in FIG. It is a figure for demonstrating the structure of the neural network model used with the neuro model prediction part of FIG. It is a flowchart for demonstrating operation | movement of the electric power generation instruction | command part of FIG. It is explanatory drawing which showed an example of the relationship between the electric power generated of a fuel cell power generation system, and an electric power use predicted value. It is a figure for demonstrating the electric power used measured by the electric power measurement part of FIG. It is a figure which shows the whole structure of the fuel cell power generation system which concerns on the 2nd Embodiment of this invention. It is a block diagram which shows the structure of the controller shown in FIG. It is a figure for demonstrating the structure of the neural network model for estimating the electric energy used by the neuro model estimation part of FIG. It is a figure for demonstrating the structure of the neural network model for estimating the hot_water | molten_metal supply amount used with the neuro model prediction part of FIG. It is a flowchart for demonstrating operation | movement of the electric power generation instruction | command part of FIG. It is explanatory drawing which shows an example of the relationship between the electric power generated of a fuel cell power generation system, and the amount of hot water storage.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1, 2 Fuel cell power generation system 100 Fuel cell power generation device 101 Controller 102 Inverter 103 Power meter 104 Household appliances 105 Commercial power supply 106 Hot water storage tank 107 Backup burner 108 Hot water supply equipment 109 Hot water supply load meter 110 City water supply part 200 Electric power use prediction part 202, 242 Neuro model prediction unit 203, 243 Neuro model learning unit 204, 244 Prediction data holding unit 205, 245 Learning data holding unit 220 Power generation command unit 230 Electric power measurement unit 240 Hot water supply amount prediction unit 250 Hot water supply amount measurement unit

Claims (13)

A fuel cell power generation system including a fuel cell power generation device that generates electric power to be supplied to an electrical device,
Power measuring means for measuring the power used by the electrical equipment;
Based on the used power measured by the power measuring means, a used power predicting means for predicting future used power for a predetermined time from a predetermined time;
An operation control means for controlling the operation of the fuel cell power generation device based on a predicted value of used power predicted by the used power prediction means. The power consumption prediction means includes prediction data holding means for holding power consumption before the predetermined time as prediction data;
A neuro model predicting means for predicting a future power consumption for a predetermined time from a predetermined time by a neural network model using the prediction data held by the prediction data holding means as an input value;
Learning data holding means for holding, as learning data, measured values of the power used in the same time zone as predicted by the neuro model prediction means;
A neuro model learning means for learning a neural network model of the neuro model prediction means based on the learning data held by the learning data holding means and a predicted value of power consumption predicted by the neuro model prediction means; The fuel cell power generation system according to claim 1, comprising:   The operation control means, when the predicted value of the used power obtained from the used power predicting means exceeds the predetermined lower limit value continuously for a predetermined time, from the time when the predicted value exceeds the predetermined lower limit value 2. The fuel cell power generation system according to claim 1, wherein the start operation of the fuel cell power generation device is performed from a time obtained by subtracting a time required for starting the fuel cell power generation device.   The operation control means, when the predicted value of used power obtained from the used power predicting means continuously falls below a predetermined lower limit value for a predetermined time, from the time when the predicted value falls below a predetermined lower limit value. The fuel cell power generation system according to claim 1, wherein the fuel cell power generation device is stopped.   2. The fuel cell power generation system according to claim 1, wherein the power measuring unit measures the upper limit value as the used power when the used power of the electric device exceeds the upper limit value of the generated power. Hot water supply measuring means for measuring the amount of hot water supplied by a hot water supply device that supplies hot water using the fuel cell power generator,
Based on the amount of hot water measured by the hot water measuring means, further comprising a hot water supply amount prediction means for predicting a future hot water supply amount for a predetermined time from a predetermined time,
The operation control means controls the operation of the fuel cell power generation device based on a predicted value of power consumption predicted by the power consumption prediction means and a predicted value of hot water amount predicted by the hot water supply amount prediction means. The fuel cell power generation system according to claim 1. The hot water supply amount prediction means includes prediction data holding means for holding the hot water supply amount before the predetermined time as prediction data;
A neuro model predicting means for predicting a future hot water supply amount for a predetermined time from a predetermined time by a neural network model using the prediction data held by the prediction data holding means as an input value;
Learning data holding means for holding, as learning data, an actual measurement value of the hot water supply amount in the same time zone as the time zone predicted by the neuro model prediction means;
A neuro model learning means for learning a neural network model of the neuro model prediction means based on the learning data held by the learning data holding means and the predicted value of the hot water amount predicted by the neuro model prediction means; The fuel cell power generation system according to claim 6, comprising:   The operation control means predicts an accumulated hot water storage amount integrated from a current hot water storage amount based on a predicted value of used power obtained from the used power prediction means and a predicted value of hot water supply obtained from the hot water supply amount prediction means. The fuel cell power generation system according to claim 6, wherein the generated power is adjusted so that the predicted accumulated hot water storage amount does not exceed a predetermined maximum hot water storage amount.   The operation control means predicts an accumulated hot water storage amount integrated from a current hot water storage amount based on a predicted value of used power obtained from the used power prediction means and a predicted value of hot water supply obtained from the hot water supply amount prediction means. When the predicted accumulated hot water storage amount exceeds a predetermined maximum hot water storage amount, a correction is made to slightly reduce the predicted power usage value, and the corrected predicted power usage value and the hot water supply amount prediction means 7. The fuel cell power generation system according to claim 6, wherein the accumulated hot water amount accumulated from the current hot water storage amount is further predicted based on the predicted value of the hot water supply amount obtained.   The operation control means predicts an accumulated hot water storage amount integrated from a current hot water storage amount based on a predicted value of used power obtained from the used power prediction means and a predicted value of hot water supply obtained from the hot water supply amount prediction means. And when the predicted accumulated hot water volume exceeds a predetermined maximum hot water storage capacity and the predicted value of the used electric power obtained from the used electric power predicting means is continuously lower than a predetermined lower limit for a predetermined time. The fuel cell power generation system according to claim 6, wherein the fuel cell power generation device is stopped from a time when the maximum hot water storage capacity is exceeded.   The operation control means predicts an accumulated hot water storage amount integrated from a current hot water storage amount based on a predicted value of used power obtained from the used power prediction means and a predicted value of hot water supply obtained from the hot water supply amount prediction means. When the predicted accumulated hot water volume exceeds a predetermined maximum hot water storage capacity, and the predicted value of the used electric power obtained from the used electric power predicting means continuously exceeds a predetermined lower limit for a predetermined time. The start operation of the fuel cell power generation device is performed from a time obtained by subtracting a time necessary for starting the fuel cell power generation device from a time when the predicted value exceeds a predetermined lower limit value. Fuel cell power generation system. A control method of a fuel cell power generation system including a fuel cell power generation device that generates electric power to be supplied to an electric device,
A power measurement step for measuring the power used by the electrical device;
Based on the power usage measured in the power measurement step, a power usage prediction step of predicting future power usage for a predetermined time from a predetermined time;
A control method for the fuel cell power generation system, comprising: an operation control step for controlling the operation of the fuel cell power generation device based on the predicted value of the power consumption predicted in the power consumption prediction step. A hot water supply measuring step for measuring the amount of hot water supplied by a hot water supply device for supplying hot water using the fuel cell power generator; and
A hot water supply amount prediction step of predicting a future hot water supply amount from a predetermined time by a predetermined time based on the hot water supply amount measured in the hot water supply measurement step,
The operation control step is a step of controlling the operation of the fuel cell power generation device based on a predicted value of the used power predicted in the used power prediction step and a predicted value of the hot water supply amount predicted in the hot water supply amount prediction step. The method for controlling a fuel cell power generation system according to claim 12, comprising: