JP3275929B2 - Design method of optimal control rule table for air conditioner - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to an air conditioner that satisfies predetermined conditions by using a genetic algorithm as an optimization method.
The present invention relates to a method for designing an optimal control rule table for a sum device .

[0002]

2. Description of the Related Art Conventionally, control of a system having nonlinearity has been performed by using a linear control theory such as PID control. In this linear control theory, optimal control is performed for all system states. And it is necessary to use a non-linear control theory such as fuzzy theory.

On the other hand, as a type of the control device, there is a control device using a one-chip microcomputer. The one-chip microcomputer has a CPU including a RAM, a ROM, a peripheral circuit, and the like. Are widely used mainly in home electric appliances at present, because they are hardly required, are inexpensive, small, and can realize various control theories by changing programs.

However, unlike a personal computer or the like, such a one-chip microcomputer has a poor calculation capability and a limited mounting memory, and therefore cannot execute a complicated and large program such as a nonlinear control theory. Have difficulty. For these reasons, current control devices based on one-chip microcomputers are mainly controlled by a control rule table.

[0005]

The above-mentioned control rule table is designed based on the experience and experiment of a skilled person who is familiar with the control object at present. Have a control rule table with a large degree of freedom that is comparable to a control device that requires a long time, a control device that requires a multidimensional control rule table such as multivariable control, or a complicated control device based on nonlinear theory. For the control device, there is a problem that it is virtually impossible to design the optimal rule table manually.

[0006] The present invention has been made in view of the above,
Its purpose is to achieve any control unit using a computer with limited capability such as a one-chip microcomputer, regardless of the complexity of the control target and the degree of freedom in the control rule table. An object of the present invention is to provide a control device that optimally satisfies conditions and an air conditioner that has a control device that satisfies arbitrary conditions such as controllability, power saving, and reliability.

[0007]

[ MEANS FOR SOLVING THE PROBLEMS] To achieve the above object, an optimal control rule for an air conditioner according to claim 1 of the present invention.
The rule table design method is applied to the air conditioner
The set temperature and the room temperature, and compare the room temperature and the set temperature.
The drive frequency of the compressor is based on the difference from
Uses a two-dimensional control rule table with each element of the numerical operation change amount
There are, air conditioner for controlling the driving frequency of the compressor
A method of designing an optimal control rule table for
The rule table design method is based on the driving of the compressor in the air conditioning unit.
Formula of the relationship between the time change of the frequency and the time change of the room temperature
Using a more modeled mathematical model and an arbitrary control rule table
Control simulation, and
Hunting amount and setting from room temperature to reaching the set temperature
Calculate the total variation of the drive frequency of the compressor until the temperature reaches,
The total hunting amount at the room temperature is evaluated with respect to the controllability value and
And the total fluctuation amount of the driving wave number for reliability and comfort of indoor space
Multiple evaluations including these evaluation values
Genetic using sum evaluation value calculation means to calculate sum of values
Algorithm and optimal control rule table extraction step
The genetic algorithm uses the control rule table as a gene.
Generating an initial control rule table for generating an initial group;
The sum evaluation one by one from the generated group of control rule tables
Input to the value calculation means, and the evaluation value for the control rule table individual
Control rule table evaluation step for obtaining the sum of
Selection is performed based on the sum of
Perform crossover and mutation on the control rule table group
To generate a new control rule table group, and select
Next generation control rule table output with the generated control rule table group
And the next-generation control rule table generation step are completed.
At the end of the genetic algorithm,
Judge whether or not the termination condition is satisfied.
Evaluation rules for the generated next-generation control rule table
Run from Tep, and if the end condition is met,
The end condition judgment step to output the group of the substitution control rule table
Extracting the optimal control rule table.
Is the control output after the end of the genetic algorithm.
Extract the one with the best sum of the evaluation values from the rule table group
The gist of this is that

The air conditioner according to claim 2 of the present invention.
The method for designing an optimal control rule table for an apparatus is described in claim 1,
In order to evaluate the control rule table, the air conditioner
Of the compressor drive frequency proportional to the power consumption of the compressor
Evaluation of energy consumption based on power consumption required from changes
It is assumed that the value has been set .

[0009]

[0010]

The optimal control rule table design of the air conditioner of the present invention
In the method, the driving frequency of the compressor in the air conditioner section
The relationship between the time change of the room temperature and the time change of the room temperature is
Control using a Dellized mathematical model and an arbitrary control rule table
Perform a simulation, and from the results of the simulation
The total hunting amount at room temperature until the set temperature is reached and the set temperature
To determine the total amount of change in the drive frequency of the compressor up to
Evaluation value regarding controllability of total hunting amount of temperature and the above
The total variation of the driving wave number is related to the reliability and the comfort of the indoor space.
Evaluation value, and the sum of multiple evaluation values including these evaluation values
Calculate the sum. Genetic algorithms, as genes
Generates the initial population of the control rules table and generates the generated control rules
Input one by one from the group of the table to the sum evaluation value calculation means.
The sum of the evaluation values for the control rule table individual,
Selection is performed based on the sum of the evaluation values, and
Further crossover and mutation for the selected control rule table group
To generate a new control rule table group,
Output with the selected control rule table group. Next generation control
When the rule table generation is completed, the genetic algorithm is terminated.
Judge whether the exit condition is satisfied, and
If not, the generated next generation control
Execute from control rule table evaluation and satisfy end condition
Output the next generation control rule group
One with the best sum of the evaluation values from the group of the control rule table
Is extracted. Created by this optimal control rule table design method
An air conditioner incorporating a control rule table is based on a control device.
The difference between the set temperature and the room temperature and the data of the change.
The control is performed by setting the compressor drive frequency from the compressor.

[0011]

[0012]

[0013]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a configuration of a control device according to an embodiment of the present invention. The control device shown in FIG., A plurality of setting control amount input means 1 _a to the controlled device 10, 1 _b, · · ·, enter the plurality of setting control amount from 1 _m, a plurality of controlled devices 10 the controlled variable is to control so as to set the control amount of the setting of the plurality of control amount input unit 1 _a, 1 _b, · · ·, each setting control amount state space dividing means from 1 _m 5 and each control amount of the controlled device 10 is controlled by control amount detection means 3 _a , 3 _b ,.
, 3 _m , and supplies each detected control amount to the state space dividing means 5. As described in detail below, the state space dividing means 5 calculates the difference between the corresponding set control amount and each control amount and the change in the difference, and performs quantization for state space division. Then, the environmental state at time t is calculated as an “environmental state vector” and supplied to the manipulated variable determining means 7.

The manipulated variable determining means 7 determines each manipulated variable of the controlled device 10 from the discretized environmental state vector based on the "control rule table" from the control rule table optimizing means 9, and determines this. The operation amounts are supplied to the controlled device 10 and controlled.

Next, a more detailed description will be given using mathematical expressions.

At time t, each set control amount input from each set control amount input means 1 _a , 1 _b ,..., 1 _m is represented by C _s1 (t) to C _sm (t), When the control amounts detected by the amount detection means 3 _a , 3 _b ,..., 3 _m are C ₁ (t) to C _m (t) and the sampling time is δ,

(Equation 1) think of. Then, in order to divide the state space, the i-th control amount ｛E given by the above equation (1)
_i (t), _{ΔE i} and _{(t)} α i1 <α} i2 <... <α ik (2) β i1 <β i2 <... < When β _il (3),

(Equation 2) Quantization is performed for state space division as shown in FIG. Based on this quantized {E ', ΔE'}

Equation 3] _{E (t) = {E 1} (t), ΔE 1 (t), E 2 (t), ΔE 2 (t), ... E m (t), ΔE m (t)} (6) As described above, the environment state at time t is represented by the state space dividing means 5 as an “environment state vector”.

Then, based on the discretized environmental state vector, the operation amount of the controlled device 10 is determined by the "control rule table" from the control rule table optimizing means 9 connected to the operation amount determination means 7.

Now, when there are m manipulated variables to be controlled in the controlled device 10, the manipulated variable at time t is:

Equation 4] _{O (t) = {O 1} (t), O 2 (t), ..., O n (t)} Assuming that expressed as (7), the operation amount at time t + [delta] is O (t + [delta] ) = O (t) + ΔO (t) (8) However, the change amount ΔO (t) of the operation amount is

ΔO (t) = {ΔO ₁ (t), ΔO ₂ (t),..., ΔO _n (t)} (9), and O (t) = f (E (t) This is the “variation vector” given by (10). Equation (1
In 0), f is a control rule table that has been optimized in advance by the control rule table optimizing means 9 to achieve an arbitrary purpose. When the control amount is m, a 2m ^2m- dimensional element having ΔO as an element is used. Expressed as a tensor.

The controlled device 10 operated according to the manipulated variable determined as described above affects the controlled variable, and each controlled variable detected by the controlled variable detecting means 3 _a , 3 _b ,..., 3 _m . Is supplied to the state space dividing means 5. By configuring the feedback loop in this manner, a stable set control amount can be obtained.

FIG. 2 is a block diagram showing a detailed configuration of the control rule table optimizing means 9 used in the control device shown in FIG.

In order to design the optimal control rule table shown in FIG.
Control rule table optimizing means 9 of the air conditioner unit
A control target 21 comprising a compressor drive frequency and a room temperature,
The change over time of the drive frequency of the compressor to be controlled and the indoor
Model the relationship with temperature over time using mathematical formulas
Control target formula modeling means 23 to be constructed, the number of control targets
Formula model 2 modeled by formula modeling means 23
5. A control system using the mathematical model 25 and an arbitrary control rule table.
Simulation, and set based on the simulation results.
Total hunting amount at room temperature until reaching the set temperature and reaching the set temperature
Calculate the total variation of the driving frequency of the compressor up to the room temperature
The evaluation value for controllability and the drive
The total fluctuation of the wave number is related to the reliability and the comfort of the indoor space.
The sum of multiple evaluation values including these evaluation values
Evaluation function 27 for calculating the control rule table as a gene
Generates an initial population and from the generated population of control rule tables
The control rules are input one by one to the total evaluation value calculation means.
The sum of the evaluation values for the table individuals is obtained, and the sum of the evaluation values is calculated.
Control based on the selection
Crossover and mutation are further performed on the rule table
Generate a new control rule table group and select the control selected in the selection.
Genetic algorithm 29 output with rule table population,
The output next-generation control rule table population is
Judge whether the end condition of the
If the conditions are not met, the generated next-generation control rules
Execute from the control rule table evaluation step for the table,
If the condition is satisfied, output a group of next-generation control rule table
Control rule table 31 and the genetic algorithm 2
After the completion of 9, evaluation is performed from the group of the output control rule table 31.
Optimal control rule table 3 for extracting one with the best sum of values
3 is comprised.

[0023] In the control rule table optimization means 9 in FIG. 2, the control
Control simulation is performed one by one from the group of rule table 31.
In order to calculate the respective evaluation functions , the control target mathematical expression modeling means 23 constructs a mathematical expression model of the control target 21 as shown in the following expression. This formula
Modeling methods are based on physical laws or physical laws.
If delling is difficult, for example, assuming a linear model,
Input M-sequence signal and use this system for least squares method.
I have decided. (For example, Takayoshi Nakamizo: Signal Analysis and System
Identification, see Corona, 1988)

C (t) = sys (O (t)) (11) The control simulation is repeated based on this mathematical model, and the optimal control rule table 3 that satisfies the evaluation function 27 by the genetic algorithm 29 which is the optimal algorithm.
Get 3.

In the optimization of the discretized control rule table, that is, in the combinatorial optimization problem, an optimum solution is obtained in principle, but the present invention is applied to an actual problem such as a multidimensional control rule table optimization problem. Then, the number of combinations increases. If the number of combinations increases dramatically, calculation becomes impossible. Therefore, it is necessary to efficiently find a suboptimal solution, and there are a heuristic method and a random method as solutions for that.

In the heuristic method, a planner who is familiar with the problem fully configures the algorithm by making full use of the inherent properties of the problem. Usually, the amount of calculation is small, but the optimality of the obtained solution is small. There is no compensation and a procedure to evaluate the solution is required.

On the other hand, the random method does not use the nature of the problem at all, but selects a large number of solutions at random, and among them, the one with the smallest evaluation function value is used as a suboptimal solution. This method has a simple algorithm, and a good solution can be obtained if it takes enough time.

In contrast, the genetic algorithm 29 (DEGoldberg: Generic Algorithm in Ser
ch, Optimization and Machine Learning, Addison-We
sley (1989)) describes an adaptation process of a natural solution proposed by Holland et al., and is an artificial model that simulates the mechanism of biological evolution, and is considered to be intermediate between heuristic and random methods. . The genetic algorithm 29 incorporates randomness into its solution, and at the same time, can introduce a unique form into the solution construction method and operation procedure. By combining both, a globally good solution can be obtained. Search efficiently.

Next, the concept of the genetic algorithm will be described in detail with reference to the flowchart shown in FIG. Note that FIG. 3 shows a generally introduced genetic algorithm.
This shows the basic flowchart of the algorithm.
The operation of the genetic algorithm 29 shown in FIG.
It is not a direct indication.

The genetic algorithm 29 first has a control rule
The event for determining the genotype of the table is modeled (step 110). In this event modeling, a symbol string inside a computer is used as a gene, and what numbers and characters are assigned to each symbol string. Decide what to assign. In the case of this embodiment, a two-dimensional gene is considered, and the control rule table itself is assigned.

Then, the control determined by modeling the event
Based on the above genotypes in the rules table,
Or in Japanese Unexamined Patent Publication No. 5-120018.
Create an initial population of genetic algorithms, like
A method for determining a control rule table using random numbers used as a method
Generates a group consisting of a large number of individuals (step 120). Next, crossover is performed. In this case, a specific gene pair is selected from the gene population, and the specific site is replaced to create a new individual (step 130). Then, a mutation is performed in which a specific site of the gene is changed at a certain probability (step 140). Then, the individual is evaluated . In this case, an evaluation function to be used as a reference of the evaluation value is determined, and the symbol string is associated with an actual object (step 150). This evaluation function plays the most important role in connecting the symbol string with the real world. Then, according to the evaluation value based on the evaluation function value, selection is performed to select an individual that can survive in a later generation (step 160). Soshi
Again to the routine of crossover and mutation,
When the number of repetitions of the routine reaches a predetermined value or
For the sum of multiple evaluation values calculated for each control rule table
The termination condition is when all the evaluations reach a predetermined level or higher.
By repeating until the end condition is reached,
170), improve the performance of the gene population
When the termination condition is reached, the current
The best union with the best evaluation function extracted from the group
And the solution to The evaluation function of the gene population
The value is the value before performing the individual evaluation (step 150) described above.
In a separate routine from the operation of the genetic algorithm at the stage
Control stains of a gene population one by one using a mathematical model
And calculate the evaluation function value.

Further , applying the genetic algorithm described above,
In the control rule table optimizing means 9 shown in FIG.
Of the control rule table determined by modeling the event
The initial population of the control rule table 31 was generated based on the genotype.
One by one from the group of the generated control rule table 31
A control simulation is performed on the mathematical model 25 to
The total room temperature from the simulation result to the set temperature
Hunting amount and compressor drive frequency until reaching set temperature
Of total hunting at room temperature
Reliability of the evaluation value and the total variation of the number of drive waves
And evaluation values for the comfort of the indoor space
Total evaluation value calculation that calculates the sum of multiple evaluation values including values
Calculating the evaluation function 27 using the means;
The population of the control rule table for which the evaluation function values have been calculated is
According to the algorithm 29, the initial stage of the control rule table as a gene
Generating an initial control rule table for generating a group of
The total evaluation value meter one by one from the group of the set control rule table
Input to the calculation means to calculate the total
A control rule table evaluation step for obtaining a sum;
Selection is performed based on the sum and the system selected by selection
Perform further crossovers and mutations on the control rule table group
To generate a new control rule table group, and select
Next generation control rule table output with the generated control rule table group
And the next-generation control rule table generation step are completed.
At the end of the genetic algorithm,
Judge whether or not the termination condition is satisfied.
Evaluation rules for the generated next-generation control rule table
Run from Tep, and if the end condition is met,
The end condition judgment step to output the group of the substitution control rule table
After the genetic algorithm,
The one with the best sum of the evaluation values from the group of the control rule table 31
One is extracted and used as the optimal control rule table 33.

As described above, since the genetic algorithm is simple, the global algorithm can be searched for more efficiently than other methods because the algorithm is simple.
That is, a sub-optimal control rule table can be efficiently obtained.

The evaluation function used in the optimization algorithm considers at least the matters described below.
FIG. 4 is a schematic diagram of these evaluation functions.

(1) As an evaluation function for determining whether controllability is good or bad,

(Equation 6) Is defined.

(2) The energy required for the operation of the controlled device 10 is considered to be substantially proportional to the manipulated variable.

(Equation 7) Is defined.

(3) The reliability of the controlled device 10 is considered to be inversely proportional to the amount of change in the manipulated variable.

(Equation 8) Is defined.

In addition, other matters that the control device designer wants to consider are defined as described above, and these arbitrary evaluation functions are defined as follows.

(Equation 9) And the control rule table is determined by an optimization algorithm such as a genetic algorithm so as to minimize the J. In the equation (15), wi is Ri positive weight coefficient der, or the design value of the control specification
It is given as a parameter obtained from it .

By considering the plurality of evaluation functions as described above, it is possible to easily design a control device that satisfies controllability, energy saving, and reliability of the controlled device.

As described above, the control target system, that is, the control target 21 is defined by the mathematical model 25, and the genetic algorithm 29 which is an optimization algorithm which can be realized by a computer.
And the like, it is possible to easily design a multi-dimensional control rule table having a high degree of freedom, which is practically impossible by hand. By considering the controllability, energy saving, and reliability of the controlled device in the evaluation function, it is possible to easily design a controller that satisfies these conditions. Can perform optimal operation to satisfy these conditions.

FIG. 5 has a control device according to the present invention .
FIG. 2 is a block diagram illustrating a configuration of an air conditioner . The air conditioner shown in FIG. 3 is obtained by applying the control device of the embodiment shown in FIG. 1 as an air conditioner, and has a set temperature corresponding to the set control amount input means 1a, 1b,. Input means 41, state space dividing means 43 corresponding to the state space dividing means 5, compressor rotational speed determining means 45 corresponding to the operation amount setting means 7, control rule table optimizing means 9
, An air conditioner unit 49 corresponding to the controlled device 10, and a suction temperature detecting unit 51 corresponding to the control amount detecting units 3a, 3b,..., 3m. Have been.

In the air conditioner shown in FIG.
the _a and the control amount, the frequency Hz of the compressor and the operation amount, thereby the room temperature T _a and controlled by the frequency Hz, and controls so that the indoor temperature T _a becomes the set temperature T _s.

In FIG. 5, the set temperature T at a certain time t is shown.
_s (t) is input from the set temperature input means 41 to the state space dividing means 43, also detects the room temperature T _a by suction temperature detecting means 51, inputs the detected room temperature T _a to the state space dividing means 43 It is supposed to.

[0044] Then, the indoor temperature T _a sensed by the set temperature T _s (t) and the suction temperature detecting means 51 which is input from the set temperature input means 41

E (t) = T _a (t) −T _s (t) (16) ΔE (t) = T _a (t) −T _a (t−δ) (17) Here, δ is a sampling time. Based on these E (t) and ΔE (t), the environmental state is classified into the state space by the state space dividing means 43 as follows, and the environmental state at time t is defined.

[0045]

[Equation 11] Using the environmental state defined by the equations (18) and (19) as an input value, the compressor rotation speed determining means 45 determines the frequency Hz of the compressor included in the air conditioner unit 49 at the time (t + δ).

The compressor speed determination means 45 is a two-dimensional control rule table of {E ′ (t), ΔE ′ (t)}, as shown in FIG. The corresponding operation change amount ΔHz (E ′ (t), ΔE ′ (t)),
According to this control rule table, the manipulated variable H at time (t + δ)
z (t + δ) is

(12) Hz (t + δ) = Hz (t) + ΔHz (E ′ (t), ΔE ′ (t)) (20)

This control rule table is designed by the control rule table optimizing means 47 having a control function in consideration of controllability, energy saving, compressor reliability, and smoothness of the control rule table.

The control rule table optimizing means 47 for designing the optimum control rule table is controlled by the control rule table optimizing means 9 shown in FIG.
1, a mathematical expression modeling means 63 based on a mathematical expression corresponding to the controlled object mathematical expression modeling means 23, and a mathematical expression model 6 corresponding to the mathematical expression model 25
5. Evaluation function 67 corresponding to evaluation function 27, genetic algorithm 69 corresponding to genetic algorithm 29, control rule table 71 corresponding to control rule table 31, and optimal control rule table 73 corresponding to optimal control rule table 33 It is configured.

[0049] In the control rule table optimization means 47 in FIG. 7, control
Control simulation one by one from the group of the rule table 71
In order to obtain the respective evaluation values, first, a mathematical expression model of the compressor frequency-suction temperature system 61 to be controlled is constructed by the mathematical expression modeling means 63. In this system,
Since it is difficult to model based on the laws of physics, assuming a linear model, an M-sequence signal is input and the system is identified by the least squares method (for example, Takayoshi Nakamizo: Signal Analysis and System Identification) Corona, 1988).

The evaluation function 67 considers three types of factors as shown in FIG. First, as for the controllability, the evaluation function J ₁ is set based on the idea that “the one that quickly reaches the set temperature and does not cause hunting is better”.

(Equation 13) Is defined as

Next, in the case of an air conditioner, the compressor occupies most of the power consumption.
Evaluation function J ₂

[Equation 14] Is defined.

[0052] Further, as those frequency variation of the compressor is small, since the reliability and user comfort is improved, the evaluation function J ₃

(Equation 15) Is defined as

Further, considering the smoothness of the control rule table,
Evaluation function J ₄

(Equation 16) Is defined as Here, m and n represent the size of the control rule table.

Finally, these evaluation functions J _{1 to} J ₄ are

Equation 17] _{J = w 1 J 1 + w} 2 J 2 + w 3 J 3 + w 4 J 4 (25) of the way to integrate the four evaluation function one uses. However, w ₁ ~w ₄ is a positive weighting factor.

The control rule table optimizing means 47 shown in FIG.
Like the control rule table optimizing means 9 of 2, the air conditioner
Change of compressor drive frequency in air conditioner
The relationship of the time change of the room temperature due to the suction temperature
Control using a modeled mathematical model and an arbitrary control rule table
Perform a simulation and check if the
Hunting amount from room temperature to set temperature and set temperature
Determine the total variation of the drive frequency of the compressor until the arrival,
Evaluation value for controllability of total hunting amount at room temperature and before
The total fluctuation of the driving wave number is related to the reliability and the comfort of the indoor space.
Of multiple evaluation values including these evaluation values
Genetic algorithm using sum evaluation value calculation means to calculate sum
Algorithm and an optimal control rule table extracting step.
The tradition algorithm is the initial of the control rule table as a gene.
Initial control rule table generating step for generating a group of
Calculating the total evaluation value one by one from the group of the set control rule tables
Input to the means and sum of the evaluation values for the control rule table individual
Control rule table evaluation step for obtaining the sum of the evaluation values
Perform selection as a criterion and select the control rule selected by selection.
Crossover and mutation are further performed on the rule table group to
Generates a simple control rule table group, and selects the control rules selected in the selection.
Next-generation control rule table generation step output with the rule table group
When the next generation control rule table generation step is completed,
Whether the termination condition of the genetic algorithm is satisfied
Is determined, and if the termination condition is not satisfied,
Control rule table evaluation step again for next-generation control rule table
From next generation, and if the termination condition is met, next-generation control
Execute the end condition judgment step to output the rule table group
Then, after the genetic algorithm is finished, the output control
Extract the one with the best sum of the evaluation values from the rule table group
Executing the optimal control rule table extracting step,
One is the optimal control rule table 73.

FIG. 9 shows a control simulation result based on a control rule table designed manually by the conventional method.
Shows a control simulation result based on the control rule table designed according to the present invention. As is clear from comparison between the two, it is apparent that the performance of the control device is clearly improved in terms of controllability, energy saving, and reliability. The above implementation
In the example, the control device and the air conditioner
Optimization means are incorporated, and the equipment itself sets the control rule table optimally.
Control based on the optimally designed control rule table.
Configuration. As another embodiment, the control rule table
Disconnect the optimization means from the control device or air conditioner and
Control rule table optimized by control table optimization means
Installed in an air conditioner or
It can be configured to control according to the appropriate control rule table.
You.

As described above, the control rules, which can realize complicated control rules, are controlled by computer simulation and the genetic algorithm known as a high-efficiency optimization method is used for controllability, energy saving, and compressor reliability ( It is possible to design a control device, that is, an air conditioner that takes into account the user's comfort. Further, it is possible to easily and quickly design an air conditioner in consideration of controllability, energy saving, and compressor reliability without the need for actual machine tests and manual trial and error. In addition, this air conditioner can provide a comfortable space to the user while ensuring energy saving.

[0058]

As described above, according to the present invention,
According to 1, evaluation of optimization means to optimize the control rule table
The values are controllability and reliability and the comfort of indoor space.
Satisfaction of the user's comfort by using the evaluation function of
Control rules incorporated in the control device of the air-conditioning system
Tables can be designed quickly and easily, and
The results generated by the rule table optimizer are
It can be implemented in products as an optimal control rule table.
You. According to the present invention, an evaluation for evaluating a plurality of control targets is performed.
Obtained from the design value of the control specification as a function weighting factor
Generate optimal control rule table just by giving parameters
Therefore, the trouble of manual tuning can be eliminated, and a control rule table having a large space or a control rule table which is difficult to design manually can be easily designed. According to claim 2 of the present invention, the user's comfort
For the conflicting conditions of suitability and energy saving,
Equipped with a control device incorporating the optimal control rule table
The air conditioner can be easily designed.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a control device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a detailed configuration of a control rule table optimizing means used in the control device shown in FIG.

FIG. 3 is a flowchart showing the operation of a genetic algorithm used in the control rule table optimizing means of FIG. 2;

FIG. 4 is a schematic diagram of an evaluation function used in the control rule table optimizing means of FIG. 2;

FIG. 5 is a block diagram showing a configuration of an air conditioner having a control device according to the present invention.

FIG. 6 is a diagram showing a control rule table constituting a compressor rotation speed determining means used in the air conditioner of FIG. 5;

FIG. 7 is a block diagram showing a configuration of a control rule table optimizing means used in the air conditioner of FIG.

FIG. 8 is a schematic diagram of an evaluation function used in the control rule table optimizing means of FIG. 7;

FIG. 9 is a diagram showing a control simulation result based on a manually designed control rule table according to a conventional method.

FIG. 10 is a diagram showing a control simulation result based on a control rule table designed according to the present invention.

[Explanation of symbols]

_1a , _1b ,..., _1m setting control amount input means _3a , _3b ,..., _3m control amount detecting means 5 state space dividing means 7 operation amount determining means 9 control rule table optimizing means Reference Signs List 10 controlled device 21 controlled object 23 controlled object mathematical expression modeling means 25 mathematical expression model 27 evaluation function 29 genetic algorithm 31 control rule table

Continuation of the front page Judge Chief Tamio Tokunaga Judge Shoji Ishikawa Judge Judge Tomoyuki Murakami (56) References JP-A-5-120018 (JP, A) National Technical Report, Vol. 37, No. 6 P639-P. 645 "Neiro Control Air Conditioner-Comfort Control by Neural Network" (58) Fields investigated (Int. Cl. ⁷ , DB name) F24F 11/02 G06N 3/00

Claims (2)

(57) [Claims] 1. An air conditioner unit which is a controlled device.
The set temperature and the room temperature, and compare the room temperature and the set temperature.
And the change in the difference, the driving frequency of the compressor
Using a two-dimensional control rule table with each operation change amount as an element
Te, air conditioner for controlling the driving frequency of the compressor
The optimal control rule table design method, wherein the optimal control rule
The table design method is based on the driving cycle of the compressor in the air conditioner unit.
The relationship between the time change of the wave number and the time change of the room temperature is expressed by an equation.
Using a mathematical model and an arbitrary control rule table
Perform a control simulation and the simulation results
Total hunting amount from room temperature to set temperature and set temperature
The total amount of change in the drive frequency of the compressor until the temperature reaches
Evaluation value for controllability of total hunting amount at the room temperature and
The total fluctuation amount of the driving wave number is used for reliability and comfort of indoor space.
Multiple evaluation values including these evaluation values
Genetic algorithm using the sum evaluation value calculation means for calculating the sum of
Algorithm and an optimal control rule table extracting step,
Genetic algorithms are the first in the control rules table for genes.
Generating an initial control rule table for generating a group of
The total evaluation value one by one from the group of the generated control rule table
Input to the calculation means and calculate the evaluation value
A control rule table evaluation step for obtaining a sum;
Selection was performed based on the summation and selected by selection
Perform further crossover and mutation on the control rule table group.
Generated a new control rule table group and selected it in the selection
Generation of next-generation control rule table output together with control rule table group
Step and next generation control rule table generation step are completed
At the stage, the termination condition of the genetic algorithm is satisfied
Judge whether the end condition is not satisfied.
The control rule table evaluation step
The next generation
From the end condition judgment step that outputs the group of the control rule table
Wherein the optimal control rule table extracting step comprises:
After the end of the genetic algorithm, the output control rules
Extract the one with the best sum of the evaluation values from the group of the table
Optimal control rules for air-conditioning equipment
Rule table design method. 2. The method according to claim 1 , further comprising the step of:
In addition, the compressor drive, which is proportional to the power consumption of the air conditioner,
Saves energy consumption from time-dependent changes in operating frequency
2. An evaluation value regarding the performance.
Design method of optimal control rule table for air conditioners in Japan.