CN110416995B - Non-invasive load decomposition method and device - Google Patents

Abstract

The invention discloses a non-invasive load decomposition method and a device, wherein the method comprises the following steps: acquiring operation records of electric equipment and power data on a bus, wherein the power data comprises: a current waveform; preprocessing the power data according to the operation record of the electric equipment to screen out first power data; determining four power parameter values corresponding to each moment according to the screened first power data; establishing a multi-target non-invasive load decomposition model with five optimization targets according to the power parameter values and the current waveforms; a type of powered device and a corresponding mode being operated on the bus are determined based on the multi-objective non-intrusive load decomposition model. The non-invasive load decomposition method provided by the invention has the advantages of high decomposition accuracy, stable result and stronger robustness.

Description

A non-invasive load decomposition method and device

Technical Field

The present invention relates to the technical field of pattern recognition, and more particularly to a non-invasive load decomposition method and device.

Background Art

In the past two decades, with the development of China's economy and the substantial improvement of people's living standards, the annual growth rate of residential electricity consumption has been around 8%. The increase in electricity demand will increase the negative impact on the environment. This is inconsistent with the Chinese government's efforts to reduce greenhouse gas emissions. The monitored household appliance energy consumption information can help decision makers and consumers understand the composition, pattern and characteristics of residential energy demand. This plays an important role in energy conservation and emission reduction, so it is particularly important to design a system that can effectively monitor the load.

Existing load monitoring methods are mainly divided into two categories: invasive and non-invasive methods. The invasive load monitoring method (ILM) requires that each power-consuming device be equipped with an instrument with communication function, which will increase the cost of deploying and maintaining measuring instruments; the non-invasive load monitoring method (NILM) requires that only one measuring instrument be installed at the user entrance of the power grid, and the total power consumption information collected is analyzed through an algorithm to monitor the power consumption status of each power-consuming device under it. For large-scale deployment, non-invasive load monitoring systems can significantly reduce installation complexity and reduce maintenance costs.

Optimization and pattern recognition are two mainstream methods for solving the problem of load decomposition and identification. The load identification algorithm based on pattern recognition determines the structure and parameters of the identification algorithm by learning the characteristics of each electrical device in the database, and finally realizes the identification of the load. The non-intrusive load monitoring system based on pattern recognition must first be trained, but the trained classifier can only identify the electrical devices and corresponding combinations that already exist in the data set. Therefore, every time a new device is added, the model needs to be trained, which consumes a lot of time and computing resources. And when there are many electrical devices, the recognition accuracy is low. The load identification algorithm based on optimization converts the load identification problem into an optimization problem. The difference between the feature vector of the unknown electrical device and the feature vector of the known electrical device is minimized to achieve the purpose of load identification. The load identification method based on optimization only needs to obtain the load characteristics of a single electrical device, and can calculate the load characteristics of multiple electrical devices when they are running simultaneously, provided that these load characteristics must meet the feature superposition standard.

Recent optimization-based research uses 1 to 3 objective functions, that is, only 1 to 3 features. However, most existing studies weight multiple objective functions into one. This will lead to two problems: (1) The weighting parameters are sensitive to the data set; (2) The weighting parameters corresponding to different objective functions are also difficult to adjust. This is because the error in load characteristics will seriously affect the accuracy of optimization-based methods. If only one feature is used as the objective function, the optimal solution obtained is not the actual operating state of the electrical equipment; and if the electrical equipment has similar or overlapping load characteristics, it is difficult to distinguish them using a single load characteristic, resulting in low load decomposition accuracy.

Therefore, how to provide a load decomposition method with high decomposition accuracy is an urgent problem to be solved by those skilled in the art.

Summary of the invention

In view of this, the present invention provides a non-invasive load decomposition method and device, which has high decomposition accuracy, stable results and strong robustness.

In order to achieve the above object, the present invention adopts the following technical solution:

A non-invasive load decomposition method comprises the following steps:

Acquiring operation records of power-consuming equipment and power data on the bus, wherein the power data includes: current waveform;

Preprocessing the power data to select first power data according to the power-consuming equipment operation record;

Determine four power parameter values corresponding to each moment according to the filtered first power data;

According to the power parameter value and the current waveform, a multi-objective non-intrusive load decomposition model with five optimization objectives is established;

Based on the multi-objective non-intrusive load decomposition model, the types of electrical equipment working on the bus and the corresponding working modes are determined.

Preferably, preprocessing the power data to select the first power data according to the power-consuming equipment operation record specifically includes:

According to the operation records of the power-consuming equipment, the power data is divided according to the operation events of each power-consuming equipment to obtain a power data division result, so that the power data in the same interval is in the same mode of the same equipment;

Perform outlier detection and elimination on the current and voltage in the same interval to obtain an outlier elimination result;

According to the power data division results and abnormal value elimination results, the average current and voltage in the interval where no electrical equipment is working are calculated. Under the same circuit, the average current and voltage in the interval where no electrical equipment is working are subtracted from the current and voltage values corresponding to all intervals where useful electrical equipment is working to obtain the first power data.

Preferably, the specific method for detecting and eliminating abnormal values of current and voltage in the same interval includes:

According to the formula MAD=median(|A _i -median(A)|), the median absolute deviation is determined, where A represents the current value or voltage value in the same interval, _Ai represents the current value or voltage value corresponding to the i-th moment, and MAD represents the median absolute deviation;

Determine whether the value of the current value or voltage value in the same interval at any time deviates from the median by more than N times the median absolute deviation. If so, determine that the current value or voltage value at this moment is an abnormal value and remove the abnormal value.

Preferably, the four power parameter values corresponding to each moment are determined according to the screened first power data, specifically including:

According to the formula: S = VI, determine the apparent power corresponding to each moment, where V represents the voltage value at a certain moment, I represents the current value at a certain moment, and S represents the apparent power;

According to the formula: P = VIcos (φ), determine the active power corresponding to each moment, where cos (φ) represents the power factor and P represents the active power;

According to the formula: Q = VIsin (φ), determine the reactive power corresponding to each moment, where Q represents the reactive power;

确定各时刻对应的谐波，其中，A(t)表示在t时刻电流值，约束条件是t＝0，1，2，…，T-1，T代表在一个电流波形周期内采样点的数量，X(k)是第k次谐波的系数。According to the formula:

Determine the harmonics corresponding to each moment, where A(t) represents the current value at moment t, the constraint is t=0, 1, 2, ..., T-1, T represents the number of sampling points in one current waveform cycle, and X(k) is the coefficient of the kth harmonic.

Preferably, according to the power parameter value and the current waveform, a multi-objective non-intrusive load decomposition model with five optimization objectives is established, specifically including:

Five optimization objectives are constructed based on power parameter values and current waveforms; wherein the power parameter values include active power, reactive power, apparent power and harmonics; the specific method is as follows:

以电流波形为基础构建第一目标函数，其中，I_ij(t)表示在t时刻设备i处于j模式下独立运行时的电流值；T代表在一个电流波形周期内采样点的数量，t＝0表示电压相位处于特定的时刻，此刻的电压波形正从最大值向最低值变化，N表示设备总数，M_i表示设备i包含的模式总数，I表示未知设备类型的组合电流波形；x_ij表示用电设备i处于工作模式j，其中i＝1，2，…，N，j＝1，2，…，M_i；According to the formula:

The first objective function is constructed based on the current waveform, where I _ij (t) represents the current value when the device i is operating independently in mode j at time t; T represents the number of sampling points in a current waveform cycle, t=0 represents that the voltage phase is at a specific moment, and the voltage waveform at this moment is changing from the maximum value to the minimum value, N represents the total number of devices, _Mi represents the total number of modes included in device i, and I represents the combined current waveform of unknown device types; x _ij represents that the power device i is in working mode j, where i=1, 2, ..., N, j=1, 2, ..., _Mi ;

以无功功率为基础构建第二目标函数，其中，Q_ij表示设备i处于j模式下独立运行时的无功功率，Q代表未知设备类型的无功功率；According to the formula:

The second objective function is constructed based on reactive power, where _Qij represents the reactive power of device i when it operates independently in mode j, and Q represents the reactive power of an unknown device type;

以有功功率为基础构建第三目标函数，其中，P_ij表示设备i处于j模式下独立运行时的有功功率，P代表未知设备类型的有功功率；According to the formula:

The third objective function is constructed based on active power, where _Pij represents the active power of device i when it operates independently in mode j, and P represents the active power of the unknown device type;

以视在功率为基础构建第四目标函数，其中，S_ij表示设备i处于j模式下独立运行时的视在功率，S代表未知设备类型的视在功率；According to the formula:

The fourth objective function is constructed based on the apparent power, where S _ij represents the apparent power of device i when it operates independently in mode j, and S represents the apparent power of the unknown device type;

以谐波为基础构建第五目标函数，其中，H_ij(k)表示设备i处于j模式下独立运行时的谐波，K是最大谐波的次序，H(k)代表未知包含设备类型的谐波；According to the formula:

The fifth objective function is constructed based on harmonics, where _Hij (k) represents the harmonics of device i when it operates independently in mode j, K is the order of the maximum harmonic, and H(k) represents the harmonics of the unknown containing device type;

According to the formula: minimize F(x) = (f ₁ (x), f ₂ (x), f ₃ (x), f ₄ (x), f ₅ (x)), the above five objective functions are constructed into a multi-objective non-intrusive load decomposition model F(x), where:

表示所求用电设备类型，需要满足约束条件x_ij∈{0，1}和

represents the type of electrical equipment required, which needs to satisfy the constraints x _ij ∈ {0, 1} and

Accordingly, determining the type of electrical equipment working on the bus and the corresponding working mode based on the multi-objective non-intrusive load decomposition model specifically includes:

A multi-objective evolutionary algorithm is used to solve the multi-objective non-intrusive load decomposition model F(x) to obtain the corresponding electrical equipment type and corresponding working mode when the five optimization objectives in F(x) are minimized at the same time;

Among them, in the multi-objective evolutionary algorithm, the constraints are expressed in an encoding and decoding manner: each bit of the code represents a type of electrical equipment, where 0 means that the electrical equipment is turned off, and k means that the electrical equipment is in the kth working mode.

A non-invasive load decomposition device, comprising:

An acquisition module is used to acquire operation records of power-consuming equipment and power data on the bus, wherein the power data includes: current waveform;

A preprocessing module, configured to preprocess the power data to select first power data according to the operation record of the power-consuming device;

A parameter calculation module, used to determine four power parameter values corresponding to each moment according to the screened first power data;

A model building module, used for building a multi-objective non-intrusive load decomposition model with five optimization objectives according to the power parameter value and the current waveform;

A determination module is used to determine the type and corresponding mode of the electrical equipment working on the bus based on the multi-objective non-intrusive load decomposition model.

Preferably, the preprocessing module specifically includes:

A division unit, configured to divide the power data according to each power-consuming device operation event according to the power-consuming device operation record, and obtain a power data division result so that the power data in the same interval is in the same mode of the same device;

An outlier elimination unit is used to detect and eliminate outliers of current and voltage in the same interval to obtain an outlier elimination result;

The calculation unit is used to calculate the average current and voltage in the interval where no electrical equipment is working according to the power data division result and the abnormal value elimination result. Under the same circuit, the average current and voltage in the interval where no electrical equipment is working is subtracted from the current and voltage values corresponding to all intervals where useful electrical equipment is working, so as to obtain the first power data.

Preferably, the outlier elimination unit specifically includes:

A calculation subunit, used to determine the median absolute deviation according to the formula MAD=median(|A _i -median(A)|), where A represents the current value or voltage value in the same interval, _Ai represents the current value or voltage value corresponding to the i-th moment, and MAD represents the median absolute deviation;

The elimination subunit is used to determine whether the value of the current value or voltage value deviating from the median in the same interval at any time is greater than N times the median absolute deviation. If so, the current value or voltage value at this moment is an abnormal value and the abnormal value is eliminated.

Preferably, the parameter calculation module specifically includes:

The apparent power calculation unit is used to determine the apparent power corresponding to each moment according to the formula: S=VI, where V represents the voltage value at a certain moment, I represents the current value at a certain moment, and S represents the apparent power;

An active power calculation unit, used to determine the active power corresponding to each moment according to the formula: P = VIcos (φ), where cos (φ) represents the power factor and P represents the active power;

A reactive power calculation unit, used to determine the reactive power corresponding to each moment according to the formula: Q=VIsin(φ), where Q represents the reactive power;

确定各时刻对应的谐波，其中，A(t)表示在t时刻电流值，约束条件是t＝0，1，2，…，T-1，T代表在一个电流波形周期内采样点的数量，X(k)是第k次谐波的系数。Harmonic calculation unit, used according to the formula:

Determine the harmonics corresponding to each moment, where A(t) represents the current value at moment t, the constraint is t=0, 1, 2, ..., T-1, T represents the number of sampling points in one current waveform cycle, and X(k) is the coefficient of the kth harmonic.

Preferably, the model building module specifically includes:

以电流波形为基础构建第一目标函数，其中，I_ij(t)表示在t时刻设备i处于j模式下独立运行时的电流值；T代表在一个电流波形周期内采样点的数量，t＝0表示电压相位处于特定的时刻，此刻的电压波形正从最大值向最低值变化，N表示设备总数，M_i表示设备i包含的模式总数，I表示未知设备类型的组合电流波形；x_ij表示用电设备i处于工作模式j，其中i＝1，2，…，N，j＝1，2，…，M_i；The first objective function building unit is used according to the formula:

The first objective function is constructed based on the current waveform, where I _ij (t) represents the current value when the device i is operating independently in mode j at time t; T represents the number of sampling points in a current waveform cycle, t=0 represents that the voltage phase is at a specific moment, and the voltage waveform at this moment is changing from the maximum value to the minimum value, N represents the total number of devices, _Mi represents the total number of modes included in device i, and I represents the combined current waveform of unknown device types; x _ij represents that the power device i is in working mode j, where i=1, 2, ..., N, j=1, 2, ..., _Mi ;

以无功功率为基础构建第二目标函数，其中，Q_ij表示设备i处于j模式下独立运行时的无功功率，Q代表未知设备类型的无功功率；The second objective function building unit is used according to the formula:

The second objective function is constructed based on reactive power, where _Qij represents the reactive power of device i when it operates independently in mode j, and Q represents the reactive power of an unknown device type;

以有功功率为基础构建第三目标函数，其中，P_ij表示设备i处于j模式下独立运行时的有功功率，P代表未知设备类型的有功功率；The third objective function building unit is used according to the formula:

The third objective function is constructed based on active power, where _Pij represents the active power of device i when it operates independently in mode j, and P represents the active power of the unknown device type;

以视在功率为基础构建第四目标函数，其中，S_ij表示设备i处于j模式下独立运行时的视在功率，S代表未知设备类型的视在功率；The fourth objective function building unit is used according to the formula:

The fourth objective function is constructed based on the apparent power, where S _ij represents the apparent power of device i when it operates independently in mode j, and S represents the apparent power of the unknown device type;

以谐波为基础构建第五目标函数，其中，H_ij(k)表示设备i处于j模式下独立运行时的谐波，K是最大谐波的次序，H(k)代表未知包含设备类型的谐波；The fifth objective function building unit is used according to the formula:

The fifth objective function is constructed based on harmonics, where _Hij (k) represents the harmonics of device i when it operates independently in mode j, K is the order of the maximum harmonic, and H(k) represents the harmonics of the unknown containing device type;

表示所求用电设备类型，需要满足约束条件x_ij∈{0，1}和

The load decomposition model building unit is used to construct the above five objective functions into a multi-objective non-intrusive load decomposition model F(x) according to the formula: minimize F(x) = (f ₁ (x), f ₂ (x), f ₃ (x), f ₄ (x), f ₅ (x)), where:

represents the type of electrical equipment required, which needs to satisfy the constraints x _ij ∈ {0, 1} and

Accordingly, the determination module is specifically used to solve the multi-objective non-intrusive load decomposition model F(x) by using a multi-objective evolutionary algorithm to obtain the corresponding electrical equipment type and corresponding working mode when the five objective functions in F(x) are simultaneously minimized;

Among them, in the multi-objective evolutionary algorithm, the constraints are expressed in an encoding and decoding manner: each bit of the code represents a type of electrical equipment, where 0 means that the electrical equipment is turned off, and k means that the electrical equipment is in the kth working mode.

Through the above technical solutions, it can be known that compared with the prior art, the present invention discloses a non-invasive load decomposition method and device, wherein the non-invasive load decomposition method specifically includes obtaining the operation records of the electrical equipment and the power data on the bus, wherein the power data includes: current waveform; according to the operation records of the electrical equipment, pre-processing the power data to filter out the first power data; determining the four power parameter values corresponding to each moment according to the filtered first power data; establishing a multi-objective non-invasive load decomposition model with five optimization objectives according to the power parameter values and the current waveform; determining the type of electrical equipment and the corresponding mode working on the bus based on the multi-objective non-invasive load decomposition model. The non-invasive load decomposition method provided by the present invention has high decomposition accuracy, stable results, and strong robustness.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

FIG1 is a flow chart of a non-invasive load decomposition method provided by the present invention;

FIG2 is a flow chart of a method for preprocessing power data to select first power data provided by the present invention;

FIG3 is a flow chart of a method for removing outliers provided by the present invention;

FIG4 is a schematic diagram of a non-invasive load decomposition device provided by the present invention;

FIG5 is a schematic diagram of a preprocessing module provided by the present invention;

FIG6 is a schematic diagram of an outlier elimination unit provided by the present invention;

FIG7 is a schematic diagram of current waveforms and voltage waveforms obtained by sampling four types of electrical equipment provided by the present invention;

FIG8 is a schematic diagram of current and voltage waveforms before pretreatment provided by the present invention;

FIG. 9 is a schematic diagram of current and voltage waveforms after preprocessing provided by the present invention.

DETAILED DESCRIPTION

The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

Referring to FIG. 1 , an embodiment of the present invention discloses a non-invasive load decomposition method, which specifically includes the following steps:

S1: Acquire the operation records of the power-consuming equipment and the power data on the bus, wherein the power data includes: current, voltage, power factor, current waveform and voltage waveform;

S2: Preprocessing the power data to select first power data according to the operation record of the power-consuming device;

S3: Determine four power parameter values corresponding to each moment according to the screened first power data;

S4: establishing a multi-objective non-intrusive load decomposition model with five optimization objectives according to the power parameter value and the current waveform;

S5: Determine the type of electrical equipment working on the bus and the corresponding working mode based on the multi-objective non-intrusive load decomposition model.

The present invention provides a non-invasive load decomposition method with high decomposition accuracy, stable results and strong robustness.

Referring to FIG. 2 , in order to further optimize the above technical solution, step S2: pre-processing the power data to select the first power data according to the operation record of the power-consuming device specifically includes:

S21: dividing the power data according to the power equipment operation records according to the power equipment operation events to obtain a power data division result so that the power data in the same interval is in the same mode of the same device;

S22: Detect and remove abnormal values of current and voltage in the same interval to obtain an abnormal value removal result;

S23: According to the power data division result and the abnormal value elimination result, the average current and voltage in the interval where no electrical equipment is working are calculated. Under the same circuit, the average current and voltage in the interval where no electrical equipment is working are subtracted from the current and voltage values corresponding to all the intervals where useful electrical equipment is working, so as to obtain the first power data.

Referring to FIG. 3 , the specific method of the above step S22: detecting and eliminating abnormal values of current and voltage in the same interval further includes:

S221: Determine the median absolute deviation according to the formula MAD=median(|A _i -median(A)|), where A represents the current value or voltage value in the same interval, _Ai represents the current value or voltage value corresponding to the i-th moment, and MAD represents the median absolute deviation;

S222: Determine whether the value of the current value or voltage value in the same interval at any time deviates from the median by more than N times the median absolute deviation, and if so, determine that the current value or voltage value at this moment is an abnormal value and remove the abnormal value. Preferably, N=3.

In order to further optimize the above technical solution, step S3: determining four power parameter values corresponding to each moment according to the screened first power data, specifically including:

According to the formula: S = VI, determine the apparent power corresponding to each moment, where V represents the voltage value at a certain moment, I represents the current value at a certain moment, and S represents the apparent power;

According to the formula: P = VIcos (φ), determine the active power corresponding to each moment, where cos (φ) represents the power factor and P represents the active power;

According to the formula: Q = VIsin (φ), the reactive power corresponding to each moment is determined, where sin (φ) represents the sine φ of the phase difference between voltage and current, which has no specific physical meaning, and Q represents reactive power;

确定各时刻对应的谐波，其中，A(t)表示在t时刻电流值，约束条件是t＝0，1，2，…，T-1，T代表在一个电流波形周期内(例如：1/50S)采样点的数量，X(k)是第k次谐波的系数。According to the formula:

Determine the harmonics corresponding to each moment, where A(t) represents the current value at moment t, the constraints are t=0, 1, 2, ..., T-1, T represents the number of sampling points in one current waveform period (e.g.: 1/50S), and X(k) is the coefficient of the kth harmonic.

It should be noted here that the order of calculating the above four power parameter values is not limited, as long as the four power parameter values are calculated.

In order to further optimize the above technical solution, step S4: according to the power parameter value and the current waveform, a multi-objective non-intrusive load decomposition model with five optimization objectives is established, specifically including:

Five optimization objectives are constructed based on power parameter values and current waveforms; wherein the power parameter values include active power, reactive power, apparent power and harmonics; the specific method is as follows:

以电流波形为基础构建第一目标函数，其中，I_ij(t)表示在t时刻设备i处于j模式下独立运行时的电流值；T代表在一个电流波形周期(例如：1/50S)内采样点的数量，t＝0表示电压相位处于特定的时刻，此刻的电压波形正从最大值向最低值变化，N表示设备总数，M_i表示设备i包含的模式总数，I表示未知设备类型的组合电流波形；x_ij表示用电设备i处于工作模式j，其中i＝1，2，…，N，j＝1，2，…，M_i；According to the formula:

The first objective function is constructed based on the current waveform, where I _ij (t) represents the current value when the device i is operating independently in mode j at time t; T represents the number of sampling points in a current waveform cycle (e.g., 1/50S), t=0 represents the voltage phase at a specific moment, and the voltage waveform at this moment is changing from the maximum value to the minimum value, N represents the total number of devices, _Mi represents the total number of modes included in device i, and I represents the combined current waveform of unknown device type; x _ij represents that the power device i is in working mode j, where i=1, 2, ..., N, j=1, 2, ..., _Mi ;

以无功功率为基础构建第二目标函数，其中，Q_ij表示设备i处于j模式下独立运行时的无功功率，Q代表未知设备类型的无功功率；According to the formula:

The second objective function is constructed based on reactive power, where _Qij represents the reactive power of device i when it operates independently in mode j, and Q represents the reactive power of an unknown device type;

以有功功率为基础构建第三目标函数，其中，P_ij表示设备i处于j模式下独立运行时的有功功率，P代表未知设备类型的有功功率；According to the formula:

The third objective function is constructed based on active power, where _Pij represents the active power of device i when it operates independently in mode j, and P represents the active power of the unknown device type;

以视在功率为基础构建第四目标函数，其中，S_ij表示设备i处于j模式下独立运行时的视在功率，S代表未知设备类型的视在功率；According to the formula:

The fourth objective function is constructed based on the apparent power, where S _ij represents the apparent power of device i when it operates independently in mode j, and S represents the apparent power of the unknown device type;

以谐波为基础构建第五目标函数，其中，H_ij(k)表示设备i处于j模式下独立运行时的谐波，K是最大谐波的次序，H(k)代表未知包含设备类型的谐波。According to the formula:

The fifth objective function is constructed based on harmonics, where _Hij (k) represents the harmonics when device i operates independently in mode j, K is the order of the maximum harmonic, and H(k) represents the harmonics of the unknown contained device type.

It should be noted here that there is no particular order in constructing the above five objective functions. There is no limitation on the order of construction here, as long as the five objective functions are constructed before establishing the multi-objective intrusive load decomposition model.

表示所求用电设备类型，需要满足约束条件x_ij∈{0，1}和

According to the formula: minimize F(x) = (f ₁ (x), f ₂ (x), f ₃ (x), f ₄ (x), f ₅ (x)), the above five objective functions are constructed into a multi-objective non-intrusive load decomposition model F(x), where:

represents the type of electrical equipment required, which needs to satisfy the constraints x _ij ∈ {0, 1} and

Accordingly, step S5: determining the type of electrical equipment working on the bus and the corresponding working mode based on the multi-objective non-intrusive load decomposition model, specifically includes:

A multi-objective evolutionary algorithm is used to solve the multi-objective non-intrusive load decomposition model F(x) to obtain the corresponding electrical equipment type and corresponding working mode when the five optimization objectives in F(x) are minimized at the same time;

Among them, in the multi-objective evolutionary algorithm, the constraints are expressed in an encoding and decoding manner: each bit of the code represents a type of electrical equipment, where 0 means that the electrical equipment is turned off, and k means that the electrical equipment is in the kth working mode.

In addition, the embodiment of the present invention also discloses a non-invasive load decomposition device, referring to FIG4 , the device comprises:

The acquisition module 1 is used to acquire the operation records of the power-consuming equipment and the power data on the bus, wherein the power data includes: current waveform;

A preprocessing module 2, configured to preprocess the power data and select first power data according to the operation record of the power-consuming device;

A parameter calculation module 3, used to determine four power parameter values corresponding to each moment according to the screened first power data;

A model building module 4 is used to build a multi-objective non-intrusive load decomposition model with five optimization objectives according to the power parameter value and the current waveform;

The determination module 5 is used to determine the type and corresponding mode of the electrical equipment working on the bus based on the multi-objective non-intrusive load decomposition model.

In order to further optimize the above technical solution, the preprocessing module 2 specifically includes:

A division unit 21 is used to divide the power data according to each power device operation event according to the power device operation record, and obtain a power data division result so that the power data in the same interval is in the same mode of the same device;

The abnormal value elimination unit 22 is used to detect and eliminate abnormal values of the current and voltage in the same interval to obtain an abnormal value elimination result;

The calculation unit 23 is used to calculate the average current and voltage in the interval where no electrical equipment is working according to the power data division result and the abnormal value elimination result. Under the same circuit, the average current and voltage in the interval where no electrical equipment is working is subtracted from the current and voltage values corresponding to all intervals where useful electrical equipment is working, so as to obtain the first power data.

It should be noted here that the same circuit can be understood as the same electricity-using household.

In order to further optimize the above technical solution, the outlier elimination unit 22 specifically includes:

The calculation subunit 221 is used to determine the median absolute deviation according to the formula MAD=median(|A _i -median(A)|), where A represents the current value or voltage value in the same interval, _Ai represents the current value or voltage value corresponding to the i-th moment, and MAD represents the median absolute deviation;

The elimination subunit 222 is used to determine whether the value of the current value or voltage value deviating from the median in the same interval at any time is greater than N times the median absolute deviation. If so, the current value or voltage value at this moment is an abnormal value and the abnormal value is eliminated. Preferably, N=3.

In order to further optimize the above technical solution, the parameter calculation module 3 specifically includes:

The apparent power calculation unit is used to determine the apparent power corresponding to each moment according to the formula: S=VI, where V represents the voltage value at a certain moment, I represents the current value at a certain moment, and S represents the apparent power;

An active power calculation unit, used to determine the active power corresponding to each moment according to the formula: P = VIcos (φ), where cos (φ) represents the power factor and P represents the active power;

A reactive power calculation unit, used to determine the reactive power corresponding to each moment according to the formula: Q=VIsin(φ), where Q represents the reactive power;

确定各时刻对应的谐波，其中，A(t)表示在t时刻电流值，约束条件是t＝0，1，2，…，T-1，T代表在一个电流波形周期内采样点的数量，X(k)是第k次谐波的系数。Harmonic calculation unit, used according to the formula:

Determine the harmonics corresponding to each moment, where A(t) represents the current value at moment t, the constraint is t=0, 1, 2, ..., T-1, T represents the number of sampling points in one current waveform cycle, and X(k) is the coefficient of the kth harmonic.

In order to further optimize the above technical solution, the model building module 4 specifically includes:

以电流波形为基础构建第一目标函数，其中，I_ij(t)表示在t时刻设备i处于j模式下独立运行时的电流值；T代表在一个电流波形周期内采样点的数量，t＝0表示电压相位处于特定的时刻，此刻的电压波形正从最大值向最低值变化，N表示设备总数，M_i表示设备i包含的模式总数，I表示未知设备类型的组合电流波形；x_ij表示用电设备i处于工作模式j，其中i＝1，2，…，N，j＝1，2，…，M_i；The first objective function building unit is used according to the formula:

The first objective function is constructed based on the current waveform, where I _ij (t) represents the current value when the device i is operating independently in mode j at time t; T represents the number of sampling points in a current waveform cycle, t=0 represents that the voltage phase is at a specific moment, and the voltage waveform at this moment is changing from the maximum value to the minimum value, N represents the total number of devices, _Mi represents the total number of modes included in device i, and I represents the combined current waveform of unknown device types; x _ij represents that the power device i is in working mode j, where i=1, 2, ..., N, j=1, 2, ..., _Mi ;

以无功功率为基础构建第二目标函数，其中，Q_ij表示设备i处于j模式下独立运行时的无功功率，Q代表未知设备类型的无功功率；The second objective function building unit is used according to the formula:

The second objective function is constructed based on reactive power, where _Qij represents the reactive power of device i when it operates independently in mode j, and Q represents the reactive power of an unknown device type;

以有功功率为基础构建第三目标函数，其中，P_ij表示设备i处于j模式下独立运行时的有功功率，P代表未知设备类型的有功功率；The third objective function building unit is used according to the formula:

The third objective function is constructed based on active power, where _Pij represents the active power of device i when it operates independently in mode j, and P represents the active power of the unknown device type;

以视在功率为基础构建第四目标函数，其中，S_ij表示设备i处于j模式下独立运行时的视在功率，S代表未知设备类型的视在功率；The fourth objective function building unit is used according to the formula:

The fourth objective function is constructed based on the apparent power, where S _ij represents the apparent power of device i when it operates independently in mode j, and S represents the apparent power of the unknown device type;

以谐波为基础构建第五目标函数，其中，H_ij(k)表示设备i处于j模式下独立运行时的谐波，K是最大谐波的次序，H(k)代表未知包含设备类型的谐波。The fifth objective function building unit is used according to the formula:

The fifth objective function is constructed based on harmonics, where _Hij (k) represents the harmonics when device i operates independently in mode j, K is the order of the maximum harmonic, and H(k) represents the harmonics of the unknown contained device type.

表示所求用电设备类型，需要满足约束条件x_ij∈{0，1}和

The load decomposition model building unit is used to construct the above five objective functions into a multi-objective non-intrusive load decomposition model F(x) according to the formula: minimize F(x) = (f ₁ (x), f ₂ (x), f ₃ (x), f ₄ (x), f ₅ (x)), where:

represents the type of electrical equipment required, which needs to satisfy the constraints x _ij ∈ {0, 1} and

Accordingly, the determination module 5 is specifically used to solve the multi-objective non-intrusive load decomposition model F(x) using a multi-objective evolutionary algorithm to obtain the corresponding electrical equipment type and corresponding working mode when the five objective functions in F(x) are simultaneously minimized;

Among them, in the multi-objective evolutionary algorithm, the constraints are expressed in an encoding and decoding manner: each bit of the code represents a type of electrical equipment, where 0 means that the electrical equipment is turned off, and k means that the electrical equipment is in the kth working mode.

It should be noted here that there are many known existing multi-objective algorithms, such as the Reference Vector Guided Evolutionary Algorithm (RVEA), but there are few reports on using multi-objective evolutionary algorithms to solve the multi-objective optimization problem.

In order to further discuss in detail the non-intrusive load decomposition method and device provided by the present invention, taking 241 valid records of bus power data of a household as an example, the technical solution of the present invention is introduced in detail.

The embodiment of the present invention obtains four power parameter values based on current, voltage, power factor, current waveform and voltage waveform, constructs five optimization objectives using the obtained power parameter values and current waveform, proposes a multi-objective non-intrusive load decomposition model, and solves the model in combination with a multi-objective evolutionary algorithm to determine the type of electrical equipment working on the bus and the corresponding working mode.

Step 1: First, obtain the operation records of the power consumption equipment and the power data on the bus, including current, voltage, power factor, current waveform and voltage waveform. According to the operation records of the power consumption equipment, pre-process the power data to filter out reasonable data, and finally determine the four power parameter values corresponding to each moment according to the filtered power data.

The three macro power parameters collected include current, voltage and power factor, and the sampling rate of the above parameters is 1Hz; the micro load features obtained are current waveform and voltage waveform, and the sampling rate of the above parameters in the 50Hz system is 6400Hz. In order to reduce the storage space, only the first 1/50 second of the waveform is used, as shown in Figure 7. A total of 16 working modes of 8 electrical appliances were tested. The 8 electrical appliances are fans, kettles, TVs, incandescent lamps, printers, computers, water dispensers and hair dryers as shown in Table 1. Dataset A only collects the situation where only one electrical appliance is running; Dataset B collects the situation where multiple electrical appliances (less than 4) are running at the same time. Using Dataset A, a 5-dimensional feature vector of 16 modes can be obtained as a known load database; Dataset B is used to verify the load decomposition method we proposed.

Table 18 electrical appliances with 16 modes corresponding to current, voltage and power factor

Using only one load characteristic cannot distinguish all electrical equipment. Five load characteristics are used to decompose the load, including macro characteristics (i.e. active power, reactive power and apparent power) and micro characteristics (i.e. current waveform and harmonics). The definitions of these characteristics are as follows:

Characteristic superposition criterion: The load characteristics used in the proposed load decomposition model must meet the characteristic superposition criterion. The detailed definition of the characteristic superposition criterion is as follows:

表示用电设备a处模式j下产生特征l对应的值；如果在t+Δt时刻用电设备a处于j模态瞬时工作且满足等式1，则表明特征l满足特征叠加准则。如果特征l满足特征叠加准则，则此特征可以用来估计叠加负荷特征即超多个电器同时运行的负荷特征值。Wherein, Ψ _l (t) is the superposition value of feature l generated by K electrical devices running simultaneously at time t,

Indicates the value corresponding to feature l generated by electrical device a in mode j; if electrical device a is in instantaneous operation in mode j at time t+Δt and satisfies equation 1, it means that feature l satisfies the feature superposition criterion. If feature l satisfies the feature superposition criterion, this feature can be used to estimate the superimposed load feature, that is, the load feature value of multiple electrical appliances running simultaneously.

The calculation formulas for active power, reactive power and apparent power are as follows:

P＝VIcos(φ), (2)

Q＝VIsin(φ), (3)

S＝VI， (4)

Where V and I are voltage and current respectively, and cosφ is the power factor. The above characteristics all meet the characteristic superposition criterion.

Current waveform: The current waveform is sampled with 128 data points in one cycle (1/50s). Due to the high sampling rate, the current waveform contains detailed electrical characteristics, but there is no obvious difference in the voltage waveform as shown in Figure 7(b). Therefore, the current waveform is selected and it meets the feature superposition criterion.

Harmonics: Perform fast Fourier transform on the current waveform to obtain current harmonics. The rectangular coordinate form of harmonics (a+jb) meets the characteristic superposition criterion. However, the physically meaningful polar coordinate form A∠θ does not meet the characteristic superposition criterion. The calculation formula for the rectangular coordinate form of harmonics is as follows:

Among them, A(t) represents the current value at time t; the constraint condition is t = 0, 1, 2, ..., T-1; T represents the number of sampling points in one current waveform period (1/50s); X(k) is the coefficient of the kth harmonic. Table 3 shows the power parameter values determined based on the power data, and this is the macroscopic feature before preprocessing. Harmonics are calculated after preprocessing.

Table 2 Active power, apparent power and reactive power corresponding to 16 modes of 8 electrical appliances before pretreatment

The data will then be preprocessed, including: (1) using the operation records of electrical equipment to separate the switching events of electrical equipment; (2) sampling the current waveform and voltage waveform at the same time to ensure that the initial phase of the voltage waveform (changing from the positive peak to the trough) is the same at each moment, and at the same time, the current waveform and the voltage waveform must have a correct phase relationship; (3) removing outliers to ensure the accuracy of subsequent processing; (4) subtracting the load characteristics of electrical equipment from the load characteristics of electrical equipment in operation to reduce the impact of noise. Some of the processed data are shown in Table 3.

Table 3 Active power, apparent power and reactive power corresponding to 16 modes of 8 electrical appliances after preprocessing

Step 2: Based on the five power parameter values, a multi-objective load decomposition model with five optimization objectives is established. Finally, a multi-objective evolutionary algorithm is used to solve the proposed multi-objective load decomposition model to determine the type and corresponding mode of electrical equipment currently running on the bus.

The load decomposition problem is converted into an optimization problem. Each feature can form an optimization problem. The traditional optimization-based load decomposition method is to weight multiple objective functions into one. However, due to the high positive or negative correlation between some features, different optimization problems cannot be simply weighted; and the weighting parameters are sensitive to different data sets and difficult to adjust. The weighting parameters have a significant impact on the accuracy of load decomposition. In order to solve these shortcomings, the load decomposition problem is converted into a multi-objective optimization problem, which can be expressed as follows:

minimize F(x)=(f ₁ (x), f ₂ (x), f ₃ (x), f ₄ (x), f ₅ (x)),

The constraints are:

_xij∈ {0，1}，

Among them, F(x) table is a five-dimensional objective function that must be optimized simultaneously; _xij represents the operating state of electrical equipment i in mode j (0 means off, 1 means on), N is the maximum number of electrical equipment in the database, and _Mi represents the total number of modes of electrical equipment i. These five optimization problems are constructed using five different load characteristics. The calculation formulas of the five objective functions are as follows:

I _ij (t) represents the current value when the electrical device i is operating independently in mode j at time t; T represents the number of sampling points in one current waveform cycle (1/50s); t=0 represents that the voltage phase is at a specific moment, and the voltage waveform at this moment is changing from the maximum value to the minimum value; N represents the total number of electrical devices; _Mi represents the total number of modes included in the electrical device i; I represents the current waveform of an unknown electrical device type; _Qij represents the reactive power of the device i when operating independently in mode j; Q represents the reactive power of an unknown electrical device type; _Pij represents the active power of the device i when operating independently in mode j; P represents the active power of an unknown electrical device type; _Sij represents the apparent power of the device i when operating independently in mode j; S represents the apparent power of an unknown electrical device type; _Hij (k) represents the harmonics of the device i when operating independently in mode j; K is the order of the maximum harmonic, and H(k) represents the harmonics of an unknown electrical device type.

新的编码方式的每一位表示一种类型用电设备，其中，0代表此用电设备关，k表示此用电设备处于第k工作模式。然后，遗传算子(位突变算子和单点交叉算子)被用来生成子代。之后，使用分配目标等级计算目标等级R，和利用多目标进化算法适应值评估用于获取适应值，同时利用这些适应值从同目标等级中筛选出下一代。根据目标等级去筛选子代，然后根据多目标进化算法得到的适应值从同目标等级中做更进一步的筛选，最终满足停止条件输出最终种群。A multi-objective evolutionary algorithm is used to solve the multi-objective load decomposition model to obtain the working status of each electrical equipment. First, a new encoding method is proposed to solve the constraints x _ij ∈ {0, 1} and

Each bit of the new encoding method represents a type of electrical equipment, where 0 represents that the electrical equipment is off, and k represents that the electrical equipment is in the kth working mode. Then, genetic operators (bit mutation operator and single-point crossover operator) are used to generate offspring. After that, the target level R is calculated using the assigned target level, and the fitness value evaluation using the multi-objective evolutionary algorithm is used to obtain the fitness value, and these fitness values are used to screen the next generation from the same target level. The offspring are screened according to the target level, and then further screened from the same target level according to the fitness value obtained by the multi-objective evolutionary algorithm, and finally the final population is output when the stopping condition is met.

Assign target level: A method of assigning target level is proposed to solve the constraint of the upper limit of power consumption equipment. First, the number of power consumption equipment of individuals is calculated, and then individuals with 1 to U power consumption equipment are assigned to R ₁ . Individuals with N power consumption equipment are assigned to R _N-U+2 .

The above technical solution is further explained below with reference to specific examples, see Table 4 and Table 5:

Table 4 Active power, apparent power, reactive power, harmonics and current waveform values corresponding to 4 modes of 3 electrical appliances after preprocessing

Table 5 Analysis and acquisition of active power, apparent power, reactive power, harmonics and current waveform values on the bus

Active power(W) Apparent power(W) Reactive power(W) Harmonics (A) Current waveform (A) 1710 1720 87 30 3

Table 4 shows the corresponding active power, apparent power, reactive power, harmonics and current waveform values of 4 modes of 3 kinds of electrical appliances. Table 5 shows the active power, apparent power, reactive power, harmonics and current waveform values measured at a certain moment on the bus and obtained after processing in step 1. After solving the multi-objective evolutionary algorithm, the types of electrical equipment and corresponding modes that make all 5 objective functions optimal are obtained; it is determined that at this moment, kettle mode 1 and printer mode 2 are in operation.

The present invention provides a non-intrusive load monitoring method based on a multi-objective evolutionary algorithm. According to the operation records of the power equipment, the power data including current, voltage, power factor, current waveform and voltage waveform are pre-processed to screen out reasonable data; according to the screened power data, the four power parameter values corresponding to each moment are determined, including useful power, reactive power, apparent power and harmonics, and the power parameter values and current waveforms are used to establish a multi-objective non-intrusive load decomposition model with five optimization objectives. The multi-objective evolutionary algorithm is used to solve the multi-objective load decomposition model to determine the type of power equipment working on the bus and the corresponding mode. Therefore, the load monitoring method provided by the present invention has high decomposition accuracy, stable results and strong robustness.

In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims (6)

1. A non-invasive load decomposition method, characterized in that it comprises the following steps: Acquiring operation records of power-consuming equipment and power data on the bus, wherein the power data includes: current waveform; Preprocessing the power data to select first power data according to the power-consuming equipment operation record; Determine four power parameter values corresponding to each moment according to the filtered first power data; According to the power parameter value and the current waveform, a multi-objective non-intrusive load decomposition model with five optimization objectives is established; Determining the type of electrical equipment working on the bus and the corresponding working mode based on the multi-objective non-intrusive load decomposition model; The determining of four power parameter values corresponding to each moment according to the filtered first power data specifically includes: According to the formula: S = VI, determine the apparent power corresponding to each moment, where V represents the voltage value at a certain moment, I represents the current value at a certain moment, and S represents the apparent power; According to the formula: P = VIcos (φ), determine the active power corresponding to each moment, where cos (φ) represents the power factor and P represents the active power; According to the formula: Q = VIsin (φ), determine the reactive power corresponding to each moment, where Q represents the reactive power; According to the formula:

Determine the harmonics corresponding to each moment, where A(t) represents the current value at moment t, the constraint is t=0, 1, 2, ..., T-1, T represents the number of sampling points in one current waveform cycle, and X(k) is the coefficient of the kth harmonic; According to the power parameter value and the current waveform, a multi-objective non-intrusive load decomposition model with five optimization objectives is established, specifically including: Five optimization objectives are constructed based on power parameter values and current waveforms; wherein the power parameter values include active power, reactive power, apparent power and harmonics; the specific method is as follows: According to the formula:

The first objective function is constructed based on the current waveform, where I _ij (t) represents the current value when the device i is operating independently in mode j at time t, T represents the number of sampling points in a current waveform cycle, t=0 represents that the voltage phase is at a specific moment, and the voltage waveform at this moment is changing from the maximum value to the minimum value, N represents the total number of devices, _Mi represents the total number of modes included in the power device i, I represents the combined current waveform of the unknown device type, and x _ij represents that the power device i is in working mode j, where i=1, 2, ..., N, j=1, 2, ..., _Mi ; According to the formula:

The second objective function is constructed based on reactive power, where _Qij represents the reactive power of device i when it operates independently in mode j, and Q represents the reactive power of an unknown device type; According to the formula:

The third objective function is constructed based on active power, where _Pij represents the active power of device i when it operates independently in mode j, and P represents the active power of the unknown device type; According to the formula:

The fourth objective function is constructed based on the apparent power, where S _ij represents the apparent power of device i when it operates independently in mode j, and S represents the apparent power of the unknown device type; According to the formula:

The fifth objective function is constructed based on harmonics, where _Hij (k) represents the harmonics of device i when it operates independently in mode j, K is the order of the maximum harmonic, and H(k) represents the harmonics of the unknown containing device type; According to the formula: minimize F(x) = (f ₁ (x), f ₂ (x), f ₃ (x), f ₄ (x), f ₅ (x)), the above five objective functions are constructed into a multi-objective non-intrusive load decomposition model F(x), where:

represents the type of electrical equipment required, which needs to satisfy the constraints x _ij ∈ {0, 1} and

Accordingly, determining the type of electrical equipment working on the bus and the corresponding working mode based on the multi-objective non-intrusive load decomposition model specifically includes: A multi-objective evolutionary algorithm is used to solve the multi-objective non-intrusive load decomposition model F(x) to obtain the corresponding electrical equipment type and corresponding working mode when the five optimization objectives in F(x) are minimized at the same time; Among them, in the multi-objective evolutionary algorithm, the constraints are expressed in an encoding and decoding manner: each bit of the code represents a type of electrical equipment, where 0 means that the electrical equipment is turned off, and k means that the electrical equipment is in the kth working mode. 2. A non-intrusive load decomposition method according to claim 1, characterized in that, according to the operation record of the electrical equipment, pre-processing the power data to select the first power data specifically comprises: According to the operation records of the power-consuming equipment, the power data is divided according to the operation events of each power-consuming equipment to obtain a power data division result, so that the power data in the same interval is in the same mode of the same equipment; Perform outlier detection and elimination on the current and voltage in the same interval to obtain an outlier elimination result; According to the power data division results and abnormal value elimination results, the average current and voltage in the interval where no electrical equipment is working are calculated. Under the same circuit, the average current and voltage in the interval where no electrical equipment is working are subtracted from the current and voltage values corresponding to all intervals where useful electrical equipment is working to obtain the first power data. 3. A non-intrusive load decomposition method according to claim 2, characterized in that the specific method of detecting and eliminating abnormal values of current and voltage in the same interval comprises: According to the formula MAD=median(|A _i -median(A)|), the median absolute deviation is determined, where A represents the current value or voltage value in the same interval, _Ai represents the current value or voltage value corresponding to the i-th moment, and MAD represents the median absolute deviation; Determine whether the value of the current value or voltage value in the same interval at any time deviates from the median by more than N times the median absolute deviation. If so, determine that the current value or voltage value at this moment is an abnormal value and remove the abnormal value. 4. A non-invasive load decomposition device, comprising: An acquisition module is used to acquire operation records of power-consuming equipment and power data on the bus, wherein the power data includes: current waveform; A preprocessing module, configured to preprocess the power data to select first power data according to the operation record of the power-consuming device; A parameter calculation module, used to determine four power parameter values corresponding to each moment according to the screened first power data; A model building module, used for building a multi-objective non-intrusive load decomposition model with five optimization objectives according to the power parameter value and the current waveform; A determination module, used to determine the type and corresponding mode of the electrical equipment working on the bus based on the multi-objective non-intrusive load decomposition model; The parameter calculation module specifically includes: The apparent power calculation unit is used to determine the apparent power corresponding to each moment according to the formula: S=VI, where V represents the voltage value at a certain moment, I represents the current value at a certain moment, and S represents the apparent power; An active power calculation unit, used to determine the active power corresponding to each moment according to the formula: P = VIcos (φ), where cos (φ) represents the power factor and P represents the active power; A reactive power calculation unit, used to determine the reactive power corresponding to each moment according to the formula: Q = VIsin (φ), where Q represents the reactive power; Harmonic calculation unit, used according to the formula:

Determine the harmonics corresponding to each moment, where A(t) represents the current value at moment t, the constraint is t=0, 1, 2, ..., T-1, T represents the number of sampling points in one current waveform cycle, and X(k) is the coefficient of the kth harmonic; The model building module specifically includes: The first objective function building unit is used according to the formula:

The first objective function is constructed based on the current waveform, where I _ij (t) represents the current value when the device i is operating independently in mode j at time t; T represents the number of sampling points in a current waveform cycle, t=0 represents that the voltage phase is at a specific moment, and the voltage waveform at this moment is changing from the maximum value to the minimum value, N represents the total number of devices, _Mi represents the total number of modes included in the power device i, and I represents the combined current waveform of the unknown device type; x _ij represents that the power device i is in working mode j, where i=1, 2, ..., N, j=1, 2, ..., _Mi ; The second objective function building unit is used according to the formula:

The second objective function is constructed based on reactive power, where _Qij represents the reactive power of device i when it operates independently in mode j, and Q represents the reactive power of an unknown device type; The third objective function building unit is used according to the formula:

The third objective function is constructed based on active power, where _Pij represents the active power of device i when it operates independently in mode j, and P represents the active power of the unknown device type; The fourth objective function building unit is used according to the formula:

The fourth objective function is constructed based on the apparent power, where S _ij represents the apparent power of device i when it operates independently in mode j, and S represents the apparent power of the unknown device type; The fifth objective function building unit is used according to the formula:

The fifth objective function is constructed based on harmonics, where _Hij (k) represents the harmonics of device i when it operates independently in mode j, K is the order of the maximum harmonic, and H(k) represents the harmonics of the unknown containing device type; The load decomposition model building unit is used to construct the above five objective functions into a multi-objective non-intrusive load decomposition model F(x) according to the formula: minimize F(x) = (f ₁ (x), f ₂ (x), f ₃ (x), f ₄ (x), f ₅ (x)), where:

represents the type of electrical equipment required, which needs to satisfy the constraints x _ij ∈ {0, 1} and

Accordingly, the determination module is specifically used to solve the multi-objective non-intrusive load decomposition model F(x) by using a multi-objective evolutionary algorithm to obtain the corresponding electrical equipment type and corresponding working mode when the five objective functions in F(x) are simultaneously minimized; Among them, in the multi-objective evolutionary algorithm, the constraints are expressed in an encoding and decoding manner: each bit of the code represents a type of electrical equipment, where 0 means that the electrical equipment is turned off, and k means that the electrical equipment is in the kth working mode. 5. A non-invasive load decomposition device according to claim 4, characterized in that the pre-processing module specifically comprises: A division unit, configured to divide the power data according to each power-consuming device operation event according to the power-consuming device operation record, and obtain a power data division result so that the power data in the same interval is in the same mode of the same device; An outlier elimination unit is used to detect and eliminate outliers of current and voltage in the same interval to obtain an outlier elimination result; The calculation unit is used to calculate the average current and voltage in the interval where no electrical equipment is working according to the power data division result and the abnormal value elimination result. Under the same circuit, the average current and voltage in the interval where no electrical equipment is working is subtracted from the current and voltage values corresponding to all intervals where useful electrical equipment is working, so as to obtain the first power data. 6. A non-intrusive load decomposition device according to claim 5, characterized in that the abnormal value removal unit specifically comprises: A calculation subunit, used to determine the median absolute deviation according to the formula MAD=median(|A _i -median(A)|), where A represents the current value or voltage value in the same interval, _Ai represents the current value or voltage value corresponding to the i-th moment, and MAD represents the median absolute deviation; The elimination subunit is used to determine whether the value of the current value or voltage value deviating from the median in the same interval at any time is greater than N times the median absolute deviation. If so, the current value or voltage value at this moment is an abnormal value and the abnormal value is eliminated.

Images