Electrical appliance working state identification method based on particle swarm optimization algorithm

Abstract

The present invention discloses an electrical appliance working state identification method based on a particle swarm optimization algorithm, and specifically relates to the technical field of electrical appliance working state identification. The present invention regards the identification problem of multiple electrical appliances and multiple working states as a combinatorial optimization problem. On the basis of representing the characteristics of each working state of each electrical appliance using multiple parameters in the user's power consumption situation data, the combinatorial optimization problem of identifying multiple electrical appliances and multiple working states is solved using a particle swarm optimization algorithm, thereby achieving accurate identification of the working state of each electrical appliance on the circuit.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical appliance working state identification, and in particular to an electrical appliance working state identification method based on a particle swarm optimization algorithm. Background Art

[0002] Electricity is a necessity for modern production and life, and is an irreplaceable form of energy in modern society. With the popularization of Internet of Things technology, the fine monitoring and management of the user-side electricity consumption process has attracted increasing attention in more and more research and applications. At present, most of the electricity consumption-related research based on the user demand side conducts steady-state analysis, transient analysis, and a hybrid analysis of the two on electrical appliances based on characteristics such as voltage, current, and power, and mostly uses smart meters to collect user electricity usage status data. When using smart meters to perceive household electricity usage status, the granularity of the user's electricity usage status is large, which is not enough to support the timely identification of the type of electrical appliances used by users and the operating status of electrical appliances. The timely identification of the type of electrical appliances used by users and the operating status of electrical appliances requires fine-grained online perception of the user's electricity usage status. Online monitoring of users' electricity usage with fine-grained perception can not only provide a more detailed understanding of electricity usage on the user side, but also help improve the safety of users' electricity usage and energy efficiency management: on the one hand, online monitoring and fine-grained perception of users' electricity usage can enable users to timely grasp the operating status of electrical appliances and discover abnormal problems in a timely manner, ensuring the safety of users' electricity use; on the other hand, online monitoring and fine-grained perception of users' electricity usage can also help to more accurately grasp the user's electricity usage process, providing detailed basis support for optimizing users' electricity usage behavior and adjusting their electricity usage habits.

[0003] According to different data collection methods, the monitoring methods of electrical appliance operation status can be divided into two categories: intrusive appliance load monitoring (ILM) and nonintrusive appliance load monitoring (NALM). Intrusive appliance status monitoring methods use devices such as smart sockets to collect power consumption data of a single appliance and identify the operation status of the appliance in a timely manner. The ILM method mainly focuses on the study of electrical current and power. Ridi A, Gisler C, Hennebert J. Automatic identification of electrical appliances using smart plugs, Systems, 2013 8th IEEE International Workshop on Signal Processing and their Applications, 2013: 301-305. A typical ILM system architecture and typical feature extraction and machine learning algorithms commonly used in ILM systems are introduced. A method for electrical equipment identification based on smart sockets is proposed. The method extracts the dynamic features of the current sequence through the first-order derivative and second-order derivative of the power consumption sequence, and uses the K nearest neighbor algorithm and Gaussian mixture model (GMM) to achieve appliance classification. Yan D, Jin Y, Sun HS, Dong B, Ye Z, Li ZX, Yuan YP. Household appliance recognitionthrough a Bayes classification model. Sustainable cities and society, 2019, v46. Smart sockets are used to collect electrical appliance power consumption data, and the power consumption behavior and energy consumption characteristics of typical household appliances are summarized and extracted from the load power sequence samples. A household appliance recognition algorithm based on the Bayes classification model is proposed.

[0004] The NALM method collects the voltage, current or power of the total load at the power input, and decomposes it to obtain the operating status of each electrical device. Generally, the power data is collected using devices such as electric meters. Hart G W. Nonintrusive appliance load monitoring [J]. Proceedings of the IEEE, 1993, 80 (12): 1870-1891. The concept of non-intrusive appliance load monitoring was first proposed. Based on the detailed analysis of the current and voltage of the total load measured at the power interface, the energy consumption of a single appliance turned on and off in a power load is determined. Kong WC, Dong ZY, Hill DJ, Ma J, Zhao JH, Luo FJ. A Hierarchical Hidden Markov Model Framework for Home Appliance Modeling [J]. IEEE Transactions on Smart Grid, 2018, 9 (4): 3079-3090. A hierarchical hidden Markov model (HHMM) framework is proposed to model home appliances, which provides a better representation for appliances with multiple modes and different energy consumption (such as washing machines and dishwashers). Himeur Y, Alsalemi A, Bensaali F, Amira A. Robust event-based non-intrusive appliance recognition using multi-scale wavelet packet tree and ensemble bagging tree [J]. Applied Energy, 2020, 267 A new non-intrusive appliance recognition system is proposed, which has the following features: (i) a novel and reliable event detection scheme is proposed, (ii) multi-scale wavelet transform is used for feature extraction, and (iii) ensemble bagging tree classifier is adopted for classification. Chen JF, Wang X, Zhang XT, Zhang WH. Temporal and Spectral Feature Learning With Two-Stream Convolutional Neural Networks for Appliance Recognition in NILM[J]. IEEE Transactions on Smart Grid, 2022, 13(1): 762-772. A two-stream convolutional neural network (TSCNN) is proposed to extract features from two types of load features, temporal and spectral, and perform classification.Aslan M and Zurel EN, An efficient hybrid model for appliances classification based on time series features [J]. Energy and Buildings, 2022. 7 (266) Ten time domain and steady-state features were extracted from the electrical current waveform, and the k-nearest neighbor rule (k-NNR) and support vector machine were used for load identification. The principal component analysis was combined with k-NNR and SVM to test the 10 features respectively and select the most effective features. Liu H; Wu HP and Yu CM, A hybrid model for appliance classification based on time series features, Energy and Buildings [J]. Energy and Buildings, 2019. 8 (196): 112-123 proposed to use S transform to extract equipment features, and then input these features into the SVM classifier to identify each electrical device.

[0005] At present, in the research and application of NALM-based electrical appliance working state identification, power data collected at low frequency is usually used. This method may confuse the working states of multiple electrical appliances on the circuit into the working state of a certain electrical appliance. When there are many types of electrical appliances connected to the circuit or the power is similar, the accuracy of identifying the working state of each electrical appliance is not high. Summary of the invention

[0006] In order to timely and accurately identify the working status of multiple electrical appliances, based on the high-frequency collected user power usage status data, the present invention regards the identification problem of multiple working status of multiple electrical appliances as a combinatorial optimization problem. On the basis of representing the characteristics of each working status of each electrical appliance with multiple parameters in the user power usage status data, the particle swarm optimization algorithm is used to solve the combinatorial optimization problem of identifying multiple working status of multiple electrical appliances, and the accurate identification of the working status of each electrical appliance in the circuit is realized.

[0007] The present invention provides the following technical solution: an electrical appliance working state identification method based on a particle swarm optimization algorithm, assuming a1, a2, ..., a m It is the parameters related to the power consumption process perceived in the circuit, such as current, voltage, active power, reactive power, apparent power, apparent power factor, power, etc., called s = {a1, a2, ..., a m} is the observable user electricity consumption situation, referred to as user electricity consumption situation; the electricity consumption situation s at time t is expressed as s(t) = (a1(t), a2(t), ..., am (t));

[0008] Let X = {X1, X2…X n} is the set of all n electrical appliances in the circuit. j∈[1,n] and j is an integer, let is the set of all possible working states. is the summary set of all working states of all n electrical appliances, and I is called the working state set of all electrical appliances;

[0009] If at time t there is only device X in the circuit j Working and its working status is e j,u , if the power consumption situation at this moment is s(t)=(a1(t),a2(t),...,a m (t)), record k∈[1,m] and k is an integer, s(t)=s(t,e j,u ),say Yes j,u The observed value of the basic working status at time t;

[0010] If the appliance X j Working status j,u Occurrence, regulation c j,u (t)=1, otherwise c j,u (t) = 0, called c j,u (t) is the working state at time t j,u The occurrence coefficient of is the set of working state occurrence coefficients of I at time t; although electrical appliance X j The working status is e j,1 、e j,2 ...e j,kj There are many possibilities, but at a specific moment t, appliance X j The working status is only k j One of the working states, namely

[0011] Considering Appliance X j Working status j,u The basic working situation is electrical appliances X j An attribute feature of Then b(e j,u ) During the observation period, only appliance X j Working and its working status is e j,u When all s(t,e j,u ) as its estimate;

[0012] Let B = {b(e j,u)|e j,u ∈I}, B is the set of basic working states of all electrical appliances in the circuit. Ideally, the user power consumption state s(t) observed in the circuit at time t is the combination of all basic working states as shown in formula (1);

[0013]

[0014] In formula (1), c j,u ∈C;

[0015] According to formula (1), the problem of identifying the working state of the electrical appliance based on the user's power consumption situation s(t) observed in the circuit at time t is essentially the problem of solving the working state coefficient set C(t) of I at time t, and formula (2) gives a model for solving the working state coefficient set C(t) of I at time t in the form of an optimization problem; the particle swarm optimization PSO algorithm is used to solve the optimization problem described in formula (2), and then the working state of the electrical appliance on the circuit is identified according to the optimal value of the working state coefficient set C(t), where t is used for general description, representing any time; if it is t0, it is a specific time.

[0016] Preferably, the process of using the particle swarm optimization PSO algorithm to identify the working status of electrical appliances on the circuit is as follows:

[0017] In the PSO algorithm, a particle is a feasible solution to the corresponding optimization problem; when implementing the PSO algorithm in a microenvironment network, each particle of the PSO algorithm must be assigned to a microenvironment, and each microenvironment has n particles, assuming that the number of particles n in each microenvironment is the same;

[0018] For the PSO algorithm implemented in the micro-environment network, assume:

[0019] 1) There are n particles in each microenvironment;

[0020] 2) For any microenvironment, record gBest L is the historical optimal position of all particles in the microenvironment;

[0021] 3) For a microenvironment, if it has m neighboring microenvironments, record gBest Li is the optimal position of all particles in its i-th neighbor microenvironment, and gBestA is defined as formula (3);

[0022] gBest A = min(gBest L ,gBest L1 ,gBest L2 …gBest Lm )(3)

[0023] 4) For the PSO algorithm based on microenvironment, for any particle in a microenvironment, if its position at time t is x(t) and its velocity is v(t), then its velocity v(t+1) and position x(t+1) at time t+1 are determined according to equations (4) and (5);

[0024] v(t+1)=ω×v(t)+c1×rand()×(pBest-x(t))+c2×rand()×(gBest A -x(t))(4)

[0025] x(t+1)=x(t)+v(t+1)(5)

[0026] Among them, formula (1) is the data exchange strategy of the PSO algorithm based on the microenvironment, while formulas (4) and (5) are the particle position update strategies in the PSO algorithm based on the microenvironment;

[0027] Step 1: Initialize the speed and position of each local particle;

[0028] Step 2: Take the initial position of each particle as the initial value of its historical optimal position:

[0029] Step 3: Calculate the optimal position gBest of all local particles L ;

[0030] The specific implementation steps of Step1-Step3 are as follows:

[0031] Input the user's power consumption situation s(t0), the working state set of all appliances I, the basic working state set B, the number of local particles n, the particle position dimension varLen, the learning factors c1 and c2, the inertia factor ω, the maximum number of iterations maxGen, the fitness function fun, and the neighbor microenvironment node list pNeighbor;

[0032] Set the particle dimension varLen to the number of elements in B;

[0033] The number of iterations k starts from 1, and it is determined whether the number of iterations k is less than the maximum number of iterations maxGen parameter to control whether the step of realizing the generation coefficient set C(t0) of the electrical appliance working state continues to be iteratively executed;

[0034] The steps to realize the electrical appliance working state occurrence coefficient set C(t0) are as follows:

[0035] For each neighbor micro-environment node in pNeighbor, query its historical optimal position gBest Li ,i takes the value of 1, 2…m, where m is the number of neighbor microenvironment nodes;

[0036] Calculate gBest according to formula (3)A ;

[0037] The particles in the particle swarm in any microenvironment use equations (4) and (5) to update their speed and position. After a finite number of iterations, the fitness value of the particle position in the particle swarm in the microenvironment can converge to fun(gBest);

[0038] in,

[0039]

[0040] If the fitness value of the new position is smaller than the value of the particle's historical best fitness, set the particle's historical best position as the particle's current position;

[0041] Update the optimal position gBest of all local particles L ;

[0042] Among them, fun(gBest L )≥fun(gBest A )≥fun(gBest);

[0043] Analyze gBest into the electrical appliance working state occurrence coefficient set C(t0) at time t0;

[0044] Finally, we get Result = I∧C(t0).

[0045] Preferably, the construction of the electrical appliance working state occurrence coefficient set C(t0) is realized by using the function given by formula (6) to judge whether each component of gBest corresponds to the electrical appliance working state in I; in formula (6), x i is a component of x, and δ is the threshold.

[0046] Preferably, the fitness of the particle is calculated by using the function fun(x) given by formula (7), where the independent variable x is the position of the particle; for the function fun(x) specified by formula (7), the better the particle, the smaller the value of fun(x); for fun(x): let x j ,u is electrical appliance X j The uth working state is e j,u The corresponding component in the independent variable x, then H(x j,u ,δ) is the working state e j,u assertion of whether it occurred; is the independent variable x about appliance X j The number of occurrences of the asserted working state; is the independent variable x about appliance X j The assertion of whether the number of working states that are asserted is greater than 1; since at a time, an appliance can have at most 1 state. Right now When, x about electrical appliances X j The assertion error of the working state; let x have more assertion errors about the working state of the appliance, the larger the value of fun(x); if M>0 and M is large enough, is regarded as a penalty term for the incorrect assertion of x about the working state of the appliance, and M is the penalty coefficient; the fitness function fun(x) specified in formula (6) is based on the objective function F(C) specified in formula (2), and the fitness of the candidate solution of the optimization problem is calculated by adding a penalty term for not satisfying the constraints of the optimization problem described in formula (2).

[0047] Preferably, the identified electrical appliance working states are extracted based on the constructed electrical appliance working state occurrence coefficient set C(t0): in C(t0), the electrical appliance working states in I corresponding to the coefficients with a value of 1 are all the electrical appliance working states identified from the user's power consumption situation s(t0).

[0048] The present invention has the following advantages:

[0049] The present invention uses an electrical appliance working state identification method based on combined optimization to identify the working state of electrical appliances. The average values ​​of the accuracy, precision, recall rate and F1 score are all above 93%, and the average value of the false alarm rate is less than 6%. The performance is good and the working state of the user-side electrical appliances can be identified more accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 Physical diagram of the experimental electrical appliances and data acquisition device provided by the present invention;

[0051] Figure 2 A graph showing the effect of the change in the number of particles PN provided by the present invention on the accuracy rate;

[0052] Figure 3 A graph showing the influence of the change in the maximum number of iterations maxGen provided by the present invention on the accuracy rate;

[0053] Figure 4 A graph showing the effect of changes in the learning factor c1 provided by the present invention on the accuracy rate;

[0054] Figure 5 A graph showing the effect of changes in the learning factor c2 provided by the present invention on the accuracy rate;

[0055] Figure 6 This is a diagram showing the influence of the change of the inertia factor ω on the accuracy provided by the present invention. DETAILED DESCRIPTION

[0056] The following is a description of the implementation of the present invention by specific embodiments. People familiar with the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are 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.

[0057] In order to test the effectiveness of the particle swarm optimization (PSO) algorithm, this embodiment collects power usage data of two electric kettles, one table lamp, one microwave oven, one electric fan and other common household appliances at a high frequency. This embodiment was carried out in Room 407 of the Anhui Provincial Key Laboratory of Intelligent Buildings and Building Energy Conservation, and the data collection period was from 12:30:00 on July 11, 2023 to 12:29:59 on July 18, 2023, a total of 168 hours. In the experiment: Under the premise of ensuring the safety of electricity use, each electrical appliance was used separately and in combination; the user's power usage status data was collected once per second. The actual electrical appliances and data collection devices in this embodiment are Figure 1 Given in. Figure 1 The data acquisition device is independently developed by the research team: the device includes 1 intelligent circuit breaker, 1 Raspberry Pi 4, and 1 5-port socket module. The intelligent circuit breaker and Raspberry Pi 4 are connected by wire using a TTL to RS485 module and communicate using the modbus protocol; the data acquisition program is deployed in the Raspberry Pi 4, and the collected data is also stored in the Raspberry Pi. The intelligent circuit breaker has been calibrated before leaving the factory. This embodiment uses the user's power consumption status including four parameter tables such as current, active power, reactive power, and apparent power to identify the working status of the appliance. Table 1 gives the rated power, rated voltage, and possible working status of each appliance.

[0058] Table 1: Electrical parameters and working conditions of the embodiment

[0059]

[0060] In the embodiment, the user's power consumption status data should be collected at 604,800 time points in all 168 hours. The statistical results show that the data collection program collected 603,934 data at 603,665 time points, and the power consumption status data was collected once repeatedly at 269 time points; no data was collected at 1,135 time points. For the time points where data was collected repeatedly, the data collected for the first time at the time point is the valid data at the time point; for the time points where data was not collected, the valid data at the previous adjacent time point is used as the estimate of the valid data at the time point to fill the missing data.

[0061] In order to test the performance of the particle swarm optimization (PSO) algorithm, the experiment first collected 1 minute of user power usage data for each working state of each appliance in Table 1, once per second, under the premise that only one appliance was connected to the circuit. Based on the 1-minute user power usage data set for each working state of each appliance, the median value was used as the basic working state of the working state, and then the basic working state set of all appliance working states required for identifying the working state of the appliance using the particle swarm optimization (PSO) algorithm was constructed.

[0062] The experiment considered the impact of five parameters that affect the performance of the PSO algorithm, including the number of particles PN, the maximum number of iterations maxGen, the learning factor c1, the learning factor c2, and the inertia coefficient ω. In the experiment, the number of particles PN changed from 50 to 3000, with a step of 50; the number of iterations maxGen changed from 100 to 1000, with a step of 100; the learning factor c1 changed from 0.1 to 2, with a step of 0.1; the learning factor c2 changed from 0.1 to 2, with a step of 0.1; the inertia coefficient ω changed from 0.1 to 2, with a step of 0.1. During the experiment, four of the above five parameters were fixed each time, and the impact of the change in the value of the other parameter on the accuracy of the PSO algorithm in identifying the working state of the electrical appliance was observed.

[0063] First, maxGen is set to 100, c1 is set to 2, c2 is set to 2, ω is set to 0.8, PN is changed from 50 to 3000, and the step size is 50. The corresponding quasi-removal rate change curve is Figure 2 Given in. Figure 2 In the figure, the circled position is the highest accuracy value of 0.94516, at which time the PN value is 2600.

[0064] Then set PN to 2600, c1 to 2, c2 to 2, ω to 0.8, maxGen to change from 100 to 1000, with a step size of 100, and the corresponding quasi-removal rate change curve is Figure 3 Given in. Figure 3 In the figure, the circled position is the highest accuracy value of 0.94514, and the value of maxGen is 400.

[0065] Then set PN to 2600, c2 to 2, ω to 0.8, maxGen to 400, c1 to change from 0.1 to 2, with a step size of 0.1. The corresponding quasi-removal rate change curve is Figure 4 Given in. Figure 4 The circled position is the highest accuracy value of 0.94542, and the value of c1 is 0.2.

[0066] Furthermore, we set PN to 2600, ω to 0.8, maxGen to 400, c1 to 0.2, step size to 0.1, and c2 to change from 0.1 to 2, step size to 0.1. The corresponding quasi-removal rate change curve is Figure 5 Given in. Figure 5 In the figure, the circled position is the highest accuracy value of 0.94828, and the value of c2 is 0.5.

[0067] Finally, PN is set to 2600, maxGen is set to 400, c1 is set to 0.2, c2 is set to 0.5, ω changes from 0.1 to 2, and the step size is 0.1. The corresponding quasi-removal rate change curve is Figure 6 Given in. Figure 6 In the figure, the circled position is the highest accuracy value of 0.94828, and the value of ω is 0.9.

[0068] In order to verify the superiority of using the particle swarm optimization PSO algorithm to identify the working state of electrical appliances, the experiment used the simulated annealing algorithm (SA) and the genetic algorithm (GA) to identify the working state of electrical appliances contained in the experimental data, and compared the identification performance with the particle swarm optimization PSO algorithm. In the experiment, for the particle swarm optimization PSO algorithm, the parameters PN are 2600, maxGen is 400, c1 is 0.2, c2 is 0.5, and ω is 0.9. Table 2 shows the performance of using the particle swarm optimization PSO algorithm, SA algorithm, and GA algorithm to identify the working state of electrical appliances.

[0069] Table 2: Performance of different algorithms in identifying the working status of electrical appliances

[0070]

[0071] As can be seen from Table 2, compared with the GA algorithm and SA algorithm, the particle swarm optimization PSO algorithm using the PSO algorithm to identify the working status of electrical appliances has better performance, with an accuracy of 94.83%, a precision of 93.28%, a recall rate of 95.25%, and an F1-score of 94.25%.

[0072] Table 3: Performance of Particle Swarm Optimization (PSO) algorithm

[0073] Minimum Q1 Median Q3 Maximum average value Standard Deviation Accuracy 0.9436 0.9467 0.9467 0.9483 0.9483 0.9471 0.0013 Accuracy 0.9276 0.9310 0.9310 0.9328 0.9328 0.9315 0.0014 Recall 0.9472 0.9507 0.9507 0.9525 0.9525 0.9512 0.0014 False reporting rate 0.0551 0.0551 0.0565 0.0565 0.0593 0.0561 0.0011 F1score 0.9373 0.9408 0.9408 0.9425 0.9425 0.9412 0.0014

[0074] In order to verify the stability of the performance of using particle swarm optimization PSO algorithm to identify the working state of electrical appliances, the experiment uses the particle swarm optimization PSO algorithm to repeatedly identify all the working states of electrical appliances contained in the experimental data 100 times, and calculates the accuracy, precision, recall, false alarm rate and F1 score of each identification. Table 3 shows the minimum, maximum, median, average, Q1 value, Q3 value and standard deviation of the relevant performance of the particle swarm optimization PSO algorithm in 100 repeated experiments. It can be seen from the data in Table 3 that the various indicators of the performance of the particle swarm optimization PSO algorithm maintain a good value, and the standard deviation is less than 0.002, showing good stability.

[0075] Although the present invention has been described in detail above by general description and specific embodiments, it is obvious to those skilled in the art that some modifications or improvements can be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. The method for identifying the working state of an electrical appliance based on a particle swarm optimization algorithm is characterized by: Let a1, a2, ..., a m is the parameter related to the power consumption process perceived in the circuit, called s = {a1, a2, ..., a m } is the user's electricity consumption situation; the electricity consumption situation s at time t is expressed as s(t) = (a1(t), a2(t), ..., a m (t)); Let X = {X1, X2…X n } is the set of all n electrical appliances in the circuit. j∈[1,n] and j is an integer, let is the set of all possible working states. is the summary set of all working states of all n electrical appliances, and I is called the working state set of all electrical appliances; If at time t there is only device X in the circuit j Working and its working status is e j,u , if the power consumption situation at this moment is s(t)=(a1(t),a2(t),...,a m (t)), record k∈[1,m] and k is an integer, s(t)=s(t,e j,u ),say Yes j,u The observed value of the basic working status at time t; If the appliance X j Working status j,u Occurrence, regulation c j,u (t)=1, otherwise c j,u (t) = 0, called c j,u (t) is the working state at time t j,u The occurrence coefficient of is the set of working state occurrence coefficients of I at time t; although electrical appliance X j The working status is There are many possibilities, but at a specific moment t, appliance X j The working status is only k j One of the working states, namely Considering Appliance X j Working status j,u The basic working situation is electrical appliances X j An attribute feature of Then b(e j,u ) During the observation period, only appliance X j Working and its working status is e j,u When all s(t,e j,u ) as its estimate; Let B = {b(e j,u )|e j,u ∈I}, B is the set of basic working states of all electrical appliances in the circuit. Ideally, the user power consumption state s(t) observed in the circuit at time t is the combination of all basic working states as shown in formula (1); In formula (1), c j,u ∈C; According to formula (1), the problem of identifying the working state of the electrical appliance based on the user's power consumption situation s(t) observed in the circuit at time t is essentially the problem of solving the working state generation coefficient set C(t) of I at time t, and formula (2) gives a model for solving the working state generation coefficient set C(t) of I at time t in the form of an optimization problem; the particle swarm optimization PSO algorithm is used to solve the optimization problem described in formula (2), and then the working state of the electrical appliance on the circuit is identified according to the optimal value of the working state generation coefficient set C(t); The process of using particle swarm optimization (PSO) algorithm to identify the working status of electrical appliances in the circuit is as follows: In the PSO algorithm, a particle is a feasible solution to the corresponding optimization problem; when implementing the PSO algorithm in a microenvironment network, each particle of the PSO algorithm must be assigned to a microenvironment, and each microenvironment has n particles, assuming that the number of particles n in each microenvironment is the same; For the PSO algorithm implemented in the micro-environment network, assume: 1) There are n particles in each microenvironment; 2) For any microenvironment, record gBest L is the historical optimal position of all particles in the microenvironment; 3) For a microenvironment, if it has m neighboring microenvironments, record gBest Li is the optimal position of all particles in its i-th neighbor microenvironment, and gBest is specified A As formula (3); gBest A =min(gBest L ,gBest L1 ,gBest L2 …gBest Lm )(3) 4) For the PSO algorithm based on microenvironment, for any particle in a microenvironment, if its position at time t is x(t) and its velocity is v(t), then its velocity v(t+1) and position x(t+1) at time t+1 are determined according to equations (4) and (5); v(t+1)=ω×v(t)+c1×rand()×(pBest-x(t))+c2×rand()×(gBest A -x(t)) (4) x(t+1)=x(t)+v(t+1) (5) Among them, formula (3) is the data exchange strategy of the PSO algorithm based on the microenvironment, while formula (4) and formula (5) are the particle position update strategies in the PSO algorithm based on the microenvironment; Step 1: Initialize the speed and position of each local particle; Step 2: Take the initial position of each particle as the initial value of its historical optimal position: Step 3: Calculate the optimal position gBest of all local particles L ; The specific implementation steps of Step1-Step3 are as follows: Input user power consumption situation s(t0), all electrical appliances working state set I, basic working state set B, local particle number n, particle position dimension varLen, learning factors c1, c2, inertia factor ω, maximum number of iterations maxGen, fitness function fun, neighbor microenvironment node list pNei ghbor; Set the particle dimension varLen to the number of elements in B; The number of iterations k starts from 1, and it is determined whether the number of iterations k is less than the maximum number of iterations maxGen parameter to control whether the step of realizing the generation coefficient set C(t0) of the electrical appliance working state continues to be iteratively executed; The steps to realize the electrical appliance working state occurrence coefficient set C(t0) are as follows: For each neighbor micro-environment node in pNeighbor, query its historical optimal position gBest Li ,i takes the value of 1, 2…m, where m is the number of neighbor microenvironment nodes; Calculate gBest according to formula (3) A ; The particles in the particle swarm in any microenvironment use equations (4) and (5) to update their speed and position. After a finite number of iterations, the fitness value of the particle position in the particle swarm in the microenvironment can converge to fun(gBest); in, By using the function given by formula (6) to determine whether each component of gBest corresponds to the electrical appliance working state in I, the construction of the electrical appliance working state occurrence coefficient set C(t0) is realized; in formula (6), x i is a component of x, and δ is the threshold; If the fitness value of the new position is smaller than the value of the particle's historical best fitness, set the particle's historical best position as the particle's current position; Update the optimal position gBest of all local particles L ; Among them, fun(gBest L )≥fun(gBest A )≥fun(gBest); Analyze gBest into the electrical appliance working state occurrence coefficient set C(t0) at time t0; Finally, we get Result = I∧C(t0).

2. The method for identifying the working state of an electrical appliance based on a particle swarm optimization algorithm according to claim 1, characterized in that: The fitness of the particle is calculated by using the function fun(x) given by equation (7), where the independent variable x is the position of the particle; for fun(x): let x j ,u is electrical appliance X j The uth working state is e j,u The corresponding component in the independent variable x, then H(x j,u ,δ) is the working state e j,u assertion of whether it occurred; is the independent variable x about appliance X j The number of occurrences of the asserted working state; is the independent variable x about appliance X j The assertion of whether the number of working states that are asserted is greater than 1; since at a time, an appliance can have at most 1 state. Right now When, x about electrical appliances X j The assertion error of the working state; let x have more assertion errors about the working state of the appliance, the greater the value of fun(x); if M>0, is regarded as a penalty term for the incorrect assertion of x about the working state of the appliance, and M is the penalty coefficient; the fitness function fun(x) specified in formula (6) is based on the objective function F(C) specified in formula (2), and the fitness of the candidate solution of the optimization problem is calculated by adding a penalty term for not satisfying the constraints of the optimization problem described in formula (2).

3. The method for identifying the working state of an electrical appliance based on a particle swarm optimization algorithm according to claim 1, characterized in that: According to the constructed electrical appliance working state occurrence coefficient set C(t0), the identified electrical appliance working state is extracted: in C(t0), the electrical appliance working state in I corresponding to the coefficient with a value of 1 is all the electrical appliance working states identified from the user's power consumption situation s(t0).

Images