Electric equipment non-intrusive load identification method and equipment based on mixed characteristics

Abstract

The invention discloses an electric equipment non-intrusive load identification method based on mixed characteristics, and discloses equipment adopting the electric equipment non-intrusive load identification method based on the mixed characteristics. According to the non-intrusive load identification method for the electric equipment based on the mixed features, active power, reactive power and total harmonic distortion in the operation process of the electric equipment are extracted to serve as the mixed features, load identification of the electric equipment is conducted, a BP neural network structure model is constructed, and the optimal solution of a particle swarm algorithm serves as the weight and the threshold value of the BP neural network. And continuously carrying out network training until the accuracy is optimal. And rapid construction and accurate identification of the load identification model are realized. The problem that a non-intrusive load identification technology is insufficient in accuracy and efficiency in the aspects of load feature extraction and load identification is solved, and the accuracy and efficiency of load identification are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of electricity monitoring and signal processing technology, and in particular to a non-intrusive load identification method and device for electrical equipment based on hybrid features. Background Technology

[0002] Load monitoring technology is a crucial foundation for building new power systems. This technology samples and analyzes total user load data to provide real-time monitoring of terminal equipment, feedback on load operations, and assessment of equipment power consumption patterns, thus standardizing the management and monitoring system for new energy sources. Non-intrusive load identification technology only requires the installation of measuring devices at the power inlet to obtain electrical signals such as voltage and current of the total load. By analyzing the collected load data, extracting the characteristic information of these electrical quantities, and using intelligent algorithms, load identification of electrical equipment is achieved.

[0003] Currently, research on non-intrusive load identification technology mainly focuses on two aspects: load feature extraction and load identification. Existing non-intrusive load identification technologies suffer from insufficient accuracy and efficiency in load feature extraction and load identification, resulting in poor practical effects. Therefore, improvements are needed in load feature extraction and load identification to enhance the accuracy and efficiency of load identification. Summary of the Invention

[0004] This invention aims to address at least one of the technical problems existing in the prior art. To this end, this invention proposes a non-intrusive load identification method based on the DAPSO-BP algorithm and hybrid features, enabling rapid construction and accurate identification of load identification models. This effectively improves the accuracy and efficiency of load identification.

[0005] The present invention also proposes a device having the above-mentioned non-intrusive load identification method for electrical equipment based on hybrid features.

[0006] The non-intrusive load identification method for electrical equipment based on hybrid features according to a first aspect of the present invention is characterized by comprising the following steps:

[0007] Collect the total load voltage and current signals of the electrical equipment, and extract the active power, reactive power and total harmonic distortion during the operation of the electrical equipment as mixed features;

[0008] A BP neural network model is constructed, wherein the input layer of the neural network model consists of three features: active power, reactive power, and total harmonic distortion, and the output layer consists of the operating status of the electrical equipment.

[0009] The weights and thresholds of the BP neural network are optimized using a dynamic adaptive particle swarm optimization algorithm, which updates parameters by dynamically adjusting inertia weights and learning factors.

[0010] The entropy error function is used as the network error objective function, and the BP neural network is trained using a variable learning rate method until the network convergence accuracy reaches a preset threshold, thus completing the construction of the load identification model.

[0011] The load identification model is used to identify the operating status of electrical equipment in real time.

[0012] The non-intrusive load identification method for electrical equipment based on hybrid features according to embodiments of the present invention has at least the following beneficial effects:

[0013] The non-intrusive load identification method provided in this application extracts active power, reactive power, and total harmonic distortion (THD) as hybrid features during the operation of electrical equipment for load identification. A BP neural network structure model is constructed, and the optimal solution of the particle swarm optimization algorithm is used as the weights and thresholds of the BP neural network. The network is continuously trained until optimal accuracy is achieved. This enables rapid construction and accurate identification of the load identification model. It solves the problems of insufficient accuracy and efficiency in load feature extraction and load identification in non-intrusive load identification technology, effectively improving the accuracy and efficiency of load identification.

[0014] According to some embodiments of the present invention, in the step of constructing a BP neural network model, wherein the input layer of the network model consists of the three features of active power, reactive power, and total harmonic distortion, and the output layer is the operating state of the electrical equipment, the output layer is constructed as follows:

[0015] For devices in a start-stop two-state state, when the electrical equipment is in the running state, the expected output of the corresponding node is 1, otherwise it is 0;

[0016] For multi-state devices, the expected output of the corresponding node is 1, 2, 3... otherwise it is 0.

[0017] According to some embodiments of the present invention, in the step of optimizing the weights and thresholds of the BP neural network using a dynamic adaptive particle swarm optimization algorithm, wherein the dynamic adaptive particle swarm optimization algorithm updates parameters by dynamically adjusting the inertia weights and learning factors, the adjustment formulas for the inertia weights and learning factors of the dynamic adaptive particle swarm optimization algorithm are as follows:

[0018]

[0019] Wherein, ω max and ω min These represent the maximum and minimum values ​​of the inertia weight, respectively; k2 is the current iteration number; and N2 is the maximum iteration number. The inertia weight for the (k2+1)th iteration. and The learning factor for the (k2+1)th iteration. The particle fitness value in the k2th iteration. This represents the average fitness value of the particles in the k2th iteration. It represents the minimum value of the particle fitness in the k2th iteration.

[0020] According to some embodiments of the present invention, in the step of using the entropy error function as the network error objective function and training the BP neural network in combination with a variable learning rate method until the network convergence accuracy reaches a preset threshold to complete the construction of the load identification model, the expression of the entropy error function is:

[0021]

[0022] Where, the subscript i represents the error of the i-th learning sample; the subscript k represents the k-th output neuron; Y represents the target output; y represents the actual output of the network output layer; and m represents the number of neurons in the network output layer.

[0023] According to some embodiments of the present invention, in the step of using the entropy error function as the network error objective function and training the BP neural network in combination with a variable learning rate method until the network convergence accuracy reaches a preset threshold to complete the construction of the load identification model, the adjustment formula of the variable learning rate is as follows:

[0024]

[0025] Where α is the learning rate, k1 is the number of learning iterations, E is the network correction error, and λ is the learning rate error adjustment coefficient, which typically ranges from 0.01 to 0.03.

[0026] According to some embodiments of the present invention, in the step of acquiring the total load voltage and current signal of the electrical equipment and extracting the active power, reactive power and total harmonic distortion (THD) as mixed features during the operation of the electrical equipment, the THD k... THD The calculation formula is:

[0027]

[0028] Where G1 is the effective value of the fundamental component, H is the preset harmonic order, and G n The effective value of all harmonic components not greater than a specific order H.

[0029] According to some embodiments of the present invention, in the step of optimizing the weights and thresholds of the BP neural network using a dynamic adaptive particle swarm optimization algorithm, wherein the dynamic adaptive particle swarm optimization algorithm updates parameters by dynamically adjusting the inertia weights and learning factors, the particle velocity and position update formulas of the particle swarm optimization algorithm are as follows:

[0030]

[0031] In the formula, Let c1 and c2 represent the velocity and position of the i-th particle in the k2-th iteration, respectively. c1 and c2 are learning factors, r1 and r2 are random numbers between [0,1], ω is the inertia weight, and p best,i For the optimal position of the i-th particle, g best This is the globally optimal position.

[0032] According to some embodiments of the present invention, in the step of constructing a BP neural network model, wherein the input layer of the network model consists of the three features of active power, reactive power, and total harmonic distortion, and the output layer represents the operating state of the electrical equipment, the number of hidden layer neurons in the BP neural network is determined by the following empirical formula:

[0033]

[0034] Where l is the number of hidden layer neurons, n is the number of input layer neurons, m is the number of output layer neurons, and a is a constant between 1 and 10.

[0035] According to some embodiments of the present invention, in the step of using the load identification model to identify the operating status of electrical equipment in real time, the load identification model has an identification rate of no less than 90% for electrical equipment.

[0036] According to a second aspect of the present invention, a non-intrusive load identification device for electrical equipment based on hybrid features is characterized in that it comprises:

[0037] The data acquisition module can acquire the total load voltage and current signals of the electrical equipment and extract the active power, reactive power and total harmonic distortion during the operation of the electrical equipment as mixed features.

[0038] The model building module can build a BP neural network model. The input layer of the neural network model consists of three features: active power, reactive power, and total harmonic distortion. The output layer is the operating status of the electrical equipment.

[0039] The parameter update module can optimize the weights and thresholds of the BP neural network using a dynamic adaptive particle swarm optimization algorithm. The dynamic adaptive particle swarm optimization algorithm updates the parameters by dynamically adjusting the inertia weights and learning factors.

[0040] The model training module can use the entropy error function as the network error objective function and combine it with the variable learning rate method to train the BP neural network until the network convergence accuracy reaches a preset threshold, thus completing the construction of the load identification model.

[0041] The identification module can use the load identification model to identify the operating status of electrical equipment in real time.

[0042] Furthermore, in the model building module, the output layer is constructed in the following way:

[0043] For devices in a start-stop two-state state, when the electrical equipment is in the running state, the expected output of the corresponding node is 1, otherwise it is 0;

[0044] For multi-state devices, the expected output of the corresponding node is 1, 2, 3... otherwise it is 0.

[0045] Furthermore, in the parameter update module, the adjustment formulas for the inertia weights and learning factors of the dynamic adaptive particle swarm algorithm are as follows:

[0046]

[0047] Where, ω max and ω min These represent the maximum and minimum values ​​of the inertia weight, respectively; k2 is the current iteration number; and N2 is the maximum iteration number. The inertia weight for the (k2+1)th iteration. and The learning factor for the (k2+1)th iteration. The particle fitness value in the k2th iteration. This represents the average fitness value of the particles in the k2th iteration. It represents the minimum value of the particle fitness in the k2th iteration.

[0048] Furthermore, in the step of using the entropy error function as the network error objective function and training the BP neural network using a variable learning rate method until the network convergence accuracy reaches a preset threshold to complete the construction of the load identification model, the expression for the entropy error function is:

[0049]

[0050] Where, the subscript i represents the error of the i-th learning sample; the subscript k represents the k-th output neuron; Y represents the target output; y represents the actual output of the network output layer; and m represents the number of neurons in the network output layer.

[0051] Furthermore, in the model training module, the adjustment formula for the variable learning rate is:

[0052]

[0053] Where α is the learning rate, k1 is the number of learning iterations, E is the network correction error, and λ is the learning rate error adjustment coefficient, which typically ranges from 0.01 to 0.03.

[0054] Furthermore, in the data acquisition module, the total harmonic distortion k THD The calculation formula is:

[0055]

[0056] Where G1 is the effective value of the fundamental component, H is the preset harmonic order, and G n The effective value of all harmonic components not greater than a specific order H.

[0057] Furthermore, in the parameter update module, the particle velocity and position update formulas of the particle swarm algorithm are as follows:

[0058]

[0059] In the formula, Let c1 and c2 represent the velocity and position of the i-th particle in the k2-th iteration, respectively. c1 and c2 are learning factors, r1 and r2 are random numbers between [0,1], ω is the inertia weight, and p best,i For the optimal position of the i-th particle, g best This is the globally optimal position.

[0060] Furthermore, in the model building module, the number of hidden layer neurons in the BP neural network is determined by the following empirical formula:

[0061]

[0062] Where l is the number of hidden layer neurons, n is the number of input layer neurons, m is the number of output layer neurons, and a is a constant between 1 and 10.

[0063] Furthermore, in the identification module, the load identification model has an identification rate of no less than 90% for electrical equipment.

[0064] According to a third aspect of the present invention, a computer-readable storage medium stores computer-executable instructions for performing the above-described non-intrusive load identification method for electrical appliances based on hybrid features.

[0065] An electronic device according to a fourth aspect of the present invention is characterized in that the device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described non-intrusive load identification method for electrical appliances based on hybrid features.

[0066] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0067] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0068] Figure 1 This is a schematic diagram illustrating the steps of the non-intrusive load identification method for electrical equipment based on hybrid features according to an embodiment of the present invention.

[0069] Figure 2 for Figure 1 A schematic diagram illustrating the steps of the DAPSO-BP load identification method, a non-intrusive load identification method for electrical equipment based on hybrid features;

[0070] Figure 3 A schematic diagram of partial sampling data in the non-intrusive load identification method for electrical equipment based on hybrid features is provided for embodiments of the present invention;

[0071] Figure 4 This is an analysis chart of the recognition rate of electrical equipment provided in an embodiment of the present invention;

[0072] Figure 5 This is a structural block diagram of a non-intrusive load identification device for electrical equipment based on hybrid features, provided in an embodiment of the present invention. Detailed Implementation

[0073] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0074] In the description of this invention, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc., are based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0075] In the description of this invention, "several" means one or more, "multiple" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0076] In the description of this invention, unless otherwise explicitly defined, terms such as "set up," "install," and "connect" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this invention in conjunction with the specific content of the technical solution.

[0077] Example 1

[0078] To address the shortcomings in accuracy and efficiency of existing non-intrusive load identification technologies in load feature extraction and load identification, this application provides a non-intrusive load identification method for electrical equipment based on hybrid features, such as... Figure 1 As shown, it includes at least the following steps:

[0079] Step S100: Collect the total load voltage and current signals of the electrical equipment, and extract the active power, reactive power and total harmonic distortion during the operation of the electrical equipment as mixed features.

[0080] Select at least five different types of electrical equipment and record the waveforms of at least one complete operation. Extract the active power, reactive power, and total harmonic distortion (THD) during the operation of the electrical equipment as hybrid features for load identification. The formulas for calculating the active power, reactive power, and THD of the electrical equipment are as follows:

[0081]

[0082] In the formula, U and I represent voltage and current, respectively, and k is the harmonic order. G represents the phase difference between voltage and current, and G1 is the effective value of the fundamental component; G n The effective value of all harmonic components not greater than a specific order H.

[0083] Step S200: Construct a BP neural network model. The input layer of the neural network model consists of three features: active power, reactive power, and total harmonic distortion. The output layer is the operating status of the electrical equipment.

[0084] Using the mixed characteristics of electrical equipment—active power, reactive power, and total harmonic distortion (THD)—as input and the operating state of the equipment as output, a BP neural network structure model is constructed. The input layer of the BP neural network is constructed based on the number of features selected in the mixed characteristics, and the output layer is constructed based on the total number of electrical equipment. The input layer of the BP neural network consists of three features: active power, reactive power, and THD, containing three neurons. The output layer outputs the operating state of the equipment. For two-state devices (start-stop), the expected output of the corresponding node is 1 when the equipment is in the running state, and 0 otherwise. For multi-state devices, the expected output of the corresponding node is 1, 2, 3…, and 0 otherwise.

[0085] Step S300: The weights and thresholds of the BP neural network are optimized using a dynamic adaptive particle swarm optimization algorithm. The dynamic adaptive particle swarm optimization algorithm updates the parameters by dynamically adjusting the inertia weights and learning factors.

[0086] The particle swarm optimization (DAPSO) algorithm with dynamic adjustment of inertia weights and learning factors is used to optimize the weights and thresholds of the BP neural network. The entropy error function is used as the network error objective function and the variable learning rate method to enhance the convergence speed of the BP neural network.

[0087] The feedback correction formulas for the weights and thresholds of the output and hidden layers of a BP neural network are as follows:

[0088]

[0089] In the formula, α is the learning rate; k1 is the number of learning iterations; and δ k This represents the network error signal during the backpropagation process.

[0090] The PSO algorithm updates the velocity and position of each particle by calculating individual and global extrema, searching for the global optimum during the iteration process. The formulas for updating particle velocity and position are:

[0091]

[0092] In the formula, Let c1 and c2 represent the velocity and position of the i-th particle in the k2-th iteration, respectively. c1 and c2 are learning factors, r1 and r2 are random numbers between [0,1], ω is the inertia weight, and p best,i For the optimal position of the i-th particle, g best This is the globally optimal position.

[0093] Step S400: Use the entropy error function as the network error objective function and combine it with the variable learning rate method to train the BP neural network until the network convergence accuracy reaches the preset threshold, thus completing the construction of the load identification model.

[0094] Specifically, this embodiment uses a combination of dynamic adaptive particle swarm optimization algorithm and BP neural network for load identification, such as... Figure 2 As shown, the specific process of the DAPSO-BP load identification method is as follows:

[0095] Step S401: Construct the neural network structure.

[0096] The input layer of the BP neural network is constructed based on the number of features in the selected mixed features, the output layer is constructed based on the total number of electrical devices, and the number of nodes in the hidden layer is determined according to an empirical formula:

[0097]

[0098] In the formula, l is the number of hidden layer neurons, m is the number of output layer neurons, n is the number of input layer neurons, and a is an arbitrary constant between 1 and 10.

[0099] Step S402: Initialization of the BP neural network

[0100] Initialize the network weights and thresholds, and set the initial iteration value for network training to k1=1. Input the network training samples and normalize the power and harmonic feature data.

[0101] Step S403: Solving the output of the BP neural network

[0102] Let the activation function be f(x) = 1 / (1+e -x The output values ​​of the hidden layer and the output layer are calculated using the following formula:

[0103]

[0104] Step S404: Calculate network error

[0105] To improve the convergence speed of BP neural networks and avoid the false saturation phenomenon that exists in traditional error functions during training, the entropy error function is taken as the objective function of the network error, such as:

[0106]

[0107] In the formula, the subscript i represents the error of the i-th learning sample; the subscript k represents the k-th output neuron; Y represents the target output; y represents the actual output of the network output layer; and m represents the number of neurons in the network output layer.

[0108] Step S405: Conditional Judgment using a BP Neural Network

[0109] The convergence condition for a BP neural network is:

[0110]

[0111] In the formula, N is the total number of training samples, ε1 is the preset convergence accuracy, and N1 is the maximum allowed number of learning iterations. If Case 1 is satisfied, proceed to step 6. If Case 2 or Case 3 is satisfied, the neural network training ends, a load identification model is provided, and load identification is performed. Otherwise, proceed to step 12 to continue network training.

[0112] Step S406: Particle parameter initialization

[0113] Let the initial number of iterations be k2 = 1, and the maximum number of iterations be N2. Let the initial value of the inertial weight be ω0, and the particle learning factors be c1 and c2. Initialize the initial position W0 and the initial velocity V0 of the particle.

[0114] Step S407, Fitness Calculation

[0115] The fitness function is constructed using the error of the BP neural network, i.e.

[0116]

[0117] Step S408: Optimal Particle Acquisition

[0118] The local and global optima of the particles are calculated.

[0119] Step S409, PSO convergence judgment

[0120] The convergence condition of the PSO algorithm is:

[0121]

[0122] In the formula, εk2 is the globally optimal fitness value in the k2th iteration, ε2 is the preset convergence accuracy, and N2 is the maximum allowed number of iterations. When the convergence condition is met, the PSO optimal solution is obtained, and then proceed to Step 12 to continue network training. If the condition is not met, proceed to Step 10.

[0123] Step S410: System Parameter Update

[0124] The PSO system parameters are dynamically adjusted using the following formula:

[0125]

[0126] In the formula, where ω max and ω min These represent the maximum and minimum values ​​of the inertia weight, respectively; k2 is the current iteration number; and N2 is the maximum iteration number. The inertia weight for the (k2+1)th iteration. and The learning factor for the (k2+1)th iteration. The particle fitness value in the k2th iteration. This represents the average fitness value of the particles in the k2th iteration. It represents the minimum value of the particle fitness in the k2th iteration.

[0127] Step S411, Particle Update

[0128] Update the particle's velocity and position, and set k2 = k2 + 1, then return to step S407.

[0129] Step S412, Network Weights and Threshold Update

[0130] When k1 = 1, the optimal solution of PSO is used as the weights and threshold of the BP neural network. Otherwise, the network weights and threshold are updated, and the network is corrected using a variable learning rate, as shown in the following formula. Then, k1 = k1 + 1, and the process proceeds to step S403 to continue network training.

[0131]

[0132] In the formula, E is the network correction error; λ is the learning rate error adjustment coefficient, which generally ranges from 0.01 to 0.03.

[0133] Finally, the optimal solution from the particle swarm optimization algorithm is used as the weights and threshold of the backpropagation (BP) neural network. The network is continuously trained until the accuracy reaches its optimum, thus completing the construction of the load identification model.

[0134] Step S500: Use the load identification model to identify the operating status of electrical equipment in real time.

[0135] Using the measured dataset, load identification was performed on five electrical appliances: laptops, incandescent light bulbs, water dispensers, hair dryers, and televisions. 70% of the data was used as the training set to build a backpropagation (BP) neural network load identification model, and 30% was used as the test set to verify the model's prediction accuracy. (Partial sample data) Figure 3 As shown, the analysis yielded the following results (see details). Figure 4 The DAPSO-BP method exhibits good recognition accuracy and convergence efficiency. It achieves a load recognition rate of over 90% for five different types of electrical equipment. Furthermore, it employs an entropy function as the network error objective function and a variable learning rate to improve the convergence speed of the BP neural network, thereby enhancing the efficiency of load identification model construction.

[0136] Example 2

[0137] Another embodiment of this application provides a non-intrusive load identification device for electrical equipment based on hybrid features, such as... Figure 5 As shown, the device 50 includes:

[0138] The data acquisition module 501 is capable of acquiring the total load voltage and current signals of the electrical equipment, and extracting the active power, reactive power and total harmonic distortion during the operation of the electrical equipment as mixed features.

[0139] The model building module 502 is capable of building a BP neural network model. The input layer of the neural network model consists of three features: active power, reactive power, and total harmonic distortion. The output layer is the operating status of the electrical equipment.

[0140] The parameter update module 503 can optimize the weights and thresholds of the BP neural network using a dynamic adaptive particle swarm optimization algorithm. The dynamic adaptive particle swarm optimization algorithm updates the parameters by dynamically adjusting the inertia weights and learning factors.

[0141] The model training module 504 can use the entropy error function as the network error objective function and combine it with the variable learning rate method to train the BP neural network until the network convergence accuracy reaches a preset threshold, thus completing the construction of the load identification model.

[0142] The identification module 505 can use the load identification model to identify the operating status of electrical equipment in real time.

[0143] Furthermore, in the model building module, the output layer is constructed in the following way:

[0144] For devices in a start-stop two-state state, when the electrical equipment is in the running state, the expected output of the corresponding node is 1, otherwise it is 0;

[0145] For multi-state devices, the expected output of the corresponding node is 1, 2, 3... otherwise it is 0.

[0146] Furthermore, in the parameter update module 503, the adjustment formulas for the inertia weights and learning factors of the dynamic adaptive particle swarm algorithm are as follows:

[0147]

[0148] Where, ω max and ω min These represent the maximum and minimum values ​​of the inertia weight, respectively; k2 is the current iteration number; and N2 is the maximum iteration number. The inertia weight for the (k2+1)th iteration. and The learning factor for the (k2+1)th iteration. The particle fitness value in the k2th iteration. This represents the average fitness value of the particles in the k2th iteration. It represents the minimum value of the particle fitness in the k2th iteration.

[0149] Furthermore, in the model training module 504, the expression for the entropy error function is:

[0150]

[0151] Where, the subscript i represents the error of the i-th learning sample; the subscript k represents the k-th output neuron; Y represents the target output; y represents the actual output of the network output layer; and m represents the number of neurons in the network output layer.

[0152] Furthermore, in the model training module 504, the adjustment formula for the variable learning rate is:

[0153]

[0154] Where α is the learning rate, k1 is the number of learning iterations, E is the network correction error, and λ is the learning rate error adjustment coefficient, which typically ranges from 0.01 to 0.03.

[0155] Furthermore, in the data acquisition module 501, the total harmonic distortion k THD The calculation formula is:

[0156]

[0157] Where G1 is the effective value of the fundamental component, H is the preset harmonic order, and G n The effective value of all harmonic components not greater than a specific order H.

[0158] Furthermore, in the parameter update module 503, the particle velocity and position update formulas of the particle swarm algorithm are as follows:

[0159]

[0160] In the formula, Let c1 and c2 represent the velocity and position of the i-th particle in the k2-th iteration, respectively. c1 and c2 are learning factors, r1 and r2 are random numbers between [0,1], ω is the inertia weight, and p best,i For the optimal position of the i-th particle, g best This is the globally optimal position.

[0161] Furthermore, in the model construction module 502, the number of hidden layer neurons in the BP neural network is determined by the following empirical formula:

[0162]

[0163] Where l is the number of hidden layer neurons, n is the number of input layer neurons, m is the number of output layer neurons, and a is a constant between 1 and 10.

[0164] Furthermore, in the identification module 505, the load identification model has an identification rate of no less than 90% for electrical equipment.

[0165] Another aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described non-intrusive load identification method for electrical appliances based on hybrid features.

[0166] Specifically, the processor connects to the memory via a bus, which may include a path for transmitting information. The bus can be a PCI bus or an EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc.

[0167] The memory may be ROM or other types of static storage devices that can store static information and instructions, RAM or other types of dynamic storage devices that can store information and instructions, or EEPROM, CD-ROM or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited thereto.

[0168] Optionally, the memory stores the code of the computer program that executes the scheme of this application, and the execution is controlled by the processor. The processor executes the application code stored in the memory to implement... Figure 1 The function of the non-intrusive load identification method for electrical equipment based on hybrid features is shown.

[0169] Another aspect of this application provides a computer-readable storage medium storing computer-executable instructions for performing the above-described... Figure 1 The method shown is a non-intrusive load identification method for electrical equipment based on hybrid features.

[0170] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0171] It will be understood by those skilled in the art that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, as is known to those skilled in the art, communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0172] The above is a detailed description of the preferred embodiments of this application. However, this application is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A non-intrusive load identification method for electrical equipment based on hybrid features, characterized in that, The following steps are involved: Collect the total load voltage and current signals of the electrical equipment, and extract the active power, reactive power and total harmonic distortion during the operation of the electrical equipment as mixed features; A BP neural network model is constructed, wherein the input layer of the neural network model consists of three features: active power, reactive power, and total harmonic distortion, and the output layer consists of the operating status of the electrical equipment. The weights and thresholds of the BP neural network are optimized using a dynamic adaptive particle swarm optimization algorithm, which updates parameters by dynamically adjusting inertia weights and learning factors. The entropy error function is used as the network error objective function, and the BP neural network is trained using a variable learning rate method until the network convergence accuracy reaches a preset threshold, thus completing the construction of the load identification model. The load identification model is used to identify the operating status of electrical equipment in real time.

2. The method according to claim 1, characterized in that, In the step of constructing a BP neural network model, where the input layer of the network model consists of three features—active power, reactive power, and total harmonic distortion—and the output layer represents the operating status of the electrical equipment, the output layer is constructed as follows: For devices in a start-stop two-state state, when the electrical equipment is in the running state, the expected output of the corresponding node is 1, otherwise it is 0; For multi-state devices, the expected output of the corresponding node is 1, 2, 3... otherwise it is 0.

3. The method according to claim 1, characterized in that, In the step of optimizing the weights and thresholds of the BP neural network using a dynamic adaptive particle swarm optimization algorithm, where the dynamic adaptive particle swarm optimization algorithm updates parameters by dynamically adjusting the inertia weights and learning factors, the adjustment formulas for the inertia weights and learning factors are as follows: Where, ω max and ω min These represent the maximum and minimum values ​​of the inertia weight, respectively; k2 is the current iteration number; and N2 is the maximum iteration number. The inertia weight for the (k2+1)th iteration. and The learning factor for the (k2+1)th iteration. The particle fitness value in the k2th iteration. This represents the average fitness value of the particles in the k2th iteration. It represents the minimum value of the particle fitness in the k2th iteration.

4. The method according to claim 1, characterized in that, In the step of using the entropy error function as the network error objective function and training the BP neural network using a variable learning rate method until the network convergence accuracy reaches a preset threshold to complete the construction of the load identification model, the expression for the entropy error function is: Where, the subscript i represents the error of the i-th learning sample; the subscript k represents the k-th output neuron; Y represents the target output; y represents the actual output of the network output layer; and m represents the number of neurons in the network output layer.

5. The method according to claim 1, characterized in that, In the step of using the entropy error function as the network error objective function and training the BP neural network using a variable learning rate method until the network convergence accuracy reaches a preset threshold to complete the construction of the load identification model, the adjustment formula for the variable learning rate is as follows: Where α is the learning rate, k1 is the number of learning iterations, E is the network correction error, and λ is the learning rate error adjustment coefficient, which typically ranges from 0.01 to 0.

03.

6. The method according to claim 1, characterized in that, In the step of collecting the total load voltage and current signals of the electrical equipment and extracting the active power, reactive power, and total harmonic distortion (THD) as hybrid features during the operation of the electrical equipment, the THD k... THD The calculation formula is: Where G1 is the effective value of the fundamental component, H is the preset harmonic order, and G n The effective value of all harmonic components not greater than a specific order H.

7. The method according to claim 1, characterized in that, In the step of optimizing the weights and thresholds of the BP neural network using a dynamic adaptive particle swarm optimization algorithm, where the algorithm updates parameters by dynamically adjusting inertia weights and learning factors, the particle velocity and position update formulas are as follows: In the formula, Let c1 and c2 represent the velocity and position of the i-th particle in the k2-th iteration, respectively. c1 and c2 are learning factors, r1 and r2 are random numbers between [0,1], ω is the inertia weight, and p best,i For the optimal position of the i-th particle, g best This is the globally optimal position.

8. The method according to claim 1, characterized in that, In the step of constructing a BP neural network model, where the input layer of the network model consists of three features—active power, reactive power, and total harmonic distortion—and the output layer represents the operating status of the electrical equipment, the number of hidden layer neurons in the BP neural network is determined using the following empirical formula: Where l is the number of hidden layer neurons, n is the number of input layer neurons, m is the number of output layer neurons, and a is a constant between 1 and 10.

9. The method according to claim 1, characterized in that, In the step of using the load identification model to identify the operating status of electrical equipment in real time, the load identification model has an identification rate of no less than 90% for electrical equipment.

10. A non-intrusive load identification device for electrical equipment based on hybrid features, characterized in that, include: The data acquisition module can acquire the total load voltage and current signals of the electrical equipment and extract the active power, reactive power and total harmonic distortion during the operation of the electrical equipment as mixed features. The model building module can build a BP neural network model. The input layer of the neural network model consists of three features: active power, reactive power, and total harmonic distortion. The output layer is the operating status of the electrical equipment. The parameter update module can optimize the weights and thresholds of the BP neural network using a dynamic adaptive particle swarm optimization algorithm. The dynamic adaptive particle swarm optimization algorithm updates the parameters by dynamically adjusting the inertia weights and learning factors. The model training module can use the entropy error function as the network error objective function and combine it with the variable learning rate method to train the BP neural network until the network convergence accuracy reaches a preset threshold, thus completing the construction of the load identification model. The identification module can use the load identification model to identify the operating status of electrical equipment in real time.

11. The apparatus according to claim 10, characterized in that, In the model building module, the output layer is constructed in the following way: For devices in a start-stop two-state state, when the electrical equipment is in the running state, the expected output of the corresponding node is 1, otherwise it is 0; For multi-state devices, the expected output of the corresponding node is 1, 2, 3... otherwise it is 0.

12. The apparatus according to claim 10, characterized in that, In the parameter update module, the adjustment formulas for the inertia weights and learning factors of the dynamic adaptive particle swarm algorithm are as follows: Where, ω max and ω min These represent the maximum and minimum values ​​of the inertia weight, respectively; k2 is the current iteration number; and N2 is the maximum iteration number. The inertia weight for the (k2+1)th iteration. and The learning factor for the (k2+1)th iteration. The particle fitness value in the k2th iteration. This represents the average fitness value of the particles in the k2th iteration. It represents the minimum value of the particle fitness in the k2th iteration.

13. The apparatus according to claim 10, characterized in that, In the model training module, the expression for the entropy error function is: Where, the subscript i represents the error of the i-th learning sample; the subscript k represents the k-th output neuron; Y represents the target output; y represents the actual output of the network output layer; and m represents the number of neurons in the network output layer.

14. The apparatus according to claim 10, characterized in that, In the model training module, the formula for adjusting the variable learning rate is: Where α is the learning rate, k1 is the number of learning iterations, E is the network correction error, and λ is the learning rate error adjustment coefficient, which typically ranges from 0.01 to 0.

03.

15. The apparatus according to claim 10, characterized in that, In the data acquisition module, the total harmonic distortion k THD The calculation formula is: Where G1 is the effective value of the fundamental component, H is the preset harmonic order, and G n The effective value of all harmonic components not greater than a specific order H.

16. The apparatus according to claim 10, characterized in that, In the parameter update module, the particle velocity and position update formulas of the particle swarm algorithm are as follows: In the formula, Let c1 and c2 represent the velocity and position of the i-th particle in the k2-th iteration, respectively. c1 and c2 are learning factors, r1 and r2 are random numbers between [0,1], ω is the inertia weight, and p best,i For the optimal position of the i-th particle, g best This is the globally optimal position.

17. The apparatus according to claim 10, characterized in that, In the model construction module, the number of hidden layer neurons in the BP neural network is determined by the following empirical formula: Where l is the number of hidden layer neurons, n is the number of input layer neurons, m is the number of output layer neurons, and a is a constant between 1 and 10.

18. The apparatus according to claim 10, characterized in that, In the identification module, the load identification model has an identification rate of no less than 90% for electrical equipment.

19. A computer-readable storage medium storing computer-executable instructions for performing the method of any one of claims 1 to 9.

20. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1-9.

Images