Convolutional neural network-based power system load characteristic event identification method and system, terminal equipment and storage medium

Abstract

The invention discloses an electric power system load characteristic event identification method and system based on a convolutional neural network, terminal equipment and a storage medium, and belongs to the field of load characteristic event identification. Respectively carrying out non-disturbance and disturbance simulation processing to obtain a response track of the load operation data; and calculating sensitivity based on the trajectory and generating an evaluation index, and after screening out key nodes, inputting load data of the key nodes into a pre-trained convolutional neural network for feature event recognition. By accurately screening key nodes and combining a deep learning algorithm, the accuracy of load characteristic event identification is improved, and the problem that in the prior art, due to the fact that an artificial intelligence algorithm is lack of a key node accurate screening mechanism, the identification precision of load characteristic events of nodes in a complex nonlinear load scene is low is solved.

Description

Technical Field

[0001] The present invention relates to the field of load characteristic event recognition, and in particular to a method, system, terminal device and storage medium for identifying load characteristic events in a power system based on a convolutional neural network. Background Art

[0002] With the continuous expansion of power systems and the improvement of their intelligence levels, the volatility and complexity of power loads have increased significantly. Accurately identifying characteristic load events in power systems is of great significance for improving grid stability, optimizing power dispatch, and forecasting power demand. However, traditional load event identification methods, which rely primarily on manual rules and statistical analysis, suffer from low recognition accuracy and poor real-time performance when faced with large numbers of complex nonlinear load fluctuations, making it difficult to meet the requirements of modern power systems for efficiency and intelligence. Therefore, how to automatically identify and classify different types of load events based on power load data has become a research focus in the fields of grid optimization, load forecasting, and fault warning.

[0003] Currently, new methods for identifying load signature events in power systems primarily utilize artificial intelligence algorithms. However, while these algorithms excel in solving nonlinear, multidimensional, and multi-objective problems, they lack the ability to extract key nodes in scenarios involving distributed energy integration and complex grid topologies, making it difficult to accurately identify load signature events.

[0004] Therefore, the existing technology lacks a precise screening mechanism for key nodes, resulting in low recognition accuracy of load characteristic events in complex nonlinear load scenarios. Summary of the Invention

[0005] The present invention provides a method, system, terminal device and storage medium for identifying load characteristic events in an electric power system based on a convolutional neural network, which can solve the problem of low accuracy in identifying load characteristic events of nodes in complex nonlinear load scenarios due to the lack of a precise screening mechanism for key nodes using artificial intelligence algorithms in existing technologies.

[0006] In order to solve the above technical problems, an embodiment of the present invention provides a method for identifying power system load characteristic events based on a convolutional neural network, comprising:

[0007] Collect load operation data of each node of the power system in real time; wherein the load operation data of each node includes node voltage, system frequency, active power and reactive power;

[0008] The node voltage, system frequency, active power and reactive power of each node collected in real time are used as the setting parameters of each node in the system simulation model. The system simulation model is subjected to undisturbed simulation processing and disturbed simulation processing respectively, and the undisturbed response trajectory and disturbed response trajectory of the load operation data of each node are obtained respectively;

[0009] Calculate the sensitivity of the load operation data of each node based on the unperturbed response trajectory and the perturbed response trajectory of the load operation data of each node;

[0010] According to the sensitivity of the load operation data of each node, the evaluation index of each node is calculated, and the node with the largest evaluation index is regarded as the key node;

[0011] The load operation data of the key nodes are input into the trained convolutional neural network model, so that the trained convolutional neural network model can identify load characteristic events according to the load data of the key nodes to obtain the load characteristic events of the key nodes.

[0012] Furthermore, the sensitivity of the load operation data of each node is calculated based on the undisturbed response trajectory and the disturbed response trajectory of the load operation data of each node, including:

[0013] According to the unperturbed response trajectory and perturbed response trajectory of the load operation data of each node, the disturbance value and the total number of samples, the sensitivity of the load operation data of each node is calculated by the following formula;

[0014]

[0015] in, is the sensitivity of the i-th load operation data of each node; is the disturbance response trajectory of the i-th load operation data of each node; is the unperturbed response trajectory of the i-th load operation data of each node; δ is the disturbance value; T is the total number of samples,

[0016] Furthermore, the calculation formula of the evaluation index of each node is:

[0017]

[0018] Among them, K n→m is the evaluation index, which represents the impact of node n on node m; N is the number of load operation data; m, n are node numbers; m,global is the unperturbed response trajectory of the mth node; θ i Represents the i-th load operation data; is the unperturbed response trajectory O of the mth node m,global Sensitivity to the i-th parameter of the m-th node; is the unperturbed response trajectory O of the mth node m,global Sensitivity of the nth node to the i-th parameter.

[0019] Furthermore, the model training of the convolutional neural network model includes:

[0020] Obtaining historical load operation data and corresponding actual load characteristic events for each node in the power system; wherein the historical load operation data for each node includes historical node voltage, historical system frequency, historical active power, and historical reactive power; the actual load characteristic events include: peak load, valley load, load fluctuation, load mutation, and extreme load;

[0021] The collected load operation history data of each node is used as the setting parameters of each node in the system simulation model. The system simulation model is subjected to undisturbed simulation processing and disturbed simulation processing respectively, and the undisturbed response trajectory and disturbed response trajectory of the load operation history data of each node are obtained respectively;

[0022] Calculate the sensitivity of the load operation history data of each node based on the undisturbed response trajectory and disturbed response trajectory of the load operation history data of each node;

[0023] According to the sensitivity of the historical load operation data of each node, the evaluation index of each node is calculated, and the node with the largest evaluation index is regarded as the key node;

[0024] Inputting the load operation history data of the key nodes into the convolutional neural network model to be trained, so that the convolutional neural network model to be trained can identify load characteristic events according to the load data of the key nodes and obtain predicted load characteristic events of the key nodes;

[0025] According to the predicted key node load characteristic events and actual load characteristic events, the loss value is calculated, and the parameters of the convolutional neural network model are optimized according to the loss value until the loss value converges to obtain a trained convolutional neural network model.

[0026] Based on the above method embodiment, the present invention provides a corresponding system embodiment;

[0027] An embodiment of the present invention provides a power system load characteristic event recognition system based on a convolutional neural network, comprising: a data acquisition module, a simulation module, a sensitivity calculation module, a key node screening module, and a load characteristic event recognition module;

[0028] The data acquisition module is used to collect load operation data of each node of the power system in real time; wherein the load operation data of each node includes node voltage, system frequency, active power and reactive power;

[0029] The simulation module is used to use the node voltage, system frequency, active power and reactive power of each node collected in real time as the setting parameters of each node in the system simulation model, perform undisturbed simulation processing and disturbed simulation processing on the system simulation model, and obtain the undisturbed response trajectory and disturbed response trajectory of the load operation data of each node respectively;

[0030] The sensitivity calculation module is used to calculate the sensitivity of the load operation data of each node according to the unperturbed response trajectory and the perturbed response trajectory of the load operation data of each node;

[0031] The key node screening module is used to calculate the evaluation index of each node according to the sensitivity of the load operation data of each node, and select the node with the largest evaluation index as the key node;

[0032] The load characteristic event recognition module is used to input the load operation data of the key node into the trained convolutional neural network model, so that the trained convolutional neural network model can perform load characteristic event recognition based on the load data of the key node to obtain the load characteristic event of the key node.

[0033] Furthermore, the sensitivity of the load operation data of each node is calculated based on the undisturbed response trajectory and the disturbed response trajectory of the load operation data of each node, including:

[0034] According to the unperturbed response trajectory and perturbed response trajectory of the load operation data of each node, the disturbance value and the total number of samples, the sensitivity of the load operation data of each node is calculated by the following formula;

[0035]

[0036] in, is the sensitivity of the i-th load operation data of each node; is the disturbance response trajectory of the i-th load operation data of each node; is the unperturbed response trajectory of the i-th load operation data of each node; δ is the disturbance value; T is the total number of samples,

[0037] Furthermore, the calculation formula of the evaluation index of each node is:

[0038]

[0039] Among them, K n→m is the evaluation index, which represents the impact of node n on node m; N is the number of load operation data; m, n are node numbers; m,global is the unperturbed response trajectory of the mth node; θ i Represents the i-th load operation data; is the unperturbed response trajectory O of the mth node m,global Sensitivity to the i-th parameter of the m-th node; is the unperturbed response trajectory O of the mth node m,global Sensitivity of the nth node to the i-th parameter.

[0040] Furthermore, the model training of the convolutional neural network model includes:

[0041] Obtaining historical load operation data and corresponding actual load characteristic events for each node in the power system; wherein the historical load operation data for each node includes historical node voltage, historical system frequency, historical active power, and historical reactive power; the actual load characteristic events include: peak load, valley load, load fluctuation, load mutation, and extreme load;

[0042] The collected load operation history data of each node is used as the setting parameters of each node in the system simulation model. The system simulation model is subjected to undisturbed simulation processing and disturbed simulation processing respectively, and the undisturbed response trajectory and disturbed response trajectory of the load operation history data of each node are obtained respectively;

[0043] Calculate the sensitivity of the load operation history data of each node based on the undisturbed response trajectory and disturbed response trajectory of the load operation history data of each node;

[0044] According to the sensitivity of the historical load operation data of each node, the evaluation index of each node is calculated, and the node with the largest evaluation index is regarded as the key node;

[0045] Inputting the load operation history data of the key nodes into the convolutional neural network model to be trained, so that the convolutional neural network model to be trained can identify load characteristic events according to the load data of the key nodes and obtain predicted load characteristic events of the key nodes;

[0046] According to the predicted key node load characteristic events and actual load characteristic events, the loss value is calculated, and the parameters of the convolutional neural network model are optimized according to the loss value until the loss value converges to obtain a trained convolutional neural network model.

[0047] Based on the above-mentioned method embodiment, another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements a power system load characteristic event identification method based on a convolutional neural network as described in the above-mentioned embodiment.

[0048] Based on the above method embodiment, another embodiment of the present invention provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the power system load characteristic event identification method based on convolutional neural network described in the above embodiment.

[0049] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0050] The present invention first uses the node voltage, system frequency, active power and reactive power of each node collected in real time as the setting parameters of each node in the system simulation model, performs undisturbed simulation processing and disturbed simulation processing on the system simulation model respectively, and obtains the undisturbed response trajectory and disturbed response trajectory of the load operation data of each node respectively; then, according to the undisturbed response trajectory and disturbed response trajectory of the load operation data of each node, the sensitivity of the load operation data of each node is calculated; then, according to the sensitivity of the load operation data of each node, the evaluation index of each node is calculated, and the node with the largest evaluation index is taken as the key node; finally, the load data of the key node is input into the trained convolutional neural network model, so that the trained convolutional neural network model performs load characteristic event recognition to obtain the load characteristic event of the key node; that is, the key nodes are first accurately screened out, and the convolutional neural network model is used to perform load characteristic event recognition according to the load data of the screened key nodes, so as to realize accurate recognition of the load characteristic event type, thereby solving the problem that the existing technology uses artificial intelligence algorithms with low recognition accuracy of the load characteristic events of nodes in complex nonlinear load scenarios due to the lack of a precise key node screening mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A flowchart of a method for identifying power system load characteristic events based on a convolutional neural network is provided in an embodiment of the present invention;

[0052] Figure 2 A structural module diagram of a power system load characteristic event recognition system based on a convolutional neural network provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0054] Example 1:

[0055] Reference Figure 1 : A flowchart of a method for identifying load characteristic events in a power system based on a convolutional neural network provided by an embodiment of the present invention; to address the problem of low accuracy in identifying load characteristic events of nodes in complex nonlinear load scenarios due to the lack of a precise screening mechanism for key nodes using artificial intelligence algorithms in the prior art, the method includes at least the following steps:

[0056] Step S1: real-time collection of load operation data of each node in the power system; wherein the load operation data of each node includes node voltage, system frequency, active power and reactive power;

[0057] For example, the power load management terminal and the master station system can be used for data collection; the power load management terminal can automatically collect the load operation data of each node, including node voltage, system frequency, active power and reactive power, etc.; this terminal supports multiple communication interfaces, such as GPRS data communication module, CDMA data communication module, dial-up MODEM module, etc., and can flexibly select the communication method according to the actual scenario to realize remote transmission of data; the master station system is responsible for receiving, storing and managing the collected data. It adopts a modular design and supports multiple methods such as GPRS, CDMA, public wireless data communication and wired network communication. It has a multi-tasking, multi-threaded concurrent communication scheduling management mechanism, which can realize parallel collection of multiple collection workstations to ensure the efficiency and stability of data collection; for example, by installing the power load management terminal at each node and setting appropriate communication parameters, the real-time collected load operation data can be transmitted to the master station system to provide a data basis for subsequent analysis and processing.

[0058] Step S2: Using the node voltage, system frequency, active power, and reactive power of each node collected in real time as setting parameters for each node in the system simulation model, performing undisturbed simulation processing and disturbed simulation processing on the system simulation model, respectively, to obtain the undisturbed response trajectory and disturbed response trajectory of the load operation data of each node;

[0059] In this embodiment, the system simulation model can be constructed using MATLAB / Simulink software, and the node voltage, system frequency, active power and reactive power of each node collected in real time are input as setting parameters into the corresponding nodes in the system simulation model constructed by MATLAB / Simulink. The system simulation model is subjected to undisturbed simulation processing and disturbed simulation processing respectively, and the undisturbed response trajectory and disturbed response trajectory of the load operation data of each node are obtained respectively.

[0060] Step S3: Calculate the sensitivity of the load operation data of each node according to the unperturbed response trajectory and the perturbed response trajectory of the load operation data of each node;

[0061] In this embodiment, the sensitivity of the load operation data of each node is calculated based on the unperturbed response trajectory and the perturbed response trajectory of the load operation data of each node, including:

[0062] According to the unperturbed response trajectory and perturbed response trajectory of the load operation data of each node, the disturbance value and the total number of samples, the sensitivity of the load operation data of each node is calculated by the following formula;

[0063]

[0064] in, is the sensitivity of the i-th load operation data of each node; is the disturbance response trajectory of the i-th load operation data of each node; is the unperturbed response trajectory of the i-th load operation data of each node; δ is the disturbance value; T is the total number of samples,

[0065] Step S4: Calculate the evaluation index of each node according to the sensitivity of the load operation data of each node, and take the node with the largest evaluation index as the key node;

[0066] In this embodiment, the calculation formula of the evaluation index of each node is:

[0067]

[0068] Among them, K n→m is the evaluation index, which represents the impact of node n on node m; N is the number of load operation data; m, n are node numbers; m,global is the unperturbed response trajectory of the mth node; θ i Represents the i-th load operation data; is the unperturbed response trajectory O of the mth node m,global Sensitivity to the i-th parameter of the m-th node; is the unperturbed response trajectory O of the mth node m,global Sensitivity of the nth node to the i-th parameter.

[0069] Step S5: Input the load operation data of the key nodes into the trained convolutional neural network model, so that the trained convolutional neural network model can identify load characteristic events according to the load data of the key nodes to obtain the load characteristic events of the key nodes.

[0070] In this embodiment, the convolutional neural network model includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer;

[0071] The input layer is used to obtain the load operation data of the key nodes and perform normalization processing on the load operation data of the key nodes to obtain the processed load operation data of the key nodes;

[0072] The convolution layer is used to perform convolution processing and nonlinear transformation processing on the processed key node load operation data to obtain convolution feature data of the key nodes; wherein the nonlinear transformation processing uses the Relu function as the activation function to perform nonlinear transformation, and the specific expression of the Relu function is: Where, Represents the output value of the convolution operation, is the weight of the i-th filter kernel in the l-th layer, is the deviation of the i-th filter kernel in the l-th layer, x l (j) is the jth local region of the lth layer, After the convolution operation on the lth layer, the i-th neuron in the l+1th layer corresponds to the input value of the j-th local area. express The activation value of

[0073] The pooling layer is used to sample the convolution feature data of the key nodes using the maximum pooling sampling function to reduce the feature space size of the convolution feature data of the key nodes; wherein the maximum pooling sampling function is specifically: Where, represents the value of the t-th neuron in the i-th feature map of the l-th layer; W represents the width of the pooling area, Represents the pooling result of the jth local area of ​​the i-th feature map in the l+1th layer.

[0074] The fully connected layer is used to map the feature data output by the pooling layer into a high-dimensional space to achieve classification and recognition of different load feature events and obtain predicted key node load feature events;

[0075] The output layer is used to output predicted key node load characteristic events.

[0076] In this embodiment, the model training of the convolutional neural network model includes:

[0077] Obtaining historical load operation data and corresponding actual load characteristic events for each node in the power system; wherein the historical load operation data for each node includes historical node voltage, historical system frequency, historical active power, and historical reactive power; the actual load characteristic events include: peak load, valley load, load fluctuation, load mutation, and extreme load;

[0078] The collected load operation history data of each node is used as the setting parameters of each node in the system simulation model. The system simulation model is subjected to undisturbed simulation processing and disturbed simulation processing respectively, and the undisturbed response trajectory and disturbed response trajectory of the load operation history data of each node are obtained respectively;

[0079] Calculate the sensitivity of the load operation history data of each node based on the undisturbed response trajectory and disturbed response trajectory of the load operation history data of each node;

[0080] According to the sensitivity of the historical load operation data of each node, the evaluation index of each node is calculated, and the node with the largest evaluation index is regarded as the key node;

[0081] Inputting the load operation history data of the key nodes into the convolutional neural network model to be trained, so that the convolutional neural network model to be trained can identify load characteristic events according to the load data of the key nodes and obtain predicted load characteristic events of the key nodes;

[0082] According to the predicted key node load characteristic events and actual load characteristic events, the loss value is calculated, and the parameters of the convolutional neural network model are optimized according to the loss value until the loss value converges to obtain a trained convolutional neural network model.

[0083] In this embodiment, the loss value is calculated by a loss function; wherein the loss function includes but is not limited to a mean square error loss function, a cross entropy loss function, and a mean absolute error loss function.

[0084] In this embodiment, the method for optimizing the parameters of the convolutional neural network model includes but is not limited to the stochastic gradient descent method, the mini-batch gradient descent method, and the Adam optimizer.

[0085] Example 2:

[0086] Reference Figure 2 : This is a structural module diagram of a power system load characteristic event identification system based on a convolutional neural network provided by an embodiment of the present invention. To address the problem of low recognition accuracy of node load characteristic events in complex nonlinear load scenarios caused by the lack of a precise key node screening mechanism using artificial intelligence algorithms in existing technologies, the system includes at least: a data acquisition module, a simulation module, a sensitivity calculation module, a key node screening module, and a load characteristic event identification module;

[0087] The data acquisition module is used to collect load operation data of each node of the power system in real time; wherein the load operation data of each node includes node voltage, system frequency, active power and reactive power;

[0088] The simulation module is used to use the node voltage, system frequency, active power and reactive power of each node collected in real time as the setting parameters of each node in the system simulation model, perform undisturbed simulation processing and disturbed simulation processing on the system simulation model, and obtain the undisturbed response trajectory and disturbed response trajectory of the load operation data of each node respectively;

[0089] The sensitivity calculation module is used to calculate the sensitivity of the load operation data of each node according to the unperturbed response trajectory and the perturbed response trajectory of the load operation data of each node;

[0090] The key node screening module is used to calculate the evaluation index of each node according to the sensitivity of the load operation data of each node, and select the node with the largest evaluation index as the key node;

[0091] The load characteristic event recognition module is used to input the load operation data of the key node into the trained convolutional neural network model, so that the trained convolutional neural network model can perform load characteristic event recognition based on the load data of the key node to obtain the load characteristic event of the key node.

[0092] In this embodiment, the sensitivity of the load operation data of each node is calculated based on the unperturbed response trajectory and the perturbed response trajectory of the load operation data of each node, including:

[0093] According to the unperturbed response trajectory and perturbed response trajectory of the load operation data of each node, the disturbance value and the total number of samples, the sensitivity of the load operation data of each node is calculated by the following formula;

[0094]

[0095] in, is the sensitivity of the i-th load operation data of each node; is the disturbance response trajectory of the i-th load operation data of each node; is the unperturbed response trajectory of the i-th load operation data of each node; δ is the disturbance value; T is the total number of samples,

[0096] In this embodiment, the calculation formula of the evaluation index of each node is:

[0097]

[0098] Among them, K n→m is the evaluation index, which represents the impact of node n on node m; N is the number of load operation data; m, n are node numbers; m,global is the unperturbed response trajectory of the mth node; θ i Represents the i-th load operation data; is the unperturbed response trajectory O of the mth node m,global Sensitivity to the i-th parameter of the m-th node; is the unperturbed response trajectory O of the mth node m,global Sensitivity of the nth node to the i-th parameter.

[0099] In this embodiment, the model training of the convolutional neural network model includes:

[0100] Obtaining historical load operation data and corresponding actual load characteristic events for each node in the power system; wherein the historical load operation data for each node includes historical node voltage, historical system frequency, historical active power, and historical reactive power; the actual load characteristic events include: peak load, valley load, load fluctuation, load mutation, and extreme load;

[0101] The collected load operation history data of each node is used as the setting parameters of each node in the system simulation model. The system simulation model is subjected to undisturbed simulation processing and disturbed simulation processing respectively, and the undisturbed response trajectory and disturbed response trajectory of the load operation history data of each node are obtained respectively;

[0102] Calculate the sensitivity of the load operation history data of each node based on the undisturbed response trajectory and disturbed response trajectory of the load operation history data of each node;

[0103] According to the sensitivity of the historical load operation data of each node, the evaluation index of each node is calculated, and the node with the largest evaluation index is regarded as the key node;

[0104] Inputting the load operation history data of the key nodes into the convolutional neural network model to be trained, so that the convolutional neural network model to be trained can identify load characteristic events according to the load data of the key nodes and obtain predicted load characteristic events of the key nodes;

[0105] According to the predicted key node load characteristic events and actual load characteristic events, the loss value is calculated, and the parameters of the convolutional neural network model are optimized according to the loss value until the loss value converges to obtain a trained convolutional neural network model.

[0106] Another embodiment of the present invention provides a terminal device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the method for identifying power system load signature events based on a convolutional neural network as described in the above embodiment. The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The terminal device may include, but is not limited to, a processor and a memory.

[0107] The processor may be a central processing unit (CPU), or other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, connecting various parts of the entire terminal device using various interfaces and lines.

[0108] The memory can be used to store the computer program, and the processor implements various functions of the terminal device by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created based on the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory and can also include a non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device or other volatile solid-state storage device.

[0109] Another embodiment of the present invention provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the power system load characteristic event identification method based on convolutional neural network described in the above embodiment.

[0110] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by the processor, the steps of each of the above-mentioned method embodiments can be implemented. The computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunications signal, and a software distribution medium.

[0111] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for identifying power system load characteristic events based on convolutional neural networks, characterized in that: include: Collect load operation data of each node of the power system in real time; wherein the load operation data of each node includes node voltage, system frequency, active power and reactive power; The node voltage, system frequency, active power and reactive power of each node collected in real time are used as the setting parameters of each node in the system simulation model. The system simulation model is subjected to undisturbed simulation processing and disturbed simulation processing respectively, and the undisturbed response trajectory and disturbed response trajectory of the load operation data of each node are obtained respectively; Calculate the sensitivity of the load operation data of each node based on the unperturbed response trajectory and the perturbed response trajectory of the load operation data of each node; According to the sensitivity of the load operation data of each node, the evaluation index of each node is calculated, and the node with the largest evaluation index is regarded as the key node; The load operation data of the key nodes are input into the trained convolutional neural network model, so that the trained convolutional neural network model can identify load characteristic events according to the load data of the key nodes to obtain the load characteristic events of the key nodes.

2. The method for identifying power system load characteristic events based on convolutional neural networks according to claim 1, characterized in that: Calculating the sensitivity of the load operation data of each node according to the undisturbed response trajectory and the disturbed response trajectory of the load operation data of each node includes: According to the unperturbed response trajectory and perturbed response trajectory of the load operation data of each node, the disturbance value and the total number of samples, the sensitivity of the load operation data of each node is calculated by the following formula; in, is the sensitivity of the i-th load operation data of each node; is the disturbance response trajectory of the i-th load operation data of each node; is the unperturbed response trajectory of the i-th load operation data of each node; δ is the disturbance value; T is the total number of samples, 3. The method for identifying power system load characteristic events based on convolutional neural networks according to claim 2, characterized in that: The calculation formula of the evaluation index of each node is: Among them, K n→m is the evaluation index, which represents the impact of node n on node m; N is the number of load operation data; m, n are node numbers; m,global is the unperturbed response trajectory of the mth node; θ i Represents the i-th load operation data; is the unperturbed response trajectory O of the mth node m,global Sensitivity to the i-th parameter of the m-th node; is the unperturbed response trajectory O of the mth node m,global Sensitivity of the nth node to the i-th parameter.

4. The method for identifying power system load characteristic events based on convolutional neural networks according to claim 3 is characterized in that: The model training of the convolutional neural network model includes: Obtaining historical load operation data and corresponding actual load characteristic events for each node in the power system; wherein the historical load operation data for each node includes historical node voltage, historical system frequency, historical active power, and historical reactive power; the actual load characteristic events include: peak load, valley load, load fluctuation, load mutation, and extreme load; The collected load operation history data of each node is used as the setting parameters of each node in the system simulation model. The system simulation model is subjected to undisturbed simulation processing and disturbed simulation processing respectively, and the undisturbed response trajectory and disturbed response trajectory of the load operation history data of each node are obtained respectively; Calculate the sensitivity of the load operation history data of each node based on the undisturbed response trajectory and disturbed response trajectory of the load operation history data of each node; According to the sensitivity of the historical load operation data of each node, the evaluation index of each node is calculated, and the node with the largest evaluation index is regarded as the key node; Inputting the load operation history data of the key nodes into the convolutional neural network model to be trained, so that the convolutional neural network model to be trained can identify load characteristic events according to the load data of the key nodes and obtain predicted load characteristic events of the key nodes; According to the predicted key node load characteristic events and actual load characteristic events, the loss value is calculated, and the parameters of the convolutional neural network model are optimized according to the loss value until the loss value converges to obtain a trained convolutional neural network model.

5. A power system load characteristic event recognition system based on convolutional neural network, characterized in that: include: Data acquisition module, simulation module, sensitivity calculation module, key node screening module and load characteristic event identification module; The data acquisition module is used to collect load operation data of each node of the power system in real time; wherein the load operation data of each node includes node voltage, system frequency, active power and reactive power; The simulation module is used to use the node voltage, system frequency, active power and reactive power of each node collected in real time as the setting parameters of each node in the system simulation model, perform undisturbed simulation processing and disturbed simulation processing on the system simulation model, and obtain the undisturbed response trajectory and disturbed response trajectory of the load operation data of each node respectively; The sensitivity calculation module is used to calculate the sensitivity of the load operation data of each node according to the unperturbed response trajectory and the perturbed response trajectory of the load operation data of each node; The key node screening module is used to calculate the evaluation index of each node according to the sensitivity of the load operation data of each node, and select the node with the largest evaluation index as the key node; The load characteristic event recognition module is used to input the load operation data of the key node into the trained convolutional neural network model, so that the trained convolutional neural network model can perform load characteristic event recognition based on the load data of the key node to obtain the load characteristic event of the key node.

6. The power system load characteristic event recognition system based on convolutional neural network according to claim 5 is characterized in that: Calculating the sensitivity of the load operation data of each node according to the undisturbed response trajectory and the disturbed response trajectory of the load operation data of each node includes: According to the unperturbed response trajectory and perturbed response trajectory of the load operation data of each node, the disturbance value and the total number of samples, the sensitivity of the load operation data of each node is calculated by the following formula; in, is the sensitivity of the i-th load operation data of each node; is the disturbance response trajectory of the i-th load operation data of each node; is the unperturbed response trajectory of the i-th load operation data of each node; δ is the disturbance value; T is the total number of samples, 7. The power system load characteristic event recognition system based on convolutional neural network according to claim 6 is characterized in that: The calculation formula of the evaluation index of each node is: Among them, K n→m is the evaluation index, which represents the impact of node n on node m; N is the number of load operation data; m, n are node numbers; m,global is the unperturbed response trajectory of the mth node; θ i Represents the i-th load operation data; is the unperturbed response trajectory O of the mth node m,global Sensitivity to the i-th parameter of the m-th node; is the unperturbed response trajectory O of the mth node m,global Sensitivity of the nth node to the i-th parameter.

8. The power system load characteristic event recognition system based on convolutional neural network according to claim 7 is characterized in that: The model training of the convolutional neural network model includes: Obtaining historical load operation data and corresponding actual load characteristic events for each node in the power system; wherein the historical load operation data for each node includes historical node voltage, historical system frequency, historical active power, and historical reactive power; the actual load characteristic events include: peak load, valley load, load fluctuation, load mutation, and extreme load; The collected load operation history data of each node is used as the setting parameters of each node in the system simulation model. The system simulation model is subjected to undisturbed simulation processing and disturbed simulation processing respectively, and the undisturbed response trajectory and disturbed response trajectory of the load operation history data of each node are obtained respectively; Calculate the sensitivity of the load operation history data of each node based on the undisturbed response trajectory and disturbed response trajectory of the load operation history data of each node; According to the sensitivity of the historical load operation data of each node, the evaluation index of each node is calculated, and the node with the largest evaluation index is regarded as the key node; Inputting the load operation history data of the key nodes into the convolutional neural network model to be trained, so that the convolutional neural network model to be trained can identify load characteristic events according to the load data of the key nodes and obtain predicted load characteristic events of the key nodes; According to the predicted key node load characteristic events and actual load characteristic events, the loss value is calculated, and the parameters of the convolutional neural network model are optimized according to the loss value until the loss value converges to obtain a trained convolutional neural network model.

9. A terminal device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements a method for identifying power system load characteristic events based on a convolutional neural network as described in any one of claims 1 to 4.

10. A storage medium, characterized in that: The storage medium includes a stored computer program, wherein, when the computer program is running, the device where the storage medium is located is controlled to execute a power system load characteristic event identification method based on a convolutional neural network as described in any one of claims 1 to 4.

Images