Operation reliability evaluation method and device, electronic equipment and computer storage medium

Abstract

The invention provides an operation reliability evaluation method and device, electronic equipment and a computer storage medium. After obtaining a to-be-evaluated system state, inputting the to-be-evaluated system state into the prediction model, and outputting to obtain the minimum load shedding amount of the power system; and determining an operation reliability evaluation result according to the minimum load shedding amount of the power system. The prediction model is a multi-kernel collaborative graph convolutional neural network. The extraction capability of the neural network on node features, topological features and physical correlation features is improved on the architecture level, so that the accuracy of the neural network on minimum load shedding calculation is effectively improved, and the adaptability to a complex variable topology scene during operation reliability evaluation is improved.

Description

Technical Field

[0001] This application relates to the technical field of power systems and their automation, and particularly relates to a method, device, electronic device, and computer storage medium for evaluating operating reliability. Background Art

[0002] Operating reliability assessment is an important tool for power system operation risk assessment and early warning. In the context of the vigorous development of renewable energy in various countries around the world, due to the intermittency and randomness of wind energy and solar energy, the power system faces strong uncertainties, and it is necessary to conduct real-time operating reliability assessment.

[0003] However, with the collection of reliability data of important power equipment, both traditional assessment methods and data-driven assessment methods of operating reliability have been challenged. On the one hand, traditional assessment methods, such as the Monte Carlo simulation method, need to iteratively solve the minimum cut load optimization problem under a large number of system states. However, the computational complexity in actual large-scale power systems is relatively high, the time consumption is long, and the cumulative computational burden hinders the online assessment of operating reliability. On the other hand, due to the collection of a large amount of important power equipment data, it is necessary to simulate more complex and diverse system states. In particular, there will be a large number of scenarios with topological changes, making it more difficult for neural network training and learning, and there are also adaptability problems with data-driven methods with fast calculation speeds. Summary of the Invention

[0004] In view of this, this application provides a method, device, electronic device, and computer storage medium for evaluating operating reliability, which can effectively improve the accuracy of the neural network in calculating the minimum cut load and the adaptability to complex variable-topology scenarios during operating reliability assessment.

[0005] The first aspect of this application provides a method for evaluating operating reliability, including:

[0006] Obtain the system state to be evaluated;

[0007] Input the system state to be evaluated into the prediction model, and output the minimum cut load of the power system; wherein, the prediction model is a multi-core collaborative graph convolutional neural network;

[0008] Determine the operating reliability assessment result according to the minimum cut load of the power system.

[0009] Optionally, the construction method of the prediction model includes:

[0010] Construct a training sample set;

[0011] Establish an initial minimum cut load model;

[0012] Construct a graph convolutional layer that is precisely embedded in the initial minimum cut load model;

[0013] Build a multi-core collaborative graph convolutional neural network based on the said graph convolutional layer;

[0014] Use the said training sample set to train the multi-core collaborative graph convolutional neural network to obtain a prediction model.

[0015] Optionally, the said operation reliability evaluation method further includes:

[0016] In the process of determining the operation reliability evaluation result according to the minimum load shedding amount of the power system, judge the convergence criterion according to the maximum number of samples and the variance of the reliability index.

[0017] Optionally, the said construction of the training sample set includes:

[0018] Obtain the power system state through Monte Carlo sampling according to the typical operation scenarios of the power system;

[0019] For each power system state, solve the minimum load shedding amount under the said power system state through the interior point method;

[0020] Construct a training sample set according to the minimum load shedding amounts under all the said power system states.

[0021] Optionally, the said construction of the graph convolutional layer accurately embedded in the initial minimum load shedding model includes:

[0022] Embed the Gauss-Seidel iteration of the power flow equation into the graph convolution to construct the graph convolutional layer accurately embedded in the initial minimum load shedding model.

[0023] Optionally, the said building of the multi-core collaborative graph convolutional neural network based on the said graph convolutional layer includes:

[0024] Determine the forward propagation formula according to the node feature convolution and the node feature neighborhood aggregation calculation function;

[0025] Based on the adjacency matrix, determine the first AGGRE function;

[0026] Based on the aggregation weight matrix of impedance, determine the second AGGRE function;

[0027] Normalize the aggregation weight matrix of the impedance to obtain the aggregation weight matrix of the target impedance;

[0028] Based on the aggregation weight matrix of the target impedance, determine the third AGGRE function;

[0029] Based on the transformed power flow equation, determine the fourth AGGRE function;

[0030] Dynamically weighted combination is performed on the first AGGRE function, the second AGGRE function, the third AGGRE function, and the fourth AGGRE function to obtain a multi-core collaborative graph convolutional neural network.

[0031] The second aspect of this application provides an operation reliability evaluation device, including:

[0032] An acquisition unit for acquiring the system state to be evaluated;

[0033] An analysis unit for inputting the system state to be evaluated into a prediction model and outputting the minimum load shedding amount of the power system; wherein, the prediction model is a multi-core collaborative graph convolutional neural network;

[0034] An evaluation unit for determining the operation reliability evaluation result according to the minimum load shedding amount of the power system.

[0035] Optionally, the construction unit of the prediction model includes:

[0036] A training sample set construction unit for constructing a training sample set;

[0037] A minimum load shedding model establishment unit for establishing an initial minimum load shedding model;

[0038] A graph convolutional layer construction unit for constructing a graph convolutional layer precisely embedded in the initial minimum load shedding model;

[0039] A graph convolutional neural network establishment unit for establishing a multi-core collaborative graph convolutional neural network according to the graph convolutional layer;

[0040] A training unit for training the multi-core collaborative graph convolutional neural network with the training sample set to obtain a prediction model.

[0041] Optionally, the operation reliability evaluation device further includes:

[0042] A convergence judgment unit for judging the convergence criterion according to the maximum number of samples and the variance of the reliability index during the process of determining the operation reliability evaluation result according to the minimum load shedding amount of the power system.

[0043] Optionally, the training sample set construction unit includes:

[0044] A sampling unit for obtaining the power system state through Monte Carlo sampling according to the typical operation scenarios of the power system;

[0045] A solving unit for solving the minimum load shedding amount under the power system state through the interior point method for each power system state;

[0046] A training sample set construction subunit, configured to construct a training sample set according to the minimum load shedding amounts under all the power system states.

[0047] Optionally, the graph convolutional layer construction unit includes:

[0048] A graph convolutional layer construction subunit, configured to embed the Gauss-Seidel iteration of the power flow equation into graph convolution to construct a graph convolutional layer precisely embedded in the initial minimum load shedding model.

[0049] Optionally, the graph convolutional neural network establishment unit includes:

[0050] A first determination unit, configured to determine a forward propagation formula according to a node feature convolution and a node feature neighborhood aggregation calculation function;

[0051] A second determination unit, configured to determine a first AGGRE function based on an adjacency matrix;

[0052] A third determination unit, configured to determine a second AGGRE function based on an aggregation weight matrix of impedance;

[0053] A normalization unit, configured to perform normalization processing on the aggregation weight matrix of impedance to obtain an aggregation weight matrix of target impedance;

[0054] A fourth determination unit, configured to determine a third AGGRE function based on the aggregation weight matrix of the target impedance;

[0055] A fifth determination unit, configured to determine a fourth AGGRE function based on the transformed power flow equation;

[0056] A dynamic weighted merging unit, configured to perform dynamic weighted merging on the first AGGRE function, the second AGGRE function, the third AGGRE function, and the fourth AGGRE function to obtain a multi-core collaborative graph convolutional neural network.

[0057] A third aspect of the present application provides an electronic device, including:

[0058] One or more processors;

[0059] A storage device, on which one or more programs are stored;

[0060] When the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the operation reliability evaluation method according to any item in the first aspect.

[0061] A fourth aspect of the present application provides a computer storage medium, on which a computer program is stored, wherein the computer program, when executed by a processor, implements the operation reliability evaluation method according to any item in the first aspect.

[0062] As can be seen from the above solution, this application provides a method, apparatus, electronic device, and computer storage medium for evaluating operating reliability. After obtaining the system state to be evaluated, the system state to be evaluated is input into a prediction model, and the minimum load shedding amount of the power system is output; according to the minimum load shedding amount of the power system, the operating reliability evaluation result is determined. The prediction model of this application is a multi-core collaborative graph convolutional neural network; at the architecture level, it improves the neural network's ability to extract node features, topological features, and their physical association features, thereby effectively improving the accuracy of the neural network's calculation of the minimum load shedding and enhancing the adaptability to complex variable topology scenarios during operating reliability evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on the provided drawings.

[0064] Figure 1 It is a specific flowchart of a method for evaluating operating reliability provided by an embodiment of this application;

[0065] Figure 2 It is a flowchart of a method for constructing a prediction model provided by another embodiment of this application;

[0066] Figure 3 It is a flowchart of a method for establishing a multi-core collaborative graph convolutional neural network provided by another embodiment of this application;

[0067] Figure 4 It is an architecture diagram of a multi-core collaborative graph convolutional neural network provided by another embodiment of this application;

[0068] Figure 5 It is an error curve of the training set during the training of different neural networks provided by another embodiment of this application;

[0069] Figure 6 It is an error curve of the test set during the training of different neural networks provided by another embodiment of this application;

[0070] Figure 7 It is the attention value of node feature e in the multi-core collaborative graph convolutional neural network provided by another embodiment of this application;

[0071] Figure 8 It is the attention value of node feature f in the multi-core collaborative graph convolutional neural network provided by another embodiment of this application;

[0072] Figure 9 The operation reliability evaluation results at different times provided by another embodiment of the present application;

[0073] Figure 10 The schematic diagram of an operation reliability evaluation device provided by another embodiment of the present application;

[0074] Figure 11 The schematic diagram of an electronic device for implementing an operation reliability evaluation method provided by another embodiment of the present application. Detailed implementation manners

[0075] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0076] The term "including" and its variations used herein are open-ended, that is, "including but not limited to". The term "based on" is "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0077] It should be noted that the information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with the relevant laws, regulations, and standards of relevant countries and regions.

[0078] It should be noted that the concepts such as "first" and "second" mentioned in the present application are only used to distinguish different devices, modules, or units, and are not used to limit the order or interdependence relationship of the functions performed by these devices, modules, or units.

[0079] It should be noted that the modifications of "one" and "multiple" mentioned in the present application are illustrative rather than restrictive. Those skilled in the art should understand that unless otherwise clearly stated in the context, it should be understood as "one or more".

[0080] The embodiments of the present application provide an operation reliability evaluation method, as Figure 1 shown, specifically including the following steps:

[0081] S101. Obtain the system status to be evaluated.

[0082] In the specific implementation process of this application, the system status can be sampled. The system status is usually sampled according to the probability distribution of load and renewable energy and the failure probability of important power equipment, and the system status to be evaluated is extracted, which is not limited here.

[0083] Among them, the system status includes but is not limited to equipment status, load level, new energy output, etc., which is not limited here.

[0084] S102. Input the system status to be evaluated into the prediction model, and output the minimum load shedding amount of the power system.

[0085] Among them, the prediction model is a multi-core collaborative graph convolutional neural network.

[0086] Optionally, in another embodiment of this application, an implementation manner of the construction method of the prediction model is as Figure 2 shown, including:

[0087] S201. Construct a training sample set.

[0088] In the specific implementation process of this application, it can be but not limited to obtaining the power system status through Monte Carlo sampling according to the typical operation scenarios of the power system, and then, for each power system status, solving the minimum load shedding amount under the power system status through the interior point method. Finally, a training sample set is constructed according to the minimum load shedding amounts under all power system statuses, which is not limited here.

[0089] S202. Establish an initial minimum load shedding model.

[0090] In the specific implementation process of this application, the establishment of the initial minimum load shedding model can be as follows:

[0091] ;

[0092] ;

[0093] ;

[0094] ;

[0095] ;

[0096] ;

[0097] ;

[0098] ;

[0099] ;

[0100] ;

[0101] Among them, and are the active and reactive load curtailments of node i, respectively; is the cost coefficient of * (taking in the above formula as an example, it represents the active cost coefficient of bus node i); and are the active and reactive loads of node i, respectively; and are the magnitudes of the active and reactive power outputs of the generator connected to node i; and are the voltage magnitude and phase angle of node i, respectively; is the phase angle difference between the i-th node and the j-th node; and represent the elements in the i-th row and j-th column of the nodal conductance matrix and nodal susceptance matrix, respectively; represents the active power between the i-th node and the j-th node; represents the generator status; is the active power output of generator i at the previous moment; is the maximum ramp rate of generator i; , , are the sets of generators, nodes, and branches, respectively; and represent the upper and lower limits of * respectively (for example: represents the upper limit of the active power between the i-th node and the j-th node, the lower limit of the active power between the i-th node and the j-th node).

[0102] S203. Construct a graph convolutional layer with precise embedding of the initial minimum load shedding model.

[0103] It should be noted that in order to achieve the precise embedding of the initial minimum load shedding model, it is necessary to maintain physical rules in the graph convolutional feature aggregation calculation. Therefore, no assumptions can be made in the derivation based on the power flow model. Therefore, in the specific implementation process of this application, the Gauss-Seidel iteration of the power flow equation can be used but not limited to be embedded into the graph convolution to construct a graph convolutional layer with precise embedding of the initial minimum load shedding model.

[0104] Among them, the Gauss-Seidel iteration will not destroy the initial power flow equation, and the physical rules can be well preserved. More importantly, it can be directly applied to solve the power flow equation.

[0105] The complex form of the power flow model is as follows:

[0106] (1)

[0107] Where Y is the nodal admittance matrix in complex form; V is the nodal voltage in complex form; I is the nodal injection current in complex form; and S is the nodal injection power in complex form.

[0108] Decompose the matrix Y into a diagonal form. The diagonal element matrix is denoted as and the non - diagonal element matrix is denoted as Equation (1) can be rewritten as Equation (2). Therefore, the Gauss - Seidel iteration function of for the i - th node can be rewritten as Equation (3).

[0109] (2)

[0110] (3)

[0111] Where and are the -th and -th iterations of respectively; is the element in the i - th row and j - th column of the admittance matrix Y; is the i - th element of the nodal injection power vector S; represents the set connected to node i, and is the element in the i - th row and i - th column of the admittance matrix Y.

[0112] It can be seen that Equation (3) is a nodal feature aggregation function that uses the system state (nodal injection power S) and grid parameters (conductance and susceptance matrices G, B). There are no assumptions in the derivation process, and Equation (3) still satisfies physical laws (such as Kirchhoff's Current Law (KCL) and Kirchhoff's Voltage Law (KVL), etc.). Therefore, the present invention uses Equation (3) as the neighborhood feature aggregation function in graph convolution and designs a new graph convolution method. However, all elements in Equation (3) are complex numbers, making it difficult to directly utilize them in the forward propagation of neural networks. The present invention decouples the nodal features in the Cartesian coordinate system, where Then the nodal feature aggregation function is expressed as Equation (4), and the expressions of its real and imaginary parts are shown in Equation (5).

[0113] (4)

[0114] (5)

[0115] Wherein, and are the i-node features in the th iteration; and respectively represent the elements of the i-th row and i-th column of the conductance matrix G and the susceptance matrix B; and represent the features aggregated from the adjacent nodes of node i, expressed as:

[0116] (6)

[0117] Wherein, and are the real part and the imaginary part of respectively.

[0118] In equations (5) and (6), and are in the denominator. For different nodes in a given topology, is a non-zero constant. However, and represent the output features of the hidden layer and may be zero, which may cause errors in the forward propagation calculation of the neural network. To avoid this situation, the present invention adds a small constant in . Finally, the matrix forms of the node feature neighborhood aggregation functions and in graph convolution can be written as:

[0119] (7)

[0120] (8)

[0121] (9)

[0122] Wherein, P and Q respectively represent the active and reactive powers injected by the node; is the Hadamard product; and respectively represent the node conductance matrix and susceptance matrix without diagonal elements; and are the diagonal matrices of the node conductance matrix and susceptance matrix respectively; and represent the coupled node feature vectors of the th layer.

[0123] S204. Establish a multi-core collaborative graph convolutional neural network based on the graph convolutional layer.

[0124] Optionally, in another embodiment of the present application, an implementation manner of step S204 is as follows Figure 3 shown, including:

[0125] S301. Determine the forward propagation formula according to the node feature convolution and the node feature neighborhood aggregation calculation function.

[0126] It can be understood that graph convolution calculation is different from traditional matrix or image convolution. It decomposes convolution into two steps: node feature neighborhood aggregation and node feature convolution calculation, so as to handle the problem of inconsistent node degrees in the graph. Its forward propagation formula can be expressed as:

[0127] ;

[0128] where represents the feature of node i in the -th layer of the neural network; N(i) represents the set of adjacent nodes of node i and itself; and are the node feature convolution and the node feature neighborhood aggregation calculation function respectively.

[0129] S302. Determine the first AGGRE function based on the adjacency matrix.

[0130] In order to reduce the number of trainable parameters in the graph convolutional neural network, most AGGRE functions only use known parameters, and the way to determine the AGGRE function becomes the primary task of constructing the graph convolutional neural network. The AGGRE function of the first type of graph convolutional neural network (the first AGGRE function) is constructed based on the adjacency matrix A, and the weight size of the neighborhood node features is determined by spectral graph theory derivation. Its mathematical formula in matrix form can be expressed as:

[0131] ;

[0132] where is the input feature with n nodes and k-dimensional features; and represent the normalized adjacency matrix and the node degree matrix, , .

[0133] S303. Determine the second AGGRE function based on the impedance-based aggregation weight matrix.

[0134] In the application of graph convolutional neural networks in power systems, physical parameters and physical models that can represent the influence of neighboring nodes are often used to construct the AGGRE function, such as branch impedance, power flow model, etc. First, the aggregation weights of neighborhood features are constructed using branch impedance. This type of AGGRE method uses a Gaussian kernel , where is the scaling factor. In different power systems, needs to be adjusted to ensure that the line weights are within a reasonable range and the lines are not ignored due to too small weights. Then, the aggregation function of the second type of graph convolutional neural network (the second AGGRE function) can be expressed as:

[0135] (10)

[0136] where W is the weight matrix calculated using line impedance (the aggregation weight matrix of impedance).

[0137] S304. Standardize the aggregation weight matrix of impedance to obtain the aggregation weight matrix of the target impedance.

[0138] To further improve the feature extraction ability of the neural network, the third type of graph convolutional neural network, based on the aggregation weight matrix W of impedance, further uses the standardization processing technique of the first type of graph convolutional neural network to process the W matrix and obtain the aggregation weight matrix of the target impedance :

[0139] .

[0140] S305. Based on the aggregation weight matrix of the target impedance, determine the third AGGRE function.

[0141] Continuing with the above example, the third AGGRE function can be expressed as:

[0142] (11)

[0143] where .

[0144] S306. Based on the transformed power flow equation, determine the fourth AGGRE function.

[0145] To further enhance the feature extraction ability of the neural network, the fourth type of graph convolutional neural network also embeds a power flow model in the Cartesian coordinate system. This model sets and as the coupled node features in the graph and uses the power flow model as the neighborhood aggregation function to update them. Among them, it is mainly assumed that the feature of the central node i is unknown and is updated by its neighborhood, then the one containing or The items of and are retained on the right side, and other elements are moved to the left side. By solving

[0146] in the power flow equation after re-conversion, the fourth node feature aggregation function (the fourth AGGRE function) is derived as:

[0147] (12)

[0148] where and are the coupled node feature vectors of the th layer; is the Hadamard product, and:

[0149] ;

[0150] ;

[0151] ;

[0152] ;

[0153] where and are the node conductance matrix and node susceptance matrix without diagonal elements; and are the diagonal matrices of the node conductance matrix and susceptance matrix.

[0154] S307. Dynamically weighted combination of the first AGGRE function, the second AGGRE function, the third AGGRE function, and the fourth AGGRE function is performed to obtain a multi-core collaborative graph convolutional neural network.

[0155] As Figure 4 shown, the present invention further uses an attention mechanism to perform dynamic weighted combination of different graph convolution methods to construct a multi-core collaborative graph convolutional neural network architecture.

[0156] That is to say, the multi-core collaborative graph convolutional neural network in the present invention uses multiple neighborhood aggregation calculation methods (such as Equation (7), Equation (10), Equation (11), Equation (12)) in the graph convolution layer to break through the problem of insufficient feature extraction of a single neighborhood aggregation method. Second, a self-attention mechanism is used to perform collaborative processing on multiple graph convolution kernels, dynamically adjust the attention of different node feature neighborhood aggregation methods, and extract important features that affect the prediction result. When is all the node coupling features input to the th layer, the aggregated features of the sth type of neighborhood aggregation method are, and it can be expressed as:

[0157] ;

[0158] The present invention integrates the neighborhood aggregation methods of equations (7), (10), (11), and (12) into a multi-core collaborative graph convolutional neural network. Therefore, . In equations (10) and (11), the coupling relationship between e and f is not considered. Therefore, and are respectively used to calculate the aggregated node representations and . Since the feature types extracted by different aggregation methods are different, concatenating the features they extract can enrich the node features and further improve the performance of the neural network. However, directly concatenating the node representations and calculated by different aggregation methods may add redundant or unimportant features, resulting in insufficient attention to the important features that affect the prediction accuracy. Therefore, the present invention multiplies the aggregated features by weights (regarded as a kind of attention) to scale the feature values. If the aggregated feature representation has an important impact on the output of the neural network, a larger weight is assigned; otherwise, a smaller weight is assigned. Different from the traditional attention mechanism, the present invention uses a global pooling layer and a fully connected layer to determine the attention size, which is expressed as equation (13). It should be noted that there are two fully connected layers for attention calculation, which are respectively used for the attention calculation of node features e and f, and are expressed as:

[0159] (13)

[0160] where, represents the global pooling calculation, and its output dimension is ; and represent two fully connected layers for attention calculation, and their output dimensions are .

[0161] Finally, the node features based on different neighborhood aggregation methods and the forward propagation function of the neural network in the th graph convolutional layer can be expressed as equations (14) and (15). Further combining the pooling layer and the fully connected layer, a complete multi-core collaborative graph convolutional neural network architecture can be constructed.

[0162] (14)

[0163] (15)

[0164] where, represents the concatenation calculation of different features; represents the coupled node features aggregated by different methods.

[0165] S205. Train the multi-core collaborative graph convolutional neural network using the training sample set to obtain a prediction model.

[0166] It can be understood that during the training process of the multi-core collaborative graph convolutional neural network in this application, a training method with a maximum number of iterations can be adopted, or a method with a preset convergence condition can be adopted, which is not limited here.

[0167] S103. Determine the operation reliability evaluation result according to the minimum load shedding amount of the power system.

[0168] It can be understood that the basic index of operation reliability is the adequacy index, such as the Probability of Load Curtailment (PLC), the Expected Demand Not Supplied (EDNS), etc. After obtaining the minimum load shedding amount of the load under different system states in the present invention, PLC and EDNS can be calculated according to the following calculation formulas:

[0169] ;

[0170] ;

[0171] where is the occurrence probability of the i-th system state; S is the set of system states where load shedding occurs; represents the minimum load shedding amount of node j under system state i.

[0172] In the specific implementation process of this application, during the process of determining the operation reliability evaluation result according to the minimum load shedding amount of the power system, it is necessary to determine when to stop the calculation. There are mainly two types of convergence judgments: the maximum number of samples and the variance of the reliability index. In the specific implementation process of this application, the convergence criterion is judged according to both the maximum number of samples and the variance of the reliability index.

[0173] The present invention will be further described below in conjunction with specific implementation examples.

[0174] Example 1:

[0175] In the specific implementation process of this application, the operation reliability evaluation in the power system usually involves load fluctuations, random renewable energy generation, and the outage of important power equipment. For load fluctuations, in this embodiment, it is assumed to follow a normal distribution, with the default value as the mean and 0.3 as the standard deviation. Two types of renewable energy, wind power and photovoltaic power, are considered in this embodiment. It is assumed that the wind speed and solar irradiance follow the Weibull distribution and the Beta distribution respectively, where the wind speed follows the Weibull distribution, where , the solar irradiance follows distribution, where . In this embodiment, multiple wind farms and photovoltaic power stations are connected to different nodes, so that the penetration rate of renewable energy exceeds 20%. In this embodiment, in the IEEE 39-node system, four wind farms with a capacity of 200 MW and four photovoltaic power stations with a capacity of 200 MW are randomly connected to different nodes of the power system. The penetration rate of renewable energy is 20.26%. In order to reduce the imbalance of training data, this embodiment samples N-1 and N-2 scenario samples to cover all training situations. In the operation reliability assessment stage, this embodiment simulates a 1% failure probability for reliability assessment. This embodiment generates 20K and 10K samples for training and testing.

[0176] To prove the effectiveness of the proposed method, this embodiment compares the traditional operation reliability assessment method with 5 data-driven operation reliability assessment methods. These methods and their related descriptions will be introduced below. All neural networks are constructed and trained using the PyTorch framework on a desktop computer with an Intel(R) Core(TM) i7-10700K CPU@ 3.80GHz 3.80GHz, 16GB RAM, and an NVIDIA GeForce RTX 2080Ti. The neural network is trained using the Adam optimizer with a learning rate of 0.001.

[0177] M0: Traditional operation reliability assessment method, which uses the interior point method to solve the minimum load shedding. This method is the benchmark method for data.

[0178] M1: This method uses a typical graph convolutional neural network to calculate the minimum load shedding, and its graph convolutional kernel is derived from the Laplacian matrix according to spectral graph theory.

[0179] M2: This method is exactly the same as M1 except for the graph convolutional kernel. The graph convolutional kernel uses a Gaussian kernel based on line impedance.

[0180] M3: This method is exactly the same as M1, but uses the graph convolutional kernel embedded in the power flow model.

[0181] M4: This method is the same as M1, but uses the graph convolutional kernel accurately embedded in the proposed physical model.

[0182] M5: This method is the same as M1, but uses the multi-core collaborative graph convolutional neural network architecture proposed in the present invention.

[0183] This embodiment uses the mean absolute error and the correlation error to evaluate the performance of different methods, and their mathematical formulas are:

[0184] ;

[0185] ;

[0186] where y is the minimum load shedding amount calculated by using a numerical method to solve the load shedding model; is the minimum load shedding amount predicted by using a neural network; is the number of test data; is the reference value. For the operation reliability index, the accurate value evaluated by M0 is used as the reference value.

[0187] such as Figure 5 shown, is the error curve of the training set during the training of different neural networks. As Figure 6 shown, is the error curve of the test set during the training of different neural networks.

[0188] In this embodiment, neural networks of different methods are first constructed and trained in the IEEE 39-bus system. The settings in M4 and M5 are 0.1. Different neural networks are trained through 2000 iterations. When different iterations are completed, the mean absolute errors of the training data and the test are as Figure 5 and Figure 6 shown. In addition, this embodiment performs K-fold cross-validation and statistically calculates the mean absolute error and standard deviation when K = 10, as shown in Table 1. It can be observed that M1 and M2 converge after several iterations, but the validation error is the largest. In M3, the embedded design of the power flow model can reduce the error to a smaller value. When the neighborhood aggregation method with the embedded power flow model is changed to the neighborhood aggregation method with the accurate embedding of the physical model, the training and test errors can be further reduced. In addition, when using M5 with the multi-core collaborative graph convolutional neural network, the minimum error and standard deviation can be obtained. The effectiveness of the neural network architecture proposed by the present invention is verified.

[0189] Table 1

[0190]

[0191] Embodiment 2:

[0192] In another embodiment of the present application, the operation reliability assessment in a power system usually involves load fluctuations, random renewable energy generation, and the outage of important power equipment. For load fluctuations, this embodiment assumes that it follows a normal distribution, with the default value as the mean and 0.3 as the standard deviation. Two types of renewable energy, wind power and photovoltaic power, are considered in this embodiment. It is assumed that the wind speed and solar irradiance follow a Weibull distribution and a Beta distribution respectively, where the wind speed follows a Weibull distribution, where , and the solar irradiance follows distribution, where In this embodiment, multiple wind farms and photovoltaic power plants are connected to different nodes, enabling the penetration rate of renewable energy to exceed 20%. In this embodiment, in the IEEE 39-node system, four wind farms with a capacity of 200 MW and four photovoltaic power plants with a capacity of 200 MW are randomly connected to different nodes of the power system. The penetration rate of renewable energy is 20.26%. To reduce the imbalance of training data, this embodiment samples N-1 and N-2 scenario samples to cover all training situations. In the operation reliability assessment stage, this embodiment simulates a 1% failure probability for reliability assessment. This embodiment generates 20K and 10K samples for training and testing.

[0193] To prove the effectiveness of the proposed method, this embodiment compares the traditional operation reliability assessment method with five data-driven operation reliability assessment methods. These methods and their related descriptions will be introduced below. All neural networks are constructed and trained using the PyTorch framework on a desktop computer with an Intel(R) Core(TM) i7-10700K CPU@ 3.80GHz 3.80GHz, 16GB RAM, and an NVIDIA GeForce RTX 2080Ti. The neural network is trained using the Adam optimizer with a learning rate of 0.001.

[0194] M0: Traditional operation reliability assessment method, which uses the interior point method to solve the minimum load shedding. This method is the benchmark method for data.

[0195] M1: This method uses a typical graph convolutional neural network to calculate the minimum load shedding, and its graph convolutional kernel is derived from the Laplacian matrix according to spectral graph theory.

[0196] M2: This method is exactly the same as M1 except for the graph convolutional kernel. The graph convolutional kernel uses a Gaussian kernel based on line impedance.

[0197] M3: This method is exactly the same as M1, but uses the graph convolutional kernel embedded in the power flow model.

[0198] M4: This method is the same as M1, but uses the graph convolutional kernel precisely embedded in the proposed physical model.

[0199] M5: This method is the same as M1, but uses the multi-core collaborative graph convolutional neural network architecture proposed in the present invention.

[0200] This embodiment uses the mean absolute error and correlation error to evaluate the performance of different methods, and their mathematical formulas are:

[0201] ;

[0202] ;

[0203] where y is the minimum load shedding amount calculated by using a numerical method to solve the load shedding model; is the minimum load shedding amount predicted by using a neural network; is the number of test data; is the reference value. For the operation reliability index, the accurate value evaluated by M0 is used as the reference value.

[0204] This embodiment further verifies the effectiveness of the proposed multi-core collaborative graph convolutional neural network design. This embodiment implements different methods in the IEEE 39-node system, and statistically analyzes the attention values of different neighborhood aggregation methods under test samples, as Figure 7 and Figure 8 shown (the bar graphs of each layer are successively , , and ). It can be observed that the attention values are different in different multi-core graph convolutional layers. The neighborhood aggregation methods derived from spectral graph theory (such as and ) have larger attention values in the th layer, while the aggregation methods embedded with physical models (such as and ) have larger attention values in the last layer. This is because the input representations of e and f are constants, and and can more effectively extract graph features based on node connections. In deeper hidden layers (such as the 3rd layer), and derived from the power flow model help to extract important features affecting the predicted minimum load shedding amount. Therefore, the multi-core design of the graph convolutional layer can more effectively perform feature extraction. In graph convolutional layers with different depths, different neighborhood aggregation methods play a dominant role, which can help the neural network extract features affecting the output more accurately and effectively, verifying the effectiveness of the proposed multi-core collaborative graph convolutional neural network architecture.

[0205] Embodiment 3:

[0206] Based on Embodiment 2, this embodiment further uses the trained neural network for operation reliability evaluation. The failure probability of important equipment is set to 1%. The maximum number of iterations is 30K. The corresponding reliability indexes and relative errors of M1 - M5 are shown in the table. It can be observed that M1 based on the typical neighborhood aggregation method has the largest evaluation error. When gradually constructing the neighborhood aggregation methods embedded with physical variables (such as line impedance) and physical models, the evaluation error gradually decreases, verifying the effectiveness of the proposed neural network architecture.

[0207] As can be seen from Table 2, when using S1 and S2 for testing, the accuracy of M0 - M3 also decreased to some extent, but the accuracy of M4 and M5 can still reach 95%. It is worth noting that in the 2383 - node system, the accuracy of VG is about 4% - 20%, but the accuracy improvement of M4 and M5 even reached 90%. Therefore, the local and global non - linear sample collection method proposed in the present invention can well improve the generalization ability of the neural network, enabling it to adapt to the testing of non - linear samples, verifying the effectiveness of the method proposed in the present invention.

[0208] Table 2

[0209]

[0210] Example 4:

[0211] This example uses the actual load curve to verify the effectiveness of the proposed neural network architecture in the evaluation during a day. As Figure 9 shown, this example applies the load curve of a certain area from January 1, 2022 to January 2, 2022 to the IEEE 39 - node system. The outage probabilities of lines and generators are set to 0.01. Subsequently, this example uses M1 - M5 for operation reliability evaluation and plots the errors of reliability indicators at different times. All the used graph convolutional neural networks are composed of three graph convolutional layers and three fully - connected layers. Each graph convolutional layer has 16 channels and each fully - connected layer has 256 neurons. Then, the load curve data on January 1 is used to generate training samples, with a total of 48000 samples (2000 samples at each moment). Then, this example conducts operation reliability evaluation at different times on January 2, and the results are as Figure Six shown. As can be seen from the figure, compared with other methods, M5 shows lower calculation errors at each moment. Especially in the time period between 12 o'clock and 17 o'clock, the error rates of other methods increase significantly, while M5 always maintains lower calculation errors within this time range, verifying the effectiveness of the method for operation reliability evaluation of the multi - core collaborative graph convolutional neural network proposed in the present invention.

[0212] From the above - mentioned solution, the present application provides a method for operation reliability evaluation. After obtaining the system state to be evaluated, the system state to be evaluated is input into the prediction model, and the minimum load shedding amount of the power system is output; according to the minimum load shedding amount of the power system, the operation reliability evaluation result is determined. The prediction model of the present application is a multi - core collaborative graph convolutional neural network; at the architecture level, it improves the ability of the neural network to extract node features, topological features and their physical association features, thereby effectively improving the accuracy of the neural network in calculating the minimum load shedding and enhancing the adaptability to complex variable - topology scenarios during operation reliability evaluation.

[0213] Another embodiment of the present application provides an operating reliability evaluation device, as Figure 10 shown, specifically including:

[0214] An acquisition unit 1001, configured to acquire the system state to be evaluated.

[0215] An analysis unit 1002, configured to input the system state to be evaluated into a prediction model, and output the minimum load shedding amount of the power system.

[0216] Among them, the prediction model is a multi-core collaborative graph convolutional neural network.

[0217] An evaluation unit 1003, configured to determine the operating reliability evaluation result according to the minimum load shedding amount of the power system.

[0218] For the specific working process of the units disclosed in the above embodiments of the present application, reference may be made to the corresponding method embodiment content, as Figure 1 shown, which will not be elaborated here.

[0219] Optionally, in another embodiment of the present application, an implementation manner of the prediction model construction unit includes:

[0220] A training sample set construction unit, configured to construct a training sample set.

[0221] A minimum load shedding model establishment unit, configured to establish an initial minimum load shedding model.

[0222] A graph convolutional layer construction unit, configured to construct a graph convolutional layer accurately embedded in the initial minimum load shedding model.

[0223] A graph convolutional neural network establishment unit, configured to establish a multi-core collaborative graph convolutional neural network according to the graph convolutional layer.

[0224] A training unit, configured to train the multi-core collaborative graph convolutional neural network by using the training sample set to obtain a prediction model.

[0225] For the specific working process of the units disclosed in the above embodiments of the present application, reference may be made to the corresponding method embodiment content, as Figure 2 shown, which will not be elaborated here.

[0226] Optionally, in another embodiment of the present application, an implementation manner of the operating reliability evaluation device includes:

[0227] A convergence judgment unit, configured to perform convergence criterion judgment according to the maximum number of samples and the variance of the reliability index during the process of determining the operating reliability evaluation result according to the minimum load shedding amount of the power system.

[0228] For the specific working process of the unit disclosed in the above embodiments of the present application, reference may be made to the corresponding method embodiment content, which will not be elaborated here.

[0229] Optionally, in another embodiment of the present application, an implementation manner of the training sample set construction unit includes:

[0230] A sampling unit, configured to obtain the power system state through Monte Carlo sampling according to the typical operation scenarios of the power system.

[0231] A solving unit, configured to solve the minimum load shedding amount under the power system state through the interior point method for each power system state.

[0232] A training sample set construction subunit, configured to construct a training sample set according to the minimum load shedding amounts under all power system states.

[0233] For the specific working process of the unit disclosed in the above embodiments of the present application, reference may be made to the corresponding method embodiment content, which will not be elaborated here.

[0234] Optionally, in another embodiment of the present application, an implementation manner of the graph convolutional layer construction unit includes:

[0235] A graph convolutional layer construction subunit, configured to embed the Gauss-Seidel iteration of the power flow equation into the graph convolution to construct a graph convolutional layer in which the initial minimum load shedding model is accurately embedded.

[0236] For the specific working process of the unit disclosed in the above embodiments of the present application, reference may be made to the corresponding method embodiment content, which will not be elaborated here.

[0237] Optionally, in another embodiment of the present application, an implementation manner of the graph convolutional neural network establishment unit includes:

[0238] A first determination unit, configured to determine the forward propagation formula according to the node feature convolution and the node feature neighborhood aggregation calculation function.

[0239] A second determination unit, configured to determine the first AGGRE function based on the adjacency matrix.

[0240] A third determination unit, configured to determine the second AGGRE function based on the aggregation weight matrix of the impedance.

[0241] A normalization unit, configured to perform normalization processing on the aggregation weight matrix of the impedance to obtain the aggregation weight matrix of the target impedance.

[0242] A fourth determination unit, configured to determine the third AGGRE function based on the aggregation weight matrix of the target impedance.

[0243] A fifth determination unit, configured to determine a fourth AGGRE function based on the converted power flow equation.

[0244] A dynamic weighted merging unit, configured to perform dynamic weighted merging on the first AGGRE function, the second AGGRE function, the third AGGRE function, and the fourth AGGRE function to obtain a multi-core collaborative graph convolutional neural network.

[0245] For the specific working process of the units disclosed in the above embodiments of the present application, reference may be made to the corresponding method embodiment content, as Figure 3 shown, which will not be elaborated here.

[0246] As can be seen from the above solutions, the present application provides an operation reliability evaluation device. After the acquisition unit 1001 acquires the system state to be evaluated; the analysis unit 1002 inputs the system state to be evaluated into the prediction model and outputs the minimum load shedding amount of the power system; the evaluation unit 1003 determines the operation reliability evaluation result according to the minimum load shedding amount of the power system. The prediction model of the present application is a multi-core collaborative graph convolutional neural network; at the architecture level, it improves the neural network's ability to extract node features, topological features, and their physical association features, thereby effectively improving the accuracy of the neural network's calculation of the minimum load shedding and enhancing the adaptability to complex variable topology scenarios during operation reliability evaluation.

[0247] The functions described above in this article can be at least partially performed by one or more hardware logic components. For example, without limitation, the exemplary types of hardware logic components that can be used include: field programmable gate arrays (FPGA), application specific integrated circuits (ASIC), application specific standard products (ASSP), system on a chip (SOC), complex programmable logic devices (CPLD), and so on.

[0248] Another embodiment of the present application provides an electronic device, as Figure 11 shown, including:

[0249] One or more processors 1101.

[0250] A storage device 1102, on which one or more programs are stored.

[0251] When the one or more programs are executed by the one or more processors 1101, the one or more processors 1101 are caused to implement the operation reliability evaluation method as described in the above embodiments.

[0252] Another embodiment of the present application provides a computer storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the operation reliability evaluation method as described in the above embodiments.

[0253] In the context of the present application, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0254] It should be noted that the computer-readable medium described above in the present application can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples of a computer-readable storage medium can include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present application, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device. And in the present application, a computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal can take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on a computer-readable medium can be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the foregoing.

[0255] The above computer-readable medium can be included in the above electronic device; or it can exist separately and not be assembled into the electronic device.

[0256] Another embodiment of the present application provides a computer program product which, when executed, is used to execute the above-mentioned running reliability evaluation method.

[0257] Specifically, according to the embodiments of the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, an embodiment of the present application includes a computer program product that includes a computer program carried on a non-transitory computer-readable medium, and the computer program includes program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device, or installed from a ROM. When the computer program is executed by a processing device, it executes the above-mentioned functions defined in the method of the embodiments of the present application.

[0258] Although the subject matter has been described in language specific to structural features and / or methodological logical acts, it should be understood that the subject matter defined in this application is not necessarily limited to the specific features or acts described above. On the contrary, the specific features and acts described above are only example forms for implementing this application.

[0259] Although several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this application. Certain features described in the context of separate embodiments can also be implemented in combination in a single embodiment. On the contrary, the various features described in the context of a single embodiment can also be implemented separately or in any suitable sub-combination in multiple embodiments.

[0260] The above description is only a preferred embodiment of this application and an explanation of the technical principles applied. Those skilled in the art should understand that the scope of the application involved in this application is not limited to the technical solutions formed by the specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above application concept. For example, the technical solutions formed by mutually replacing the above features with the technical features (but not limited to) having similar functions applied in this application.

Claims

1. A method for evaluating operation reliability, characterized in that: include: Get the system status to be evaluated; The system state to be evaluated is input into a prediction model, and the minimum load shedding amount of the power system is obtained as an output; wherein the prediction model is a multi-core collaborative graph convolutional neural network; An operation reliability evaluation result is determined according to the minimum load shedding amount of the power system.

2. The operation reliability evaluation method according to claim 1, characterized in that: The method for constructing the prediction model comprises: Construct a training sample set; Establishing the initial minimum load shedding model; Constructing a graph convolutional layer that accurately embeds the initial minimum load shedding model; Establishing a multi-core collaborative graph convolutional neural network according to the graph convolutional layer; The multi-core collaborative graph convolutional neural network is trained using the training sample set to obtain a prediction model.

3. The operation reliability evaluation method according to claim 1, characterized in that: Also includes: In the process of determining the operation reliability evaluation result according to the minimum load shedding amount of the power system, the convergence criterion is judged according to the maximum number of samples and the variance of the reliability index.

4. The operation reliability evaluation method according to claim 2, characterized in that: The step of constructing a training sample set includes: According to the typical operation scenarios of the power system, the power system status is obtained through Monte Carlo sampling; For each power system state, the minimum load shedding amount under the power system state is solved by the interior point method; A training sample set is constructed based on the minimum load shedding amount under all the power system states.

5. The operation reliability evaluation method according to claim 2, characterized in that: The step of constructing a graph convolutional layer that accurately embeds the initial minimum load shedding model includes: The Gauss-Seidel iteration of the power flow equation is embedded into the graph convolution to construct a graph convolution layer in which the initial minimum load shedding model is accurately embedded.

6. The operation reliability evaluation method according to claim 2, characterized in that: The step of establishing a multi-core collaborative graph convolutional neural network according to the graph convolutional layer includes: Determine the forward propagation formula based on the node feature convolution and node feature neighborhood aggregation calculation function; Based on the adjacency matrix, determining a first AGGRE function; Determining a second AGGRE function based on the aggregation weight matrix of the impedance; Standardizing the aggregate weight matrix of the impedance to obtain an aggregate weight matrix of the target impedance; Determining a third AGGRE function based on the aggregate weight matrix of the target impedance; Based on the transformed power flow equation, the fourth AGGRE function is determined; The first AGGRE function, the second AGGRE function, the third AGGRE function and the fourth AGGRE function are dynamically weighted and merged to obtain a multi-core collaborative graph convolutional neural network.

7. An operation reliability assessment device, characterized in that: include: An acquisition unit, used for acquiring a system state to be evaluated; An analysis unit, used for inputting the system state to be evaluated into a prediction model, and outputting a minimum load shedding amount of the power system; wherein the prediction model is a multi-core collaborative graph convolutional neural network; An evaluation unit is used to determine an operation reliability evaluation result according to a minimum load shedding amount of the power system.

8. The operation reliability evaluation device according to claim 7, characterized in that: The construction unit of the prediction model includes: A training sample set construction unit, used for constructing a training sample set; A minimum load shedding model establishing unit, used to establish an initial minimum load shedding model; A graph convolutional layer construction unit, used to construct a graph convolutional layer accurately embedded in the initial minimum load shedding model; A graph convolutional neural network establishing unit, used for establishing a multi-core collaborative graph convolutional neural network according to the graph convolution layer; A training unit is used to train the multi-core collaborative graph convolutional neural network using the training sample set to obtain a prediction model.

9. An electronic device, characterized in that: include: one or more processors; a storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors are enabled to implement the operation reliability assessment method according to any one of claims 1 to 6.

10. A computer storage medium, characterized in that: A computer program is stored thereon, wherein when the computer program is executed by a processor, the operation reliability evaluation method as claimed in any one of claims 1 to 6 is implemented.

Images