Power grid starting scheme checking method, system, equipment, medium and product

Abstract

The invention relates to the technical field of electric power systems, and discloses a power grid starting scheme checking method, system, device, medium and product, a power grid starting scheme is generated through a generative adversarial network based on a double attention mechanism by using a data set formed by current operation data of each node in a power grid, and a preset power grid starting rule is combined to check the power grid starting scheme. The power grid starting scheme is checked, the generative adversarial network based on the double attention mechanism is utilized, the accuracy of the generated power grid starting scheme can be improved, even under the conditions of power grid topology change, equipment state adjustment and the like, rapid adaptation can still be achieved, an accurate checking result can be provided, and therefore the change of a power grid is processed in real time, and the checking efficiency is improved. And the safety and the stability of the power grid are enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power systems, and particularly relates to a power grid starting scheme checking method, system, device, medium and product. BACKGROUND

[0002] At present, the generation and checking of the power grid starting scheme mainly depend on traditional algorithms, and these algorithms have some obvious defects in processing large-scale power grid data: the traditional algorithms usually need a large amount of data for training and optimization, and perform poorly in the case of small data. When the power grid topology is complex or data is missing, the performance of the traditional algorithms will decrease significantly, and it is difficult to handle the case of missing or incomplete data. The existing methods are often difficult to process the changes of the power grid in real time, resulting in low checking efficiency, lack of adaptive ability, and limited data processing and analysis ability. SUMMARY

[0003] Therefore, the present application provides a power grid starting scheme checking method, system, device, medium and product, which solves the technical problem that the existing methods are often difficult to process the changes of the power grid in real time, resulting in low checking efficiency, lack of adaptive ability, and limited data processing and analysis ability.

[0004] The first aspect of the present application provides a power grid starting scheme checking method, comprising:

[0005] obtaining current operation data of each node in the power grid to form a data set;

[0006] generating the power grid starting scheme based on a dual attention mechanism-based generative adversarial network and using the data set;

[0007] obtaining starting operation data of the power grid starting scheme, and checking the power grid starting scheme in combination with a preset power grid starting rule.

[0008] Preferably, the current operation data includes device state data, topology connection data and electrical parameter data.

[0009] The method further comprises:

[0010] performing data missing compensation processing on the current operation data based on a K-nearest neighbor algorithm.

[0011] Preferably, the generative adversarial network includes a generator network and a discriminator network.

[0012] The dual attention mechanism-based generative adversarial network generates the power grid starting scheme using the data set, comprising:

[0013] receiving random noise data through the generator network, performing feature extraction on the random noise data based on a double attention mechanism, and generating an initial power grid startup scheme;

[0014] inputting the data set and the initial power grid startup scheme into the discriminator network, performing authenticity discrimination on the initial power grid startup scheme based on the data set by the discriminator network, and outputting a discrimination result;

[0015] adjusting parameters of the generator network according to the discrimination result until the discrimination result output by the discriminator network reaches a preset convergence condition, and obtaining an optimized power grid startup scheme.

[0016] Preferably, the double attention mechanism includes a spatial attention mechanism and a channel attention mechanism.

[0017] The receiving random noise data through the generator network, performing feature extraction on the random noise data based on a double attention mechanism, and generating an initial power grid startup scheme, comprises:

[0018] The random noise data is converted into sequence data with position information through position encoding, and the sequence data is input into the generator network; wherein the random noise data is a multi-dimensional feature vector.

[0019] The sequence data is subjected to weighted processing of position information based on the spatial attention mechanism, and a first weighted output feature is obtained.

[0020] The sequence data is subjected to feature importance evaluation based on the channel attention mechanism, the importance weight of each channel is obtained, and the sequence data is subjected to weighted processing in the channel dimension according to the importance weight, and a second weighted output feature is obtained.

[0021] The first weighted output feature and the second weighted output feature are weighted and fused to obtain a fused feature vector.

[0022] The initial power grid startup scheme is generated based on the fused feature vector.

[0023] Preferably, the discriminator network performs authenticity discrimination on the initial power grid startup scheme based on the data set, and outputs a discrimination result, which comprises:

[0024] The initial power grid startup scheme is compared with the data set, and the similarity between the initial power grid startup scheme and the data set is calculated.

[0025] If the similarity is higher than a preset similarity threshold, it is determined that the initial power grid startup scheme is true, and a discrimination result indicating true is output.

[0026] If the similarity is lower than or equal to a preset similarity threshold, it is determined that the initial power grid starting scheme is false, and a determination result indicating the false is output.

[0027] Preferably, the starting operation data includes current, voltage, load power and connection state.

[0028] The starting operation data of the power grid starting scheme is acquired, and the power grid starting scheme is checked in combination with a preset power grid starting rule.

[0029] It is determined whether the current is greater than a preset current threshold, and in the case that the current is greater than the preset current threshold, it is determined that the power grid starting scheme is unqualified, and the step of acquiring the current operation data of each node in the power grid to form a data set is re-executed; in the case that the current is not greater than the preset current threshold, the next step is executed.

[0030] According to the connection state of each device, a topological effective number is determined, and it is determined whether the topological effective number is equal to a preset total device number, and in the case that the topological effective number is not equal to the preset total device number, it is determined that the power grid starting scheme is unqualified, and the step of acquiring the current operation data of each node in the power grid to form a data set is re-executed; in the case that the topological effective number is equal to the preset total device number, the next step is executed.

[0031] It is determined whether the power grid is overloaded according to the current, the voltage and the load power, and in the case that the power grid is overloaded, it is determined that the power grid starting scheme is unqualified, and the step of acquiring the current operation data of each node in the power grid to form a data set is re-executed; in the case that the power grid is not overloaded, it is determined that the power grid starting scheme is qualified, and the process is ended.

[0032] In a second aspect, the present application further provides a power grid starting scheme checking system, comprising:

[0033] A data acquisition module is configured to acquire current operation data of each node in a power grid to form a data set.

[0034] A scheme generation module is configured to generate a power grid starting scheme by using a data set based on a dual attention mechanism-based generative adversarial network.

[0035] A scheme checking module is configured to acquire starting operation data of the power grid starting scheme, and check the power grid starting scheme in combination with a preset power grid starting rule.

[0036] In a third aspect, the present application also provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer program is executed by the processor to enable the processor to perform the steps of the power grid startup scheme checking method according to the first aspect.

[0037] In a fourth aspect, the present application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed to implement the steps of the power grid startup scheme checking method according to the first aspect.

[0038] In a fifth aspect, the present application also provides a computer program product, which comprises a computer program stored on a non-transitory computer readable storage medium, and the computer program comprises program instructions, wherein when the program instructions are executed by a computer, the computer is enabled to perform the steps of the power grid startup scheme checking method according to the first aspect.

[0039] From the above technical solutions, it can be seen that, by using the data set composed of the current operation data of each node in the power grid, the power grid startup scheme is generated based on the generative adversarial network based on the double attention mechanism, and the power grid startup scheme is checked in combination with the preset power grid startup rule. The generative adversarial network based on the double attention mechanism can generate the accuracy of the power grid startup scheme. Even in the case of power grid topology changes, device state adjustment, etc., it can still quickly adapt and provide accurate checking results, so as to process the changes of the power grid in real time, improve the checking efficiency, and enhance the safety and stability of the power grid. BRIEF DESCRIPTION OF DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0041] Figure 1 An application environment diagram of the power grid startup scheme checking method provided by the embodiment of the present application is shown in the figure;

[0042] Figure 2 A flowchart of the power grid startup scheme checking method provided by the embodiment of the present application is shown in the figure;

[0043] Figure 3 A structural schematic diagram of the power grid startup scheme checking system provided by the embodiment of the present application is shown in the figure;

[0044] Figure 4 A structural schematic diagram of the electronic device provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0045] In order to make the personnel in the technical field better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only 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 labor are within the protection scope of the present application.

[0046] The power grid starting scheme checking method provided by the embodiments of the present application can be applied to the application environment as shown in Figure 1 . Wherein, the terminal 101 communicates with the server 102 through the network. The data storage system can store the data required to be processed by the server 102. The data storage system can be integrated on the server 102, or placed on the cloud or other network servers. The terminal 101 or the server 102 obtains the current running data of each node in the power grid to form a data set; generates the power grid starting scheme by using the data set based on the dual attention mechanism of the generative adversarial network; obtains the starting running data of the power grid starting scheme, and combines the preset power grid starting rule to check the power grid starting scheme.

[0047] The terminal 101 can be, but is not limited to, various personal computers, notebook computers, smart phones and tablet computers and the like.

[0048] The server 102 can be a stand-alone physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0049] As shown in Figure 2 , the present application provides a power grid starting scheme checking method. Taking the terminal 101 or the server 102 in Figure 1 as an example, the method comprises the following steps S1 to S3. Wherein:

[0050] Step S1, obtaining the current running data of each node in the power grid to form a data set.

[0051] Wherein, the current running data includes device state data, topology connection data and electrical parameter data. The device state data includes the working state (running / stop operation / maintenance) of the power transmission / distribution line and other devices, the state of the circuit breaker / switch; the topology connection data includes the connection information between devices, which clearly defines the power grid topology structure; the electrical parameter data includes the line / device current, the voltage / current / power and load data of each region / device.

[0052] Step S2, a dual attention mechanism-based generative adversarial network generates the power grid startup scheme based on the dataset.

[0053] The dual attention mechanism-based generative adversarial network combines the advantages of dual attention mechanism and generative adversarial network to improve the generation efficiency and accuracy of the power grid startup scheme. The dual attention mechanism enables the model to focus on key information in the input data, whether it is spatial dimension position information or channel dimension feature importance. This mechanism enhances the model's understanding and processing ability of complex power grid data, making the generated power grid startup scheme more in line with the actual operation state and demand of the power grid.

[0054] Through the introduction of the generative adversarial network, the generation quality of the scheme is further improved through the adversarial training between the generator and the discriminator. The discriminator discriminates the authenticity of the generated power grid startup scheme, forcing the generator to continuously optimize its output until the preset convergence condition is reached, thereby ensuring that the generated scheme not only meets the startup rules of the power grid, but also effectively responds to various changes in the power grid.

[0055] The power grid startup scheme includes startup sequence, startup time, startup current, startup connection relationship, startup voltage, and startup load power parameters. These parameters are based on the current operating data of each node in the power grid and are obtained through deep learning and complex calculations of the generative adversarial network. The startup sequence ensures the safety and stability of the power grid equipment during the startup process, avoiding equipment overload or damage; the startup time and startup current, voltage, and other parameters are adjusted according to the actual situation of the power grid to ensure smooth transition and efficient operation during the power grid startup process. In addition, the power grid startup scheme also considers the startup connection relationship, i.e., the topology of the power grid and the connection information between devices, to ensure the connectivity and stability of the power grid during the startup process.

[0056] Step S3, obtain the startup operation data of the power grid startup scheme and combine the preset power grid startup rules to check the power grid startup scheme.

[0057] The preset power grid startup rules are formulated based on the actual operation experience and safety standards of the power grid, including current, voltage, and load power limit conditions, as well as the effectiveness requirements of device connection state. The startup operation data is collected in real time from the power grid startup process, including current, voltage, load power, and connection state key information. By comparing and analyzing these startup operation data with the preset power grid startup rules, it can be determined whether the power grid startup scheme meets the safety operation requirements of the power grid. If there are any problems or the scheme does not meet the rules, the system will give feedback in a timely manner and may trigger the process of regenerating and optimizing the power grid startup scheme to ensure the safe, stable, and efficient operation of the power grid.

[0058] It should be noted that, by means of the generative adversarial network based on the double attention mechanism, the power grid starting scheme is generated by using the data set composed of the current operation data of each node in the power grid, and the power grid starting scheme is checked in combination with the preset power grid starting rule. The generative adversarial network based on the double attention mechanism can generate the accuracy of the power grid starting scheme. Even in the case of power grid topology change, device state adjustment and the like, accurate checking results can still be quickly adapted and provided, so that the changes of the power grid are processed in real time, the checking efficiency is improved, and the safety and stability of the power grid are enhanced.

[0059] In some embodiments, the method further comprises:

[0060] The current operation data is compensated for data loss based on a K nearest neighbor algorithm.

[0061] The K nearest neighbor algorithm can effectively compensate for missing features in the power grid data, ensuring that even when the data quality is incomplete, the checking accuracy can still be maintained.

[0062] Specifically, the principle of the K nearest neighbor algorithm is as follows:

[0063] Low-quality data samples not only have a small number, but also have missing feature values of some parts. If the missing feature values are not processed, the clustering effect will be affected. When a feature value of a certain data is missing, the data similar to it will contain certain hidden information. Therefore, the K nearest neighbor algorithm is used to preprocess the data and compensate for the missing attributes of the data. The K nearest neighbor algorithm measures the data attributes by calculating the similarity of the samples. For sample data a and sample b with complete features, the similarity calculation is as follows:

[0064]

[0065] In the formula, is the similarity distance between sample a and sample b, and s is the number of dimensions of the data sample. is the feature value of sample a in the i-th dimension, is the feature value of sample b in the i-th dimension. If some samples in the sample data set have missing data features, the similarity calculation process of the two samples is as follows:

[0066]

[0067] In the formula, is the distance between sample a and sample b after compensation for missing data, and s is the number of dimensions of the data sample. is the feature value of sample a in the i-th dimension. is the feature value of sample b in the i-th dimension. An information indicator representing sample a and sample b on the i-th feature, usually taking values:

[0068]

[0069] The current running data is compensated for data missing by the K-neighbor algorithm, and after the processing is completed, a data set with complete data characteristics is obtained.

[0070] In some embodiments, the generative adversarial network includes a generator network and a discriminator network;

[0071] The generative adversarial network based on the double attention mechanism generates a power grid startup scheme using a data set, including:

[0072] Step S201, receiving random noise data through the generator network, and extracting features of the random noise data based on the double attention mechanism to generate an initial power grid startup scheme.

[0073] The random noise data is generated by a specific distribution (e.g., Gaussian distribution or uniform distribution) to provide diversity for the generator network. The generator network uses the double attention mechanism to extract features of the random noise data to generate an initial power grid startup scheme. The double attention mechanism enables the generator network to focus on key information in the input data, whether it is spatial dimension position information or channel dimension feature importance, and is fully considered to improve the accuracy and efficiency of generating the power grid startup scheme.

[0074] Specifically, the double attention mechanism includes a spatial attention mechanism and a channel attention mechanism.

[0075] The random noise data is received through the generator network, and the random noise data is extracted based on the double attention mechanism to generate an initial power grid startup scheme, including:

[0076] Step S2011, converting the random noise data into sequence data with position information through position encoding, and inputting the sequence data into the generator network; wherein the random noise data is a multi-dimensional feature vector.

[0077] Step S2012, performing weighted processing on the sequence data based on the spatial attention mechanism to obtain first weighted output features.

[0078] The spatial attention mechanism is used to make the network pay more attention to certain feature areas, and can convert input information into key coefficients. Using the self-attention mechanism, a spatial attention network is constructed according to the features between different data as follows:

[0079]

[0080] where M represents the final attention matrix or weight matrix, which is usually used to weight the feature map or feature channel. It determines the weight of each feature or position in the subsequent processing. X is the input feature, usually a tensor, representing the output of a certain layer in the neural network. It can be a feature map of an image or other forms of features. The Softmax function is used to convert the output of to a probability distribution, ensuring that the sum of the weights of the output is 1. a(·) and β(·) are convolution layers with a convolution size of 1x1. After the inner product calculation, the two-dimensional feature data is obtained by processing with the Softmax function. The final output result is:

[0081]

[0082] where, represents the weighted output feature map, i.e., the first weighted output feature, is an activation function, and M is an attention weight matrix, which is a transformation operation on the input feature map X.

[0083] Step S2013, based on the channel attention mechanism, the importance of the sequence data is evaluated, the importance weight of each channel is obtained, and the sequence data is weighted in the channel dimension according to the importance weight, to obtain the second weighted output feature.

[0084] wherein, in the channel attention mechanism, after the feature data is globally averaged and maximally pooled, two 1x1 convolution layers are obtained, and then the two convolution layers are added. The channel weight is calculated using the Softmax layer, and the sequence data is weighted in the channel dimension using the channel weight, and the second weighted output feature is obtained.

[0085] Step S2014, the first weighted output feature and the second weighted output feature are weighted and fused to obtain a fused feature vector.

[0086] wherein, by setting different weights, the first weighted output feature and the second weighted output feature are weighted and fused to obtain a fused feature vector.

[0087] Step S2015, based on the fused feature vector, an initial power grid startup scheme is generated.

[0088] wherein, the fused feature vector is input into a fully connected layer, and through the processing of the fully connected layer, the relevant parameters of the power grid startup scheme are obtained, including the startup order, the startup time, the startup current, the startup connection relationship, the startup voltage and the startup load power, etc. These parameters jointly constitute the initial power grid startup scheme.

[0089] Step S202, input the data set and the initial power grid startup scheme to the discriminator network, the discriminator network judges the authenticity of the initial power grid startup scheme based on the data set, and outputs the judgment result.

[0090] The discriminator network receives the data set and the initial power grid startup scheme as input, and uses the learned discrimination rule to judge the authenticity of the initial power grid startup scheme. The discriminator network internally analyzes and compares the input data through a complex neural network structure to determine whether the initial power grid startup scheme matches the actual situation of the power grid represented by the data set. If the discriminator network considers that the initial power grid startup scheme has a high matching degree with the data set, i.e., it meets the startup rules and actual needs of the power grid, it outputs a "true" judgment result; otherwise, if the matching degree is low, it outputs a "false" judgment result. Through the authenticity judgment of the discriminator network, high-quality power grid startup schemes can be further screened out.

[0091] Specifically, the discriminator network judges the authenticity of the initial power grid startup scheme based on the data set, and outputs the judgment result, including:

[0092] Step S2021, compare the initial power grid startup scheme with the data set, and calculate the similarity between the initial power grid startup scheme and the data set.

[0093] The similarity calculation is performed by a similarity measurement method (such as cosine similarity, Euclidean distance), which compares the initial power grid startup scheme with the actual power grid operation data in the data set. This step aims to quantify the closeness of the initial scheme to the actual situation of the power grid, thereby providing a basis for subsequent authenticity judgment. Similarity calculation may involve multiple indicators, such as the matching degree of current, voltage, load power, and the compliance of device connection state, etc.

[0094] Step S2022, if the similarity is higher than the preset similarity threshold, the initial power grid startup scheme is determined to be true, and a judgment result indicating true is output;

[0095] Step S2023, if the similarity is lower than or equal to the preset similarity threshold, the initial power grid startup scheme is determined to be false, and a judgment result indicating false is output.

[0096] Step S203, adjust the parameters of the generator network according to the judgment result until the judgment result output by the discriminator network reaches the preset convergence condition, and obtain the optimized power grid startup scheme.

[0097] In one example, a generative adversarial network (GAN) includes a generator network G and a discriminator network D, which play a small and large game in the network, and the game function is:

[0098]

[0099] In the formula, is the objective function of the GAN, which represents the game process between the generator network G and the discriminator network D. The generator network G is responsible for generating pseudo samples (i.e. fake data samples). The discriminator network D is responsible for judging whether the sample is real (from the real data distribution) or generated (from the generator). z is a random vector from a noise distribution, input to the generator. is the input data of the generator network, and g(z) is the generated fake sample, is the real sample data, and D(x) is the judgment of the sample data by the discriminator network. It can be seen that when the generator network generates data with the same characteristics as the sample data, the GAN will generate an optimal solution and reach a balanced state. The solving process of the GAN optimal solution is usually calculated using divergence, and the KL divergence is calculated as:

[0100]

[0101] In the formula, is the real probability distribution (or target distribution), is the approximate probability distribution (or predicted distribution). This integral is a weighted average of all possible solutions. Therefore, the KL divergence is a weighted sum (or integral) of the logarithmic difference of each point on the real distribution P(x).

[0102] Although the GAN model has a small amount of data requirement, it has obvious disadvantages, that is, when the training performance of the discriminator is too good, the generator gradient will collapse, and the optimization of the generator also has many difficulties, thus causing a significant impact on the training efficiency. To improve this problem, the Wasserstein distance is used as the function for solving, and the definition of the distance is:

[0103]

[0104] In the formula, K is a constant, usually related to the regularization of the model, used to control the scaling of the distance. and Pdata are the real data distribution and the generated data distribution, respectively. is the Wasserstein distance, which represents the distance between the real data distribution and the generated data distribution. is a 1-Lipschitz function (i.e., a function that satisfies a certain smoothness constraint), which is a function mapping of the data samples. This function is used to measure the difference between the generated samples and the real samples, and the optimization process aims to maximize this difference. is the Lipschitz constant of the function f (indicating the "smoothness" of the function), which limits the rate of change of f(x). This constraint is a key part of the Wasserstein distance formula, ensuring that the generator remains stable during training. , and are the expected values on real data and generated data, respectively, representing the weighted average of all samples.

[0105] This function has better smoothness properties, so using the Wasserstein distance function instead of the KL divergence function can solve the gradient problem, and the richness of the generated sample features will also be significantly improved. In this way, the WGAN model can provide reliable indicators while also addressing the problem of poor training stability. However, the forced pruning of weights by the Wasserstein distance function can also cause gradient ascent or gradient explosion problems, so a gradient penalty parameter is added to the WGAN model to make the model meet the lip condition and correct the gradient problem. The core function of the WGAN-GP model is:

[0106]

[0107] where, is the objective function of the generative adversarial network (GAN), representing the game process between the generator G and the discriminator D. represents the objective function that the discriminator D needs to maximize in order to distinguish between real samples and generated samples. This is the expectation of the output value of the real data x in the discriminator D. This is the expectation of the discriminator's output value for the generated data g(z). The generator G generates fake samples g(z) through noise z, and the discriminator's goal is to judge it as "fake" (close to 0). By minimizing this expectation, the generator will strive to generate more and more realistic samples. is the gradient penalty term, which adds a regularization term to punish the discriminator's gradient for not meeting the Lipschitz continuity condition (i.e., the discriminator's gradient exceeds 1). This part helps prevent the problem of gradient vanishing or gradient explosion, thereby stabilizing the training process. λ is the weight of the gradient penalty term.

[0108] Compared with the WGAN function, the loss function of the WGAN-GP model adds a regularization term, which is proportional to the GP parameter. Since the parameter of the penalty term is valid for the data x, the penalty term will shorten the optimization area of pz and pdata, thereby reducing the computational load.

[0109] It can be understood that the generator network G in the generative adversarial network usually receives random noise (such as a z vector) as input and generates fake data. By introducing the double attention mechanism, the input of the generator network or the intermediate features generated by the generator network can be weighted, so that the generator network can pay more attention to key features (such as the status of important equipment, the connection of key nodes, etc.) when generating the power grid startup scheme. After combining the double attention mechanism, the objective function of the GAN will be optimized, so that the generator pays more attention to generating data in key areas, and the discriminator network will also become more effective and can identify and punish data lacking key features. Through such optimization, the generator network and the discriminator network can find more accurate patterns in more complex power grid startup scheme data, ensuring that the generated power grid startup scheme is more realistic.

[0110] In some embodiments, the startup operation data includes current, voltage, load power and connection state; in this case, the startup operation data of the power grid startup scheme is obtained, and the power grid startup scheme is checked in combination with the preset power grid startup rule, including:

[0111] Step S301, determine whether the current is greater than the preset current threshold value, if the current is greater than the preset current threshold value, it is determined that the power grid startup scheme is unqualified, and the step of obtaining the current operation data of each node in the power grid to form a data set is re-executed; if the current is not greater than the preset current threshold value, the next step is executed.

[0112] Wherein, the preset current threshold value is the maximum current of the equipment, that is:

[0113]

[0114] In the formula, V max is the maximum value of the voltage, and R is the resistance.

[0115] Step S302, according to the connection state of each device, determine the effective number of topology, and determine whether the effective number of topology is equal to the preset total number of devices, if the effective number of topology is not equal to the preset total number of devices, it is determined that the power grid startup scheme is unqualified, and the step of obtaining the current operation data of each node in the power grid to form a data set is re-executed; if the effective number of topology is equal to the preset total number of devices, the next step is executed.

[0116] Wherein, when checking the topology state of the power grid, the topology connectivity can be used to determine that there is no disconnected or mismatched device connection in the starting process:

[0117]

[0118] In the formula, T valid is the topology validity, C i is the state of the i-th device or connection (1 if valid, otherwise 0), and n is the total number of devices or connections. If T valid is not equal to the total number of devices, it indicates that the topology state is not matched, and the starting scheme is unqualified, and T valid is determined by evaluating the connectivity of the topology of the power grid.

[0119] In step S303, it is determined whether the power grid is overloaded according to the current, voltage and load power. If the power grid is determined to be overloaded, it is determined that the starting scheme of the power grid is unqualified, and the steps of obtaining the current running data of each node in the power grid and constructing a data set are re-executed. If the power grid is determined not to be overloaded, it is determined that the starting scheme of the power grid is qualified, and the process is ended.

[0120] Wherein, for risk assessment in the starting of the power grid, the risk value r can be calculated by comprehensively considering the current, voltage and load power parameters, and a weighted average method is used:

[0121]

[0122] In the formula, V actual and P load are the voltage and load power respectively, V max and P max are the maximum voltage and maximum load power respectively, and w1, w2 and w3 are weight factors, indicating the importance of each parameter in the risk calculation. If r>1, it indicates that there is an overload risk in the starting scheme, which needs to be corrected.

[0123] Based on the same inventive concept, the embodiments of the present application also provide a power grid starting scheme checking system for implementing the power grid starting scheme checking method described above.

[0124] The implementation scheme provided by the system for solving the problem is similar to the implementation scheme described in the above method, so the specific limitations in one or more power grid starting scheme checking system embodiments provided below can be referred to the limitations of the power grid starting scheme checking method in the above, which will not be repeated here.

[0125] As Figure 3 shown, the embodiments of the present application provide a power grid starting scheme checking system, which comprises:

[0126] The data acquisition module 100 is configured to acquire current operation data of each node in the power grid to form a data set.

[0127] The scheme generation module 200 is configured to generate a power grid starting scheme based on a dual attention mechanism-based generative adversarial network and the data set.

[0128] The scheme checking module 300 is configured to acquire starting operation data of the power grid starting scheme and check the power grid starting scheme in combination with preset power grid starting rules.

[0129] In some embodiments, the current operation data includes device state data, topology connection data and electrical parameter data.

[0130] The system further includes a preprocessing module configured to:

[0131] The current operation data is compensated for data loss based on a K-nearest neighbor algorithm.

[0132] In some embodiments, the generative adversarial network includes a generator network and a discriminator network.

[0133] The scheme generation module 200 includes:

[0134] The generator module is configured to receive random noise data through the generator network, extract features of the random noise data based on the dual attention mechanism, and generate an initial power grid starting scheme.

[0135] The discriminator module is configured to input the data set and the initial power grid starting scheme into the discriminator network, perform authenticity discrimination on the initial power grid starting scheme based on the data set, output a discrimination result, adjust parameters of the generator network according to the discrimination result until the discrimination result output by the discriminator network reaches a preset convergence condition, and obtain an optimized power grid starting scheme.

[0136] In some embodiments, the dual attention mechanism includes a spatial attention mechanism and a channel attention mechanism.

[0137] The generator module is configured to:

[0138] The random noise data is converted into sequence data with position information through position encoding, and the sequence data is input into the generator network; wherein the random noise data is a multi-dimensional feature vector.

[0139] The sequence data is weighted based on the spatial attention mechanism to obtain first weighted output features.

[0140] The sequence data is evaluated for feature importance based on a channel attention mechanism to obtain importance weights of each channel, and the sequence data is weighted in the channel dimension according to the importance weights to obtain second weighted output features;

[0141] The first weighted output features and the second weighted output features are weighted and fused to obtain a fused feature vector;

[0142] An initial power grid startup scheme is generated based on the fused feature vector.

[0143] In some embodiments, a discriminator module is configured to:

[0144] The initial power grid startup scheme is compared with the data set to calculate the similarity between the initial power grid startup scheme and the data set;

[0145] If the similarity is higher than a preset similarity threshold, the initial power grid startup scheme is determined to be true, and a true discrimination result is output;

[0146] If the similarity is lower than or equal to the preset similarity threshold, the initial power grid startup scheme is determined to be false, and a false discrimination result is output.

[0147] In some embodiments, the startup operation data includes current, voltage, load power, and connection state; and the discriminator module is configured to:

[0148] It is determined whether the current is greater than a preset current threshold, and if the current is greater than the preset current threshold, it is determined that the power grid startup scheme is unqualified, and the steps of obtaining the current operation data of each node in the power grid to form the data set are re-executed; if the current is not greater than the preset current threshold, the next step is performed;

[0149] The topology effective number is determined according to the connection state of each device, and it is determined whether the topology effective number is equal to a preset total device number, and if it is determined that the topology effective number is not equal to the preset total device number, it is determined that the power grid startup scheme is unqualified, and the steps of obtaining the current operation data of each node in the power grid to form the data set are re-executed; if it is determined that the topology effective number is equal to the preset total device number, the next step is performed;

[0150] It is determined whether the power grid is overloaded according to the current, voltage, and load power, and if it is determined that the power grid is overloaded, it is determined that the power grid startup scheme is unqualified, and the steps of obtaining the current operation data of each node in the power grid to form the data set are re-executed; if it is determined that the power grid is not overloaded, it is determined that the power grid startup scheme is qualified, and the process is ended.

[0151] As Figure 4As shown, the embodiment of the present application provides an electronic device, the electronic device 10 includes a memory 20 and a processor 30, the memory 20 stores a computer program, the computer program is executed by the processor 30, so that the processor 30 executes the steps of the power grid starting scheme checking method in the above embodiment.

[0152] The embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed to realize the steps of the power grid starting scheme checking method in the above embodiment.

[0153] The embodiment of the present application provides a computer program product, the computer program product includes a computer program stored on a non-transitory computer readable storage medium, the computer program includes program instructions, wherein when the program instructions are executed by a computer, the computer executes the steps of the power grid starting scheme checking method in the above embodiment.

[0154] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working process of the above-mentioned system, electronic device, computer storage medium and computer program product can refer to the corresponding process in the foregoing method embodiment, which will not be repeated here.

[0155] It should be noted that the terms "include" and "have" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0156] It should be understood that, although each step in the flowchart involved in each of the above embodiments is displayed in sequence according to the arrow, these steps are not necessarily executed in sequence according to the arrow. Unless otherwise stated herein, the execution of these steps has no strict sequence limitation, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in each of the above embodiments can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least part of other steps or steps or stages in other steps.

[0157] In several embodiments provided by the present application, it should be understood that the disclosed system, electronic device, computer storage medium, computer program product and method can be implemented in other manners. For example, the described device embodiments are merely illustrative, and the division of units can be different. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed coupling or direct coupling or communication connection between units can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or in other forms.

[0158] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place or distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0159] In addition, the functional units in each embodiment of the present application can be integrated into a processing unit, or each unit can be physically present separately, or two or more units can be integrated into one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0160] The integrated unit, if realized in the form of a software functional unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application, essentially or the part that contributes to the prior art, or all or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a number of instructions for executing all or part of the steps of the method described in the embodiments of the present application by a computer device (which can be a personal computer, a server, or a network device, etc.). The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (English full name: Read-Only Memory, English abbreviation: ROM), a random access memory (English full name: Random Access Memory, English abbreviation: RAM), a magnetic disk or an optical disk, and various program code storage media.

[0161] The above examples are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that the technical solutions recorded in the foregoing examples can be modified, or some technical features can be replaced by equivalent features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for verifying a power grid startup scheme, characterized in that, include: Obtain the current operating data of each node in the power grid to form a dataset; A generative adversarial network based on a dual attention mechanism is used to generate a power grid startup scheme using the dataset. The startup and operation data of the power grid startup scheme are obtained, and the power grid startup scheme is verified in combination with the preset power grid startup rules.

2. The power grid startup scheme verification method according to claim 1, characterized in that, The current operating data includes device status data, topology connection data, and electrical parameter data; The method further includes: The K-nearest neighbor algorithm is used to perform data missing compensation processing on the current running data.

3. The power grid startup scheme verification method according to claim 1, characterized in that, The generative adversarial network includes a generator network and a discriminator network; The generative adversarial network based on the dual attention mechanism uses the dataset to generate a power grid startup scheme, including: The generator network receives random noise data, extracts features from the random noise data based on a dual attention mechanism, and generates an initial power grid startup scheme. The dataset and the initial power grid startup scheme are input into the discriminator network, which performs a true / false judgment on the initial power grid startup scheme based on the dataset and outputs the judgment result. The parameters of the generator network are adjusted according to the discrimination result until the discrimination result output by the discriminator network reaches the preset convergence condition, thereby obtaining the optimized power grid startup scheme.

4. The power grid startup scheme verification method according to claim 3, characterized in that, The dual attention mechanism includes spatial attention mechanism and channel attention mechanism; The step of receiving random noise data through the generator network, extracting features from the random noise data based on a dual attention mechanism, and generating an initial power grid startup scheme includes: The random noise data is converted into sequence data with location information through position encoding, and the sequence data is input into the generator network; wherein, the random noise data is a multi-dimensional feature vector; Based on the spatial attention mechanism, the sequence data is weighted by positional information to obtain the first weighted output feature. The sequence data is evaluated for feature importance based on the channel attention mechanism, the importance weight of each channel is obtained, and the sequence data is weighted according to the channel dimension based on the importance weight to obtain the second weighted output feature. The first weighted output feature and the second weighted output feature are weighted and fused to obtain the fused feature vector; The initial power grid startup scheme is generated based on the fused feature vector.

5. The power grid startup scheme verification method according to claim 3, characterized in that, The discriminator network performs a authenticity check on the initial power grid startup scheme based on the dataset, and outputs the discrimination result, including: The initial power grid startup plan is compared with the dataset, and the similarity between the initial power grid startup plan and the dataset is calculated. If the similarity is higher than a preset similarity threshold, the initial power grid startup scheme is determined to be true, and a judgment result indicating that it is true is output. If the similarity is lower than or equal to a preset similarity threshold, the initial power grid startup scheme is determined to be false, and a judgment result indicating falseness is output.

6. The power grid startup scheme verification method according to claim 1, characterized in that, The startup and operation data includes current, voltage, load power, and connection status; The process of acquiring the startup and operation data of the power grid startup scheme and verifying the power grid startup scheme in conjunction with preset power grid startup rules includes: If the current is greater than a preset current threshold, the power grid startup scheme is deemed unqualified, and the step of acquiring the current operating data of each node in the power grid and constructing a dataset is re-executed; if the current is not greater than the preset current threshold, the next step is executed. Based on the connection status of each device, determine the effective number of topologies and determine whether the effective number of topologies is equal to the preset total number of devices. If the effective number of topologies is not equal to the preset total number of devices, the power grid startup scheme is deemed unqualified, and the step of obtaining the current operating data of each node in the power grid to form a dataset is re-executed; if the effective number of topologies is equal to the preset total number of devices, proceed to the next step. The system determines whether the power grid is overloaded based on the current, voltage, and load power. If the power grid is overloaded, the system determines that the power grid startup plan is unqualified and the step of obtaining the current operating data of each node in the power grid to form a dataset is re-executed. If the power grid is not overloaded, the system determines that the power grid startup plan is qualified and the process ends.

7. A power grid startup scheme verification system, characterized in that, include: The data acquisition module is used to acquire the current operating data of each node in the power grid to form a dataset; The scheme generation module is used to generate power grid startup schemes using the dataset based on a generative adversarial network with a dual attention mechanism. The scheme verification module is used to acquire the startup and operation data of the power grid startup scheme and verify the power grid startup scheme in combination with the preset power grid startup rules.

8. An electronic device, characterized in that, The electronic device includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, causes the processor to perform the steps of the power grid startup scheme verification method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it implements the steps of the power grid startup scheme verification method as described in any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, wherein when the program instructions are executed by a computer, the computer performs the steps of the power grid startup scheme verification method as described in any one of claims 1-6.

Images