Power system transient stability evaluation method and device based on priority strategy

Abstract

The application relates to the technical field of power system evaluation, and discloses a power system transient stability evaluation method and device based on a priority strategy, which comprises the following steps: forming a first evaluation sequence according to multiple operation sample data of a power system; updating the first evaluation sequence according to the numerical comparison relationship of information entropy indexes corresponding to the multiple operation sample data to form a second evaluation sequence; using a pre-trained neural network cascade model to predict the transient stability category of the second evaluation sequence; and using the prediction result to optimize the pre-trained neural network cascade model; wherein the pre-trained neural network cascade model is used for inputting operation sample data and outputting the prediction result of the transient stability category corresponding to the operation sample data, so that the transient stability of large-scale power operation data can be efficiently evaluated.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system evaluation, in particular to a power system transient stability evaluation method and device based on a priority strategy. BACKGROUND

[0002] With the continuous development and complication of power systems, transient stability evaluation of power systems has become a key link to ensure safe and stable operation of power grids. Traditional transient stability evaluation methods often have problems such as large amount of calculation, long evaluation period, and poor adaptability to time-varying environment.

[0003] In recent years, deep learning technology has gradually increased in application in power systems, providing a new idea for transient stability evaluation. However, when facing large-scale power operation data, it is difficult to efficiently evaluate the transient stability of large-scale power operation data. SUMMARY

[0004] Therefore, the present application provides a power system transient stability evaluation method and device based on a priority strategy, which solves the technical problem that it is difficult to efficiently evaluate the transient stability of large-scale power operation data when facing large-scale power operation data.

[0005] The first aspect of the present application provides a power system transient stability evaluation method based on a priority strategy, which comprises:

[0006] forming a first evaluation sequence according to a plurality of operation sample data of a power system;

[0007] updating the first evaluation sequence according to a numerical comparison relationship of information entropy indexes corresponding to the plurality of operation sample data to form a second evaluation sequence;

[0008] using a pre-trained neural network cascade model to predict the transient stability category of the second evaluation sequence, and using the prediction result to optimize the pre-trained neural network cascade model; wherein the pre-trained neural network cascade model is used to input the operation sample data and output the prediction result of the transient stability category corresponding to the operation sample data.

[0009] Preferably, before the step of updating the first evaluation sequence according to the numerical comparison relationship of information entropy indexes corresponding to the plurality of operation sample data to form a second evaluation sequence, the step of calculating information entropy indexes corresponding to the plurality of operation sample data is further included.

[0010] The step of calculating information entropy indexes corresponding to the plurality of operation sample data comprises:

[0011] inputting the operation sample data into the pre-trained neural network model, and outputting a prediction probability of each transient stability class corresponding to the operation sample data;

[0012] performing weighted summation calculation on the prediction probability of each transient stability class corresponding to the operation sample data to obtain an information entropy index corresponding to the operation sample data.

[0013] Preferably, the step of updating the first evaluation sequence according to the numerical comparison relationship of the information entropy indexes corresponding to the plurality of operation sample data to form a second evaluation sequence comprises: comparing the information entropy indexes corresponding to the plurality of operation sample data with an information entropy threshold, and screening out the plurality of operation sample data with the information entropy indexes greater than the information entropy threshold as key operation sample data.

[0014] arranging the plurality of key operation sample data in descending order according to the numerical comparison relationship of the information entropy indexes to obtain a second evaluation sequence.

[0015] Preferably, the step of performing transient stability class prediction on the second evaluation sequence by using the pre-trained neural network cascade model and optimizing the pre-trained neural network cascade model by using the prediction result comprises:

[0016] After the network parameter optimization of the pre-trained neural network model, the information entropy index corresponding to each operation sample data is recalculated based on the pre-trained neural network model after the network parameter optimization;

[0017] The second evaluation sequence is updated by using the numerical comparison relationship of the recalculated information entropy indexes corresponding to the operation sample data.

[0018] Preferably, the pre-trained neural network cascade model comprises a plurality of serially connected neural network models.

[0019] The step of performing transient stability class prediction on the second evaluation sequence by using the pre-trained neural network cascade model and optimizing the pre-trained neural network cascade model by using the prediction result comprises:

[0020] inputting the second evaluation sequence into a first neural network model in the plurality of pre-trained neural network cascade models for prediction to obtain a transient stability class prediction result corresponding to the second evaluation sequence;

[0021] determine the confidence of the second evaluation sequence by using the transient stability class prediction result corresponding to the second evaluation sequence and the transient stability class simulation result;

[0022] When the confidence of the second evaluation sequence does not reach the preset confidence threshold, the multiple running sample data after expanding the current running simulation time window are updated into the second evaluation sequence;

[0023] The updated second evaluation sequence is input into the second neural network model adjacent to the first neural network model for prediction, so as to obtain the transient stability class prediction result corresponding to the updated second evaluation sequence;

[0024] The confidence of the updated second evaluation sequence is determined by using the transient stability class prediction result corresponding to the updated second evaluation sequence and the transient stability class simulation result;

[0025] When the confidence of the updated second evaluation sequence does not reach the preset confidence threshold, the multiple running sample data after expanding the current running simulation time window are continuously updated into the second evaluation sequence, and iteration is performed in this way until the confidence of the second evaluation sequence reaches the preset confidence threshold or reaches the preset maximum running time window, the iteration stops, and the neural network cascade model after the iteration stops is output.

[0026] Preferably, the step of updating the second evaluation sequence by using the multiple running sample data after expanding the current running simulation time window comprises:

[0027] The multiple running sample data after expanding the current running simulation time window are obtained.

[0028] The second evaluation sequence is updated according to the value comparison relationship of the information entropy indexes respectively corresponding to the obtained multiple running sample data after expanding the current running simulation time window.

[0029] Preferably, after the neural network cascade model stops iteration, the network parameter optimization of the neural network models in the neural network cascade model is performed by using the input quantity and the output result of the last iteration of the neural network cascade model.

[0030] In a second aspect, the present application further provides a power system transient stability evaluation device based on a priority strategy, comprising:

[0031] A sequence forming module is configured to form a first evaluation sequence according to multiple running sample data of a power system.

[0032] A sequence updating module is configured to update the first evaluation sequence according to the value comparison relationship of information entropy indexes respectively corresponding to the multiple running sample data, so as to form a second evaluation sequence.

[0033] a model evaluation module, configured to perform transient stability class prediction on the second evaluation sequence by using a pre-trained neural network cascade model, and optimize the pre-trained neural network cascade model by using a prediction result; wherein the pre-trained neural network cascade model is configured to input the operation sample data and output the prediction result of the transient stability class corresponding to the operation sample data.

[0034] In a third aspect, the present application further provides an electronic device, which comprises a memory and a processor, the memory stores a computer program, and the computer program is executed by the processor to make the processor execute the steps of the power system transient stability evaluation method based on the priority strategy.

[0035] In a fourth aspect, the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed to implement the steps of the power system transient stability evaluation method based on the priority strategy.

[0036] From the above technical solutions, it can be seen that the numerical comparison relationship of the information entropy indexes corresponding to the plurality of operation sample data is used to update the sorting of the operation sample data, so that the operation sample data with higher complexity is identified and preferentially included in the evaluation sequence, ensuring that the operation sample data with more significant model performance improvement is preferentially processed, and then the pre-trained neural network cascade model is used to perform transient stability class prediction on the updated evaluation sequence, and the pre-trained neural network cascade model is optimized by using the prediction result, thereby improving the accuracy and efficiency of model evaluation, enhancing the adaptability of the model to the time-varying operation mode and topology of the power system, and efficiently performing transient stability evaluation on large-scale power operation data. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 An application environment schematic diagram of the power simulation system provided by the embodiment of the present application;

[0038] Figure 2 A flowchart of the power system transient stability evaluation method based on the priority strategy provided by the embodiment of the present application;

[0039] Figure 3 A flowchart of updating the first evaluation sequence provided by the embodiment of the present application;

[0040] Figure 4 A neural network cascade model evaluation principle schematic diagram provided by the embodiment of the present application;

[0041] Figure 5A batch transient stability evaluation principle schematic diagram of a neural network cascade model provided by the embodiment of the present application is shown in the figure.

[0042] Figure 6 A structure schematic diagram of a power system transient stability evaluation device based on a priority strategy provided by the embodiment of the present application is shown in the figure.

[0043] Figure 7 A structure schematic diagram of an electronic device provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0044] In order for those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0045] The power system transient stability evaluation method based on the priority strategy provided by the embodiment of the present application can be applied to the application environment of the power simulation system as shown in the figure. Figure 1 The simulator 101 is in communication connection with the server 102, and the data storage system 103 can store the data required to be processed by the server 102. The simulator 101 can simulate the operation data of the power equipment, such as the generator power angle time series data. The server 102 obtains the operation data simulated by the simulator 101, and executes the power system transient stability evaluation method based on the priority strategy by using the simulated operation data. The data storage system 103 can be integrated on the server 102, or placed on the cloud or other network servers. The server 102 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0046] As shown in the figure, Figure 2 The power system transient stability evaluation method based on the priority strategy provided by the embodiment of the present application is executed in the server 102 as shown in the figure, Figure 1 and includes steps S1-S3. Among them:

[0047] Step S1, forming a first evaluation sequence according to a plurality of operation sample data of the power system.

[0048] Among them, the operation sample data can be the generator power angle time series data of the power system in the current simulation time window, and the plurality of operation sample data can form the first evaluation sequence according to the time sequence relationship.

[0049] Step S2, updating the first evaluation sequence according to the numerical comparison relationship of the information entropy indexes corresponding to the plurality of operation sample data respectively to form a second evaluation sequence.

[0050] The information entropy index can quantify the information amount and critical stability degree contained in each operation sample data, and the higher the information entropy index of the operation sample data, the higher the information amount and critical stability degree contained in the operation sample data, and the operation sample data should be arranged in the front. Therefore, the first evaluation sequence is updated according to the numerical comparison relationship of the information entropy indexes corresponding to the plurality of operation sample data respectively to form a second evaluation sequence sorted according to the information entropy indexes, so that the operation sample data with higher complexity identified is preferentially included in the evaluation sequence, and the operation sample data with more significant model performance improvement is preferentially processed.

[0051] Step S3, using the pre-trained neural network cascade model to predict the transient stability class of the second evaluation sequence, and using the prediction result to optimize the pre-trained neural network cascade model; wherein the pre-trained neural network cascade model is used to input the operation sample data and output the prediction result of the transient stability class corresponding to the operation sample data.

[0052] The pre-trained neural network cascade model is a pre-trained neural network cascade model, the input of which is the operation sample data labeled with a label, and the output is the label of the transient stability class corresponding to the operation sample data. The transient stability class includes transient stability, critical stability and instability. The last layer structure of the neural network cascade model is a softmax layer, and the final output is the probability distribution of each transient stability class. The neural network cascade model selects the maximum item in the prediction probability of the three transient stability classes as the final prediction result of the transient stability class of the power system.

[0053] It should be noted that the embodiment of the present application updates the sorting of the operation sample data according to the numerical comparison relationship of the information entropy indexes corresponding to the plurality of operation sample data respectively, so that the operation sample data with higher complexity identified is preferentially included in the evaluation sequence, and the operation sample data with more significant model performance improvement is preferentially processed. Then, the pre-trained neural network cascade model is used to predict the transient stability class of the updated evaluation sequence, and the prediction result is used to optimize the pre-trained neural network cascade model, so as to improve the accuracy and efficiency of model evaluation, and enhance the adaptability of the model to the time-varying operation mode and topology of the power system, thereby efficiently evaluating the transient stability of large-scale power operation data.

[0054] In some embodiments, the higher the information entropy index of the running sample data, the greater the uncertainty of its prediction results, and the more complex or critical the information it may contain, thus making it more necessary to prioritize its evaluation. To accurately calculate the information entropy index of each running sample data, a step of calculating the information entropy index corresponding to each of the multiple running sample data is included before step S2.

[0055] The steps for calculating the information entropy index corresponding to multiple running sample data include: steps S201~S202.

[0056] Step S201: For each running sample data, input the running sample data into the pre-trained neural network model and output the predicted probability of each transient stable category corresponding to the running sample data.

[0057] The pre-trained neural network model is a neural network model that has been trained in advance. Its input is the labeled running sample data, and its output is the predicted probability of the label of the transient stable category corresponding to the running sample data.

[0058] It should be noted that a neural network model can be a single neural network, a neural network model within a cascaded neural network model, or a neural network model independent of the cascaded neural network model; no restrictions are imposed here.

[0059] Step S202: Calculate the information entropy index corresponding to the running sample data by weighted summation of the predicted probabilities of each transient stable category.

[0060] The method for calculating the information entropy index of the running sample data is as follows:

[0061]

[0062] In the formula, To measure the information entropy of sample data i, Let represent the output vector of the i-th running sample data evaluated by the neural network model, where each element represents the predicted probability that the sample is stable, critically stable, or unstable. is the index of the transiently stable category, and L is the number of transiently stable categories, which is set to 3 in this invention.

[0063] In some embodiments, such as Figure 3 As shown, step S2, which updates the first evaluation sequence based on the numerical comparison relationship of the information entropy index corresponding to multiple running sample data to form the second evaluation sequence, includes steps S211 to S212. Wherein:

[0064] Step S211, compare the information entropy indexes corresponding to the plurality of running sample data respectively with the information entropy threshold value, and screen out the plurality of running sample data with the information entropy index greater than the information entropy threshold value as key running sample data.

[0065] The information entropy threshold value is a preset information entropy empirical value.

[0066] Step S212, arrange the plurality of key running sample data in descending order according to the numerical comparison relationship of the information entropy indexes, and obtain a second evaluation sequence.

[0067] The identified key running sample data is preferentially included in the evaluation sequence, and is arranged in order from high to low according to the information entropy index, to form the second evaluation sequence, so as to ensure the priority setting of the model evaluation process, so that the sample with higher information value and potential influence can be evaluated earlier.

[0068] In some embodiments, the embodiment of the present application sets a dynamic updating mechanism of the evaluation queue, enhances the adaptability of the model to the complex power system environment, and guarantees the real-time reliability of the evaluation. After step S3 in the present application, steps S301-S202 are further included. Wherein:

[0069] Step S301, after the network parameter optimization of the pre-trained neural network model, the information entropy index corresponding to each running sample data is recalculated based on the pre-trained neural network model after the network parameter optimization.

[0070] It should be noted that when the neural network cascade model is difficult to obtain the transient stability class sample due to insufficient credibility and accumulates to a certain number, for example, the neural network cascade model does not obtain the transient stability class unstable after reaching the preset maximum number of iterations, the network parameter optimization of the pre-trained neural network model is needed, and the information entropy index corresponding to each running sample data is recalculated based on the pre-trained neural network model after the network parameter optimization, so that the second evaluation sequence is updated synchronously, so as to further optimize the evaluation process and improve the evaluation efficiency and accuracy.

[0071] Step S302, update the second evaluation sequence by using the numerical comparison relationship of the recalculated information entropy indexes corresponding to the running sample data.

[0072] In some embodiments, the pre-trained neural network cascade model includes a plurality of serially connected neural network models, and the neural network model can adopt a recurrent neural network (RNN). The plurality of serially connected neural network models can include a first neural network model, a second neural network model, a third neural network model, and an nth neural network model.

[0073] In the embodiments of the present application, the process of using the pre-trained neural network cascade model to perform transient stability class prediction on the second evaluation sequence and using the prediction result to optimize the pre-trained neural network cascade model in step S3 includes steps S311-S316. Wherein:

[0074] Step S311, input the second evaluation sequence into the first neural network model in the plurality of pre-trained neural network cascade models for prediction to obtain the transient stability class prediction result corresponding to the second evaluation sequence.

[0075] The first neural network model can be the first neural network model in the neural network cascade model, or can be other non-last neural network model.

[0076] Step S312, determine the confidence of the second evaluation sequence using the transient stability class prediction result corresponding to the second evaluation sequence and the transient stability class simulation result.

[0077] The transient stability class simulation result of the second evaluation sequence is determined by the simulator transient stability time domain simulation, whether the final state is transient stability or instability, or critical instability. The transient stability class simulation result can be understood as the actual transient stability class, and the confidence of the second evaluation sequence is determined by the comparison result of the transient stability class prediction result corresponding to the second evaluation sequence and the transient stability class simulation result.

[0078] Step S313, when the confidence of the second evaluation sequence does not reach the preset confidence threshold, update the plurality of running sample data after expanding the current running simulation time window in the second evaluation sequence.

[0079] When the confidence of the second evaluation sequence does not reach the preset confidence threshold, it means that the data of the current running simulation time window is not enough to make an accurate judgment, at this time the model will request the simulator to continue running to expand the time window to obtain more data. With the advancement of the simulation process, new and longer time window data is passed to the next neural network, which performs more in-depth feature extraction and stability evaluation based on more abundant information.

[0080] Exemplarily, as Figure 4As shown, the evaluator adopts a neural network cascade model, which includes a plurality of serially connected neural network models. The neural network models can adopt RNN recurrent neural networks. Since the update and length of the simulation time window are different, it is for evaluating a single sample based on the cascade neural network structure. When evaluating a sample, the simulator is used to simulate the sample, and the simulation data of 0-0.5s is input into the first neural network model CNN-1 under the evaluator; the simulation is continued to 0-1s, and the simulation data of 0-1s is input into the first neural network model CNN-2, and so on. With the continuous increase of the simulation time, the simulation time window of the input data received by the neural network at the back is longer, the evaluation is more accurate, until the confidence of the transient stability class as stable is 99.9, the result of the transient stability class as stable is accepted, and the data set of the transient stability class is updated.

[0081] In one of the examples, in order to ensure the dynamics and real-time of the evaluation process, the information entropy index of the newly added sample in the evaluation queue is continuously calculated. By continuously calculating the information entropy index of the newly added running sample data, those newly added samples which have important influence on model optimization are quickly identified, and the sorting of the second evaluation queue is adjusted accordingly, so as to ensure the efficient use of evaluation resources.

[0082] Therefore, the step of updating the second evaluation sequence with the plurality of running sample data after expanding the current running simulation time window includes steps S3131-S3132. Among them:

[0083] Step S3131, obtaining a plurality of running sample data after expanding the current running simulation time window.

[0084] Step S3132, updating the second evaluation sequence according to the value comparison relationship of the information entropy index corresponding to the plurality of running sample data after expanding the current running simulation time window.

[0085] It can be understood that by dynamically adjusting the evaluation queue in real time according to the new sample addition and model update, the accuracy and efficiency of the deep learning-based power system transient stability evaluation model are significantly improved, and the rapid adaptation ability of the model to the time-varying operation environment of the power system is enhanced.

[0086] Step S314, inputting the updated second evaluation sequence into the second neural network model adjacent to the first neural network model for prediction to obtain the transient stability class prediction result corresponding to the updated second evaluation sequence.

[0087] Step S315, determining the confidence of the updated second evaluation sequence by using the transient stability class prediction result corresponding to the updated second evaluation sequence and the transient stability class simulation result.

[0088] Step S316, when the confidence of the updated second evaluation sequence does not reach the preset confidence threshold, continue to update the second evaluation sequence with the multiple running sample data after expanding the current running simulation time window, and iterate in this way until the confidence of the second evaluation sequence reaches the preset confidence threshold or reaches the preset maximum running time window, the iteration stops, and the neural network cascade model after the iteration stops is output.

[0089] It can be understood that the cascade mechanism of the series connection of the neural network cascade model ensures that the model can dynamically adjust the simulation time window according to the actual evaluation needs, which avoids unnecessary long-time simulation and ensures the accuracy and reliability of the evaluation results.

[0090] Since the cascade architecture of the neural network cascade model is composed of multiple layers of neural networks, the output is the transient stable class output by the last layer of neural networks. At the same time, the input data time window of the last layer of neural networks of the cascade architecture is the longest, and the received data is the most abundant, so the output result of the upstream network is more accurate. Accordingly, the final output of the cascade architecture can be used to correct and optimize the upstream neural network.

[0091] Therefore, in some embodiments, when the neural network cascade model iteration stops, the input quantity and output result of the last iteration of the neural network cascade model are used to optimize the network parameters of the neural network model in the neural network cascade model.

[0092] Exemplarily, as shown in Figure 5 The second evaluation sequence is denoted as "head-end-task queue-tail end", and 1, 2, …, i+1, i+2 in the second evaluation sequence represent different running sample data. The running sample data in the second evaluation sequence is simulated by using the time domain simulator, and the second evaluation sequence is predicted by using the neural network cascade model.

[0093] Among them, the area of the horizontal dashed line in the time domain simulator represents the time domain simulation window length, such as the first horizontal dashed line representing simulation to 0.5s, and the running sample data obtained by simulating 0~0.5s is given to the first neural network model for preliminary evaluation; the second and third horizontal dashed lines represent simulation to 1s and 2s, respectively, and the running sample data obtained by simulating 0~1s and 0~2s is respectively given to the second and third neural network models for evaluation. When the confidence of which neural network model meets the requirements, stop subsequent simulation, and there is no need to use the next level neural network for prediction.

[0094] For the i-th running sample data, the first neural network model CNN1-(input is 0~0.5s simulation data) evaluates the probability of being stable as 69.9%, which does not meet the requirement, and the prediction result is not retained, so the simulation is continued, and a longer simulation time window (0~1s simulation data) is input into the second neural network model CNN-2, the probability of being stable is evaluated as 85.2%, which still does not meet the requirement, and the prediction result is also not retained, further simulation is carried out, and a longer simulation time window (0~2s simulation data) is input into the third neural network model CNN-3, at this time, the probability of being stable is evaluated as 99.9%, which meets the requirement, and therefore the result is taken as the final evaluation result. At this time, the transient stability class is stable and is fed back to CNN-1 and CNN-2, that is, the two neural network models CNN-1 and CNN-2 are informed that the labels corresponding to the 0~0.5s and 0~1s data input by them are stable, which is equivalent to obtaining two input time window short-label determined sub-samples, and when the sub-samples accumulate to a certain number, the network parameters of the two neural network models CNN-1 and CNN-2 can be updated.

[0095] In addition, the information entropy index proposed in the present application is calculated by the first neural network CNN-1 in the cascade architecture. After simulation to the i-th sample, the CNN-1 is used to preliminarily predict all running sample data, the three kinds of stable state probabilities obtained by prediction are used to calculate the information entropy index of each sample, the subsequent samples to be evaluated are sorted according to the index size, and the task queue is updated. For example, the original sample evaluation order is i, i+1, i+2…, after the information entropy value sorting, the new order becomes i, i+25, i+16…, so that the sample with the most significant model performance improvement is always processed preferentially.

[0096] Based on the same inventive concept, the embodiment of the present application also provides a priority-based power system transient stability evaluation device for implementing the priority-based power system transient stability evaluation method described above.

[0097] The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more priority-based power system transient stability evaluation device embodiments provided below can be referred to the limitations of the priority-based power system transient stability evaluation method in the above text, which will not be described here again.

[0098] As shown in Figure 6 The embodiment of the present application provides a priority-based power system transient stability evaluation device, which comprises:

[0099] The sequence forming module 100 is used to form a first evaluation sequence according to a plurality of running sample data of the power system;

[0100] The sequence updating module 200 is configured to update the first evaluation sequence according to a numerical comparison relationship of the information entropy indexes corresponding to the plurality of operation sample data, and form a second evaluation sequence.

[0101] The model evaluation module 300 is configured to perform transient stability class prediction on the second evaluation sequence by using the pre-trained neural network cascade model, and optimize the pre-trained neural network cascade model by using the prediction result. The pre-trained neural network cascade model is configured to input the operation sample data and output the prediction result of the transient stability class corresponding to the operation sample data.

[0102] In some embodiments, the device further comprises an information entropy calculation module configured to calculate the information entropy indexes corresponding to the plurality of operation sample data, including:

[0103] For each operation sample data, the operation sample data is input into the pre-trained neural network model, and the prediction probability of each transient stability class corresponding to the operation sample data is output.

[0104] The prediction probabilities of each transient stability class corresponding to the operation sample data are weighted and summed to obtain the information entropy index corresponding to the operation sample data.

[0105] In some embodiments, the sequence updating module 200 is specifically configured to compare the information entropy indexes corresponding to the plurality of operation sample data with an information entropy threshold, and filter out the plurality of operation sample data with the information entropy index greater than the information entropy threshold as key operation sample data. The plurality of key operation sample data are arranged in descending order according to the numerical comparison relationship of the information entropy indexes, and the second evaluation sequence is obtained.

[0106] In some embodiments, the device further comprises a first network optimization module configured to, after the pre-trained neural network model performs network parameter optimization, recalculate the information entropy index corresponding to each operation sample data based on the pre-trained neural network model after network parameter optimization. The second evaluation sequence is updated by using the numerical comparison relationship of the recalculated information entropy indexes corresponding to the operation sample data.

[0107] In some embodiments, the pre-trained neural network cascade model comprises a plurality of serially connected neural network models.

[0108] The model evaluation module 300 is specifically configured to input the second evaluation sequence into a first neural network model in the multi-previously trained neural network cascade model to perform prediction, to obtain a transient stability category prediction result corresponding to the second evaluation sequence; determine a confidence degree of the second evaluation sequence by using the transient stability category prediction result corresponding to the second evaluation sequence and the transient stability category simulation result; when the confidence degree of the second evaluation sequence does not reach a preset confidence degree threshold, update the plurality of running sample data after the current running simulation time window is expanded in the second evaluation sequence; input the updated second evaluation sequence into a second neural network model adjacent to the first neural network model to perform prediction, to obtain a transient stability category prediction result corresponding to the updated second evaluation sequence; determine a confidence degree of the updated second evaluation sequence by using the transient stability category prediction result corresponding to the updated second evaluation sequence and the transient stability category simulation result; when the confidence degree of the updated second evaluation sequence does not reach the preset confidence degree threshold, continue to update the plurality of running sample data after the current running simulation time window is expanded in the second evaluation sequence, and iteratively continue in this way until the confidence degree of the second evaluation sequence reaches the preset confidence degree threshold or reaches a preset maximum running time window, the iteration stops, and the neural network cascade model after the iteration stops is output.

[0109] In some embodiments, updating the plurality of running sample data after the current running simulation time window is expanded in the second evaluation sequence comprises: obtaining the plurality of running sample data after the current running simulation time window is expanded; and updating the second evaluation sequence according to a value comparison relationship of information entropy indexes respectively corresponding to the plurality of running sample data after the current running simulation time window is expanded.

[0110] In some embodiments, after the neural network cascade model stops iteration, network parameter optimization is performed on the neural network models in the neural network cascade model by using an input quantity and an output result of the last iteration of the neural network cascade model.

[0111] As shown in Figure 7 The embodiment of the present application also provides an electronic device, the electronic device 10 comprises a memory 20 and a processor 30, the memory 20 stores a computer program, and the computer program is executed by the processor 30, so that the processor 30 executes the steps of the power system transient stability evaluation method based on the priority strategy in any one of the above embodiments.

[0112] The embodiment of the present application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed to realize the steps of the power system transient stability evaluation method based on the priority strategy in any one of the above embodiments.

[0113] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the system, the electronic device and the computer storage medium described above can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0114] It should be noted that the terms "first", "second", and the like in the description and claims of the present application and the above drawings are used to distinguish similar objects, and do not necessarily have to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes 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.

[0115] In several embodiments provided by the present application, it can be understood that each block in the flowchart or block diagram can represent a module, a program segment or a part of code, and the module, the program segment or the part of code include one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur in different order from that noted in the drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and they can also be executed in reverse order, depending on the functions involved.

[0116] In several embodiments provided by the present application, it should be understood that the disclosed system, electronic device, computer storage medium and method can be implemented in other ways. For example, the above-described device embodiments are only schematic, for example, the division of units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0117] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, i.e. can be located in one place, or can be distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0118] In addition, each function unit in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software function unit.

[0119] When the integrated unit is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the entire or part of the technical solutions that essentially contribute to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for executing all or part of the steps of the method described in each embodiment 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.

[0120] The above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1. A method for power system transient stability assessment based on priority strategy, characterized in that, The method comprises the following steps: forming a first evaluation sequence according to a plurality of operation sample data of a power system; updating the first evaluation sequence according to a numerical comparison relationship of information entropy indexes corresponding to the plurality of operation sample data to form a second evaluation sequence; using a pre-trained neural network cascade model to predict a transient stability class of the second evaluation sequence, and using a prediction result to optimize the pre-trained neural network cascade model; wherein the pre-trained neural network cascade model is used to input the operation sample data and output a prediction result of the transient stability class corresponding to the operation sample data; the pre-trained neural network cascade model comprises a plurality of serially connected neural network models; the step of using the pre-trained neural network cascade model to predict the transient stability class of the second evaluation sequence, and using the prediction result to optimize the pre-trained neural network cascade model, comprises: inputting the second evaluation sequence into a first neural network model in the pre-trained neural network cascade model for prediction to obtain a transient stability class prediction result corresponding to the second evaluation sequence; using the transient stability class prediction result corresponding to the second evaluation sequence and a transient stability class simulation result to determine a confidence degree of the second evaluation sequence; when the confidence degree of the second evaluation sequence does not reach a preset confidence degree threshold, updating the second evaluation sequence with a plurality of operation sample data after expanding a current operation simulation time window; inputting the updated second evaluation sequence into a second neural network model adjacent to the first neural network model for prediction to obtain a transient stability class prediction result corresponding to the updated second evaluation sequence; using the transient stability class prediction result corresponding to the updated second evaluation sequence and a transient stability class simulation result to determine a confidence degree of the updated second evaluation sequence; when the confidence degree of the updated second evaluation sequence does not reach the preset confidence degree threshold, continuing to update the second evaluation sequence with a plurality of operation sample data after expanding a current operation simulation time window, and iteratively performing the same until the confidence degree of the second evaluation sequence reaches the preset confidence degree threshold or reaches a preset maximum operation time window, and the iteration stops, and an iteration-stopped neural network cascade model is output.

2. The priority policy based power system transient stability assessment method according to claim 1, wherein, Before the step of updating the first evaluation sequence according to a numerical comparison relationship of information entropy indexes corresponding to the plurality of operation sample data to form a second evaluation sequence, the method further comprises the step of calculating the information entropy indexes corresponding to the plurality of operation sample data; the step of calculating the information entropy indexes corresponding to the plurality of operation sample data comprises: for each operation sample data, inputting the operation sample data into a pre-trained neural network model to output a prediction probability of each transient stability class corresponding to the operation sample data; using the prediction probability of each transient stability class corresponding to the operation sample data to perform weighted summation calculation to obtain an information entropy index corresponding to the operation sample data.

3. The priority policy based power system transient stability assessment method according to claim 1, wherein, The step of updating the first evaluation sequence according to the numerical comparison relationship of the information entropy indexes respectively corresponding to the plurality of operation sample data to form a second evaluation sequence comprises: Comparing the information entropy indexes respectively corresponding to the plurality of operation sample data with an information entropy threshold value, and screening out the plurality of operation sample data with the information entropy indexes greater than the information entropy threshold value as key operation sample data; Arranging the plurality of key operation sample data in descending order according to the numerical comparison relationship of the information entropy indexes to obtain a second evaluation sequence.

4. The method for priority policy based power system transient stability assessment as claimed in claim 2, wherein, The step of using the pre-trained neural network cascade model to predict the transient stability class of the second evaluation sequence and using the prediction result to optimize the pre-trained neural network cascade model comprises: After the network parameter optimization of the pre-trained neural network model, recalculating the information entropy index corresponding to each operation sample data based on the pre-trained neural network model with optimized network parameters; Using the numerical comparison relationship of the recalculated information entropy indexes corresponding to the operation sample data to update the second evaluation sequence.

5. The priority policy based power system transient stability assessment method according to claim 1, wherein, The step of updating the second evaluation sequence with the plurality of operation sample data after expanding the current operation simulation time window comprises: Obtaining the plurality of operation sample data after expanding the current operation simulation time window; Updating the second evaluation sequence according to the numerical comparison relationship of the information entropy indexes respectively corresponding to the plurality of operation sample data after expanding the current operation simulation time window.

6. The priority policy based power system transient stability assessment method according to claim 1, wherein, When the neural network cascade model iteration stops, using the input quantity and output result of the last iteration of the neural network cascade model to optimize the network parameters of the neural network model in the neural network cascade model.

7. A device for power system transient stability assessment based on priority strategy, characterized in that, Comprise: A sequence forming module configured to form a first evaluation sequence according to a plurality of operation sample data of a power system; A sequence updating module configured to update the first evaluation sequence according to a numerical comparison relationship of information entropy indexes respectively corresponding to the plurality of operation sample data to form a second evaluation sequence; A model evaluation module configured to use a pre-trained neural network cascade model to predict the transient stability class of the second evaluation sequence and use the prediction result to optimize the pre-trained neural network cascade model; wherein the pre-trained neural network cascade model is configured to input the operation sample data and output the prediction result of the transient stability class corresponding to the operation sample data; The pre-trained neural network cascade model comprises a plurality of serially connected neural network models; The step of using the pre-trained neural network cascade model to predict the transient stability class of the second evaluation sequence and using the prediction result to optimize the pre-trained neural network cascade model comprises: inputting the second evaluation sequence into a first neural network model in the pre-trained neural network cascade model for prediction to obtain a transient stability category prediction result corresponding to the second evaluation sequence; determining a confidence degree of the second evaluation sequence by using the transient stability category prediction result corresponding to the second evaluation sequence and a transient stability category simulation result; when the confidence degree of the second evaluation sequence does not reach a preset confidence degree threshold, updating the second evaluation sequence by using a plurality of running sample data after expanding a current running simulation time window; inputting the updated second evaluation sequence into a second neural network model adjacent to the first neural network model for prediction to obtain a transient stability category prediction result corresponding to the updated second evaluation sequence; determining a confidence degree of the updated second evaluation sequence by using the transient stability category prediction result corresponding to the updated second evaluation sequence and a transient stability category simulation result; when the confidence degree of the updated second evaluation sequence does not reach the preset confidence degree threshold, continuing to update the second evaluation sequence by using a plurality of running sample data after expanding a current running simulation time window, and iteratively performing the same until the confidence degree of the second evaluation sequence reaches the preset confidence degree threshold or reaches a preset maximum running time window, and stopping the iteration, and outputting a neural network cascade model after the iteration is stopped.

8. An electronic device, comprising: The electronic device comprises a memory and a processor, the memory stores a computer program, and the computer program is executed by the processor to make the processor execute the steps of the power system transient stability evaluation method based on the priority strategy according to any one of claims 1 to 6.

9. A computer-readable storage medium having stored thereon a computer program, characterized in that The computer program is executed to implement the steps of the power system transient stability evaluation method based on the priority strategy according to any one of claims 1 to 6.

Images