Generation method and device of power grid operation mode, computer equipment and storage medium

Abstract

The invention relates to a power grid operation mode generation method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring historical data of a power grid operation mode; wherein the historical data comprises environment data and operation data; inputting the historical data into a preset data generation model to obtain sample data; wherein the data generation model comprises a first sub-model and a second sub-model, and the first sub-model is used for generating candidate sample data by using historical data and random noise; the second sub-model is used for judging data sources of historical data and the candidate sample data and determining the sample data; and inputting the sample data into a preset environment identification model to obtain an environment mode of the power grid operation mode and corresponding target operation data. By adopting the method, intelligent adjustment and optimization of the operation data of the power grid operation mode can be realized, and the stability and the safety of a power system are improved.

Description

Technical Field

[0001] The present application relates to the field of electrical technology, and in particular to a method, device, computer equipment, storage medium, and computer program product for generating a power grid operation mode. Background Art

[0002] With the continuous advancement of electrical technology, the high concentration of urban populations and industries has led to a surge in electricity demand, which continues to grow. This demand is complex and volatile, affected by various factors such as weather changes, holidays, and emergencies, making electricity demand forecasting significantly more difficult. For these reasons, traditional power forecasting techniques typically simulate extreme conditions (such as extreme power shortages) to identify weaknesses in grid structure, equipment configuration, and operating methods. However, these traditional extreme scenario generation methods rely primarily on historical data and statistical analysis, and therefore have significant limitations in addressing emerging risks. Summary of the Invention

[0003] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium and computer program product for generating a power grid operation mode to address the above technical problems.

[0004] In a first aspect, the present application provides a method for generating a power grid operation mode. The method comprises:

[0005] Acquiring historical data on power grid operation modes; wherein the historical data includes environmental data and operation data;

[0006] Inputting the historical data into a preset data generation model to obtain sample data; wherein the data generation model includes a first sub-model and a second sub-model, wherein the first sub-model is used to generate candidate sample data using historical data and random noise; and the second sub-model is used to distinguish the data sources of the historical data and the candidate sample data and determine the sample data;

[0007] The sample data is input into a preset environment recognition model to obtain the environment model of the power grid operation mode and the corresponding target operation data.

[0008] In one embodiment, inputting the historical data into a preset data generation model to obtain sample data includes:

[0009] Using the historical data of the power grid operation mode, establishing first sample sub-data;

[0010] Inputting the historical data into a preset data generation model to obtain second sample sub-data;

[0011] The first sample sub-data and the second sample sub-data are determined as sample data.

[0012] In one embodiment, the second sub-model is further used to:

[0013] Inputting the candidate sample data into the second sub-model to predict the data source of the candidate sample data; wherein the data source includes historical data and model generation;

[0014] When the proportion of the predicted candidate sample data whose data source is historical data reaches a preset threshold, the predicted candidate sample data whose data source is historical data is determined as sample data.

[0015] In one embodiment, the first sub-model is obtained by:

[0016] Establishing an initial first sub-model; wherein the first sub-model is used to generate random noise, and generate candidate sample data using the random noise and historical data;

[0017] Inputting the historical data into the initial first sub-model, and outputting candidate sample data;

[0018] Inputting the candidate sample data and historical data into the second sub-model, and adjusting the parameters of the initial first sub-model based on the prediction results;

[0019] When the proportion of historical data that is the source of candidate sample data and historical data predicted by the second sub-model reaches a preset threshold, the initial first sub-model is determined as the first sub-model.

[0020] In one embodiment, obtaining the environment recognition model includes:

[0021] Establishing an initial environment recognition model; wherein the initial environment recognition model includes a reward mechanism and a penalty mechanism; determining target operating data under each environment mode by establishing the mechanism and the penalty mechanism;

[0022] The sample data is input into the environment recognition model for training to obtain the environment recognition model.

[0023] In one embodiment, inputting the historical data into a preset data generation model to obtain sample data includes:

[0024] performing standardization processing on the historical data to obtain standard data;

[0025] The standard data is input into a preset data generation model to obtain sample data.

[0026] In one embodiment, the method further comprises:

[0027] Get current data on how the grid is operating;

[0028] determining a current environmental mode of the power grid operation mode based on the current data and reference data corresponding to each environmental mode of the power grid operation mode;

[0029] The operation data of the power grid operation mode is adjusted to the target operation data corresponding to the current environmental mode for operation.

[0030] In a second aspect, the present application further provides a device for generating a power grid operation mode. The device comprises:

[0031] A data acquisition module is used to acquire historical data of the power grid operation mode; wherein the historical data includes environmental data and operation data;

[0032] A sample generation module is configured to input the historical data into a preset data generation model to obtain sample data; wherein the data generation model includes a first sub-model and a second sub-model, wherein the first sub-model is configured to generate candidate sample data using historical data and random noise; and the second sub-model is configured to identify the data sources of the historical data and the candidate sample data and determine the sample data;

[0033] The mode prediction module is used to input the sample data into a preset environment recognition model to obtain the environmental mode of the power grid operation mode and the corresponding target operation data.

[0034] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method for generating a power grid operation mode as described in any one of the embodiments of the present disclosure is implemented.

[0035] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for generating a power grid operation mode as described in any one of the embodiments of the present disclosure.

[0036] In a fifth aspect, the present application further provides a computer program product, comprising a computer program that, when executed by a processor, implements the method for generating a power grid operation mode as described in any one of the embodiments of the present disclosure.

[0037] The above-mentioned method, device, computer equipment, storage medium and computer program product for generating the power grid operation mode obtain the historical data of the power grid operation mode, generate sample data according to the data generation model, and input the sample data into the environment recognition model, thereby obtaining the environmental pattern and target operation data of the power grid operation mode. This solution realizes the intelligent processing and utilization of historical data by constructing a data generation model and an environment recognition model. The data generation model can generate sample data that is similar to but not exactly the same as the historical data, which increases the diversity and richness of the data and provides strong support for the subsequent environmental pattern recognition and determination of the target operation data. The environment recognition model can accurately identify the environmental pattern of the power grid operation mode based on the input sample data, and determine the corresponding target operation data according to the environmental pattern, thereby realizing the intelligent adjustment and optimization of the operation data of the power grid operation mode and improving the stability and security of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is a flow chart of a method for generating a power grid operation mode in one embodiment;

[0039] Figure 2 A schematic diagram of a process for determining sample data in one embodiment;

[0040] Figure 3 A schematic diagram of a flow chart for use of the second sub-model in one embodiment;

[0041] Figure 4 A schematic diagram of a process for obtaining a first sub-model in one embodiment;

[0042] Figure 5 A schematic diagram of a process for obtaining an environment recognition model in one embodiment;

[0043] Figure 6 Schematic diagram of a process for data standardization in one embodiment;

[0044] Figure 7 A schematic diagram of a process for using target operation data in one embodiment;

[0045] Figure 8 is a structural block diagram of a device for generating a power grid operation mode in one embodiment;

[0046] Figure 9 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0048] In one embodiment, Figure 1 As shown, a method for generating a power grid operation mode is provided. This embodiment uses the method applied to a terminal as an example for illustration. It is understandable that the method can also be applied to a server, or to a system including a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0049] Step S100: Acquire historical data of power grid operation mode; wherein the historical data includes environmental data and operation data.

[0050] In an exemplary embodiment, the grid operation mode may include the conversion of voltage and current within the power system, the location where electrical energy is received and distributed, and the like. The environmental data may include the time of occurrence, weather conditions, temperature, humidity, and the like. The operational data may include electrical data within the grid operation mode, specifically, grid line power, power plant active power, node voltage, load, and other data. This data may be cleaned, standardized, and subjected to data structure changes, and then subsequently processed. The historical data may include historical records of grid operation modes, such as environmental and operational data from a past period of time.

[0051] In an exemplary embodiment, the power grid may include a system consisting of substations and transmission and distribution lines of various voltages within the power system; for example, a distribution network. Specifically, the power grid may include multiple units such as substations, transmission lines, and distribution lines. The power grid operation mode may include different operating states and modes of the power grid, such as normal, emergency, and recovery states, as well as the topology, equipment configuration, and operating parameters of the power grid in these states. Specifically, the historical power grid operation mode data may include substation operation data.

[0052] Step S200, input the historical data into a preset data generation model to obtain sample data; wherein, the data generation model includes a first sub-model and a second sub-model, the first sub-model is used to generate candidate sample data using historical data and random noise; the second sub-model is used to determine the data source of the historical data and the candidate sample data, and determine the sample data.

[0053] In one exemplary embodiment, the data generation model can be built based on a deep learning algorithm, such as a generative adversarial network (GAN). The first sub-model can serve as a generator, generating candidate sample data based on input historical data and random noise. The second sub-model can serve as a discriminator, identifying the source of the input data—that is, determining whether the data originates from historical data or candidate sample data generated by the first sub-model—and determining the final sample data based on the discrimination result. In this way, sample data similar to, but not identical to, the historical data can be generated for subsequent environmental pattern recognition and determination of target operational data.

[0054] In an exemplary embodiment, the random noise can be randomly generated by the first sub-model, or can be input simultaneously when historical data is input; the first sub-model can generate candidate sample data based on historical data and random noise, and the candidate sample data can include data consisting only of historical data and data generated by the model based on the fusion of historical data and random noise; after obtaining the candidate data, the candidate data can be input into the second sub-model, and the second sub-model can be used to determine whether the candidate data is historical data or generated by the model. When the possibility that the second sub-model determines that the candidate sample data is historical data is less than a preset threshold, the parameters of the first sub-model are adjusted and the candidate sample data is regenerated. When the possibility that the second sub-model determines that the candidate sample data is historical data is greater than a preset threshold, the candidate sample data determined by the second sub-model to be historical data is determined as sample data, etc.

[0055] Step S300: input the sample data into a preset environment recognition model to obtain the environment model of the power grid operation mode and the corresponding target operation data.

[0056] In an exemplary embodiment, the environmental recognition model can be constructed based on a machine learning algorithm, such as a support vector machine (SVM), random forest (RF), or neural network (NN). The environmental recognition model can identify the environmental mode in which the power grid operates based on input sample data and determine corresponding target operating data based on the environmental mode. Environmental modes can include different weather conditions, temperature ranges, load levels, etc., and target operating data can include optimal operating data for the power grid under different environmental modes, such as voltage, current, and power. In this way, the operating data of the power grid operation mode can be adjusted based on the current environmental mode of the power grid operation mode to ensure the stability and security of the power system.

[0057] In one exemplary embodiment, the environmental recognition model may include constructing an extreme scenario agent-optimized Markov decision model. Specifically, the model may include a state space for characterizing the environmental model and operating data of the power grid's operation; an action space for characterizing adjustable parameters of the power grid's operation; a reward function for calculating a reward value based on the current state and the action taken, where the reward value is determined based on the optimization objective of the environmental model; and a policy network for outputting the optimal action based on the current state. The policy network is trained using reinforcement learning methods. In one exemplary embodiment, the reinforcement learning methods may include deep deterministic policy gradient (DDPG) and proximal policy optimization (PPO). The trained policy network can output optimal adjustment parameters based on the current environmental model and operating data of the power grid to achieve the optimization objective of the environmental model, such as reducing losses and improving voltage quality.

[0058] In an exemplary embodiment, the environmental modes may include various extreme scenarios of power grid operation, such as extreme power shortages. It is understood that by simulating extreme scenarios, weaknesses in the power grid structure, equipment configuration, and operation can be identified. Training and prediction can be performed using multiple sample data to predict optimal operating data for each environmental mode (extreme scenario). This optimal operating data is then used when the power grid operates in that environmental mode to ensure the stability of the power grid's operation.

[0059] In the above-mentioned method for generating the grid operation mode, the environmental mode and target operation data of the grid operation mode are obtained by obtaining the historical data of the grid operation mode, generating sample data according to the data generation model, and inputting the sample data into the environment recognition model. This scheme realizes the intelligent processing and utilization of historical data by constructing a data generation model and an environment recognition model. The data generation model can generate sample data that is similar to but not exactly the same as the historical data, which increases the diversity and richness of the data and provides strong support for the subsequent environmental mode recognition and determination of the target operation data. The environment recognition model can accurately identify the environmental mode of the grid operation mode based on the input sample data, and determine the corresponding target operation data according to the environmental mode, thereby realizing the intelligent adjustment and optimization of the grid operation mode operation data and improving the stability and security of the power system.

[0060] In one embodiment, Figure 2 As shown, the historical data is input into a preset data generation model to obtain sample data, including:

[0061] Step S201: Create first sample sub-data using the historical data of the power grid operation mode.

[0062] Step S202: input the historical data into a preset data generation model to obtain second sample sub-data.

[0063] Step S203: Determine the first sample sub-data and the second sample sub-data as sample data.

[0064] In an exemplary embodiment, the data generation model can generate first sample data based on historical data, and generate second sample data using random noise and historical data, and the first sample data and the second sample data are both sample data, etc.

[0065] In one exemplary embodiment, when the probability that the second sub-model predicts that the candidate sample data is historical data exceeds a preset threshold, all candidate sample data output by the first sub-model can be determined as sample data. Specifically, the parameters of the first sub-model can be trained in advance to accurately generate sample data. Subsequently, historical data can be acquired at preset intervals and sample data can be regenerated, thereby ensuring the real-time and accuracy of the sample data. In this way, the sample data can be continuously updated and optimized to more closely reflect the actual operation of the power grid and environmental changes, thereby improving the accuracy of environmental pattern recognition and the certainty of target operating data.

[0066] In this embodiment, by combining the first sample sub-data generated by historical data and the second sample sub-data generated by the model, not only the data set is enriched, but also the diversity and authenticity of the data set are improved, providing a more comprehensive and reliable data foundation for subsequent environmental pattern recognition and determination of target operation data.

[0067] In one embodiment, Figure 3 As shown, the second sub-model is also used for:

[0068] Step S211: input the candidate sample data into the second sub-model to predict the data source of the candidate sample data; wherein the data source includes historical data and model generation.

[0069] In step S212, when the proportion of the predicted candidate sample data whose data source is historical data reaches a preset threshold, the predicted candidate sample data whose data source is historical data is determined as sample data.

[0070] In an exemplary embodiment, the second sub-model can not only have the function of distinguishing the source of data, but also intelligently screen and determine the sample data when the prediction results meet certain conditions. Specifically, when the second sub-model predicts that the proportion of candidate sample data derived from historical data reaches a preset threshold, these candidate sample data predicted to be derived from historical data can be regarded as valid sample data. This mechanism ensures the quality of the sample data, making it closer to the actual historical data distribution, thereby improving the accuracy of subsequent environmental pattern recognition and target operation data determination. Through this intelligent screening process, we can more effectively utilize the generated data, optimize the operation strategy of the power grid operation mode, and enhance the stability and security of the power system.

[0071] In one exemplary embodiment, only the data generated by the model can be input into the second sub-model, and the second sub-model can be used to determine the source of the data. In another exemplary embodiment, historical data and the data generated by the model can be input into the second sub-model together, and the second sub-model can be used to determine the source of the data, etc.

[0072] In one exemplary embodiment, when the proportion of predicted candidate sample data sourced from historical data reaches a preset threshold, the candidate sample data sourced from historical data is determined as the sample data. Simultaneously, the prediction of the newly input candidate sample data can be used to modify the parameters of the first sub-model to further optimize the performance of the first sub-model and improve the quality of the candidate sample data. In this way, the data generation process can be continuously iterated and optimized, ensuring that the generated sample data is more realistic and effective, providing stronger support for environmental pattern recognition and the determination of target operation data.

[0073] In this embodiment, the intelligent filtering function of the second sub-model not only improves the quality of sample data but also enhances the adaptability of the data generation model. When the proportion of candidate sample data derived from historical data reaches a preset threshold, these candidate sample data are automatically screened as valid sample data. This mechanism ensures the authenticity and reliability of the sample data. Simultaneously, through iterative optimization of the first sub-model parameters, the performance of the data generation model is further improved, enabling it to generate sample data that is more closely aligned with the actual operation of the power grid and environmental changes.

[0074] In one embodiment, Figure 4 As shown, the method for obtaining the first sub-model includes:

[0075] Step S221, establishing an initial first sub-model; wherein, the first sub-model is used to generate random noise, and use the random noise and historical data to generate candidate sample data.

[0076] Step S222: input the historical data into the initial first sub-model, and output candidate sample data.

[0077] Step S223: Input the candidate sample data and historical data into the second sub-model, and adjust the parameters of the initial first sub-model based on the prediction result.

[0078] Step S224: when the second sub-model predicts that the proportion of the data source of the candidate sample data and the historical data as historical data reaches a preset threshold, the initial first sub-model is determined as the first sub-model.

[0079] In an exemplary embodiment, the initial first sub-model may include a generator, etc. The initial first sub-model may use the input historical data and random noise to generate candidate sample data. The initial first sub-model may also use the input historical data to randomly generate random noise, and use random noise and historical data to generate candidate sample data, etc.

[0080] In an exemplary embodiment, the candidate sample data and historical data can be input into the second sub-model, and the second sub-model predicts the data sources of the candidate sample data and historical data. Based on the prediction results, the degree of similarity between the candidate sample data and the historical data can be understood, and then the parameters of the initial first sub-model can be adjusted to improve the ability of the first sub-model to generate candidate sample data. When the second sub-model predicts that the proportion of historical data that is the data source of the candidate sample data and historical data reaches a preset threshold, it means that the initial first sub-model already has a certain generation capability. At this time, the initial first sub-model can be determined as the first sub-model for subsequent data generation tasks. In this way, the accuracy and reliability of the first sub-model can be ensured, providing strong support for subsequent data generation and environmental pattern recognition.

[0081] In an exemplary embodiment, the parameters of the first sub-model can be further fine-tuned based on actual needs. For example, based on the prediction results of the second sub-model, optimization algorithms such as gradient descent can be applied to the weights, biases, and other parameters in the generator of the first sub-model to ensure that the candidate sample data generated by the first sub-model more closely matches the actual historical data distribution. This step can be repeated until the performance of the first sub-model meets the preset conditions. In this way, the generation capability and accuracy of the data generation model can be further improved, providing more reliable data support for subsequent environmental pattern recognition and determination of target operation data.

[0082] In this embodiment, by establishing an initial first sub-model, the model can generate candidate sample data using input historical data and random noise. The generated candidate sample data is input into the second sub-model together with the historical data, and the second sub-model predicts their data sources. The purpose of this step is to evaluate the similarity between the candidate sample data and the historical data, thereby guiding the parameter adjustment of the initial first sub-model. The generation capability and accuracy of the data generation model can be further improved, providing more reliable data support for subsequent environmental pattern recognition and determination of target operation data. It ensures that the first sub-model in the data generation model has strong generation capability and accuracy. This design not only improves the quality and diversity of the sample data, but also provides a more comprehensive and reliable data basis for subsequent environmental pattern recognition and determination of target operation data.

[0083] In one embodiment, Figure 5 As shown, the acquisition of the environment recognition model includes:

[0084] Step S301, establishing an initial environment recognition model; wherein, the initial environment recognition model includes a reward mechanism and a penalty mechanism; and determining target operating data in each environment mode by establishing the mechanism and the penalty mechanism.

[0085] Step S302: input the sample data into the environment recognition model for training to obtain the environment recognition model.

[0086] In an exemplary embodiment, the initial environment recognition model may include structures such as a deep learning network for feature extraction and pattern recognition of input sample data. Reward and penalty mechanisms can be designed based on the operational objectives and optimization requirements of the power grid operation mode. For example, corresponding reward and penalty rules can be set for optimization objectives such as reducing losses and improving voltage quality. During the training process, the parameters of the environment recognition model are continuously adjusted so that it can accurately identify target operating data under different environmental modes. The purpose of this step is to build a model that can accurately identify the environmental mode of the power grid operation mode and determine the corresponding target operating data, providing strong support for subsequent intelligent adjustment and optimization. Once the environment recognition model is trained, it can be applied to the adjustment and optimization of operating data of the actual power grid operation mode, thereby improving the stability and security of the power system. This design not only improves the accuracy and reliability of the environment recognition model but also provides a more comprehensive and effective solution for the intelligent operation and optimization of the power grid operation mode.

[0087] In one exemplary embodiment, the environmental recognition model can include constructing an extreme scenario agent optimization and adjustment Markov decision model. Through continuous learning and optimization, this model can accurately predict and adjust the optimal operating strategy for various extreme scenarios. In specific implementations, historical and real-time data on the power grid's operating mode can be input into the trained environmental recognition model. The model will output corresponding target operating data based on the current environmental pattern (such as weather conditions, temperature range, load level, etc.) and preset optimization objectives (such as reducing losses and improving voltage quality). Based on the target operating data output by the model, operators can adjust the operating parameters of the power grid's operating mode accordingly to improve the stability and security of the power system.

[0088] In an exemplary embodiment, the environmental recognition model can construct a state space set encompassing generator sets, balancing units, weak line load factors, weak node voltages, and single-step dynamic active power adjustment, as well as an action space set. The state space set represents the current environmental mode and operating status of the power grid, while the action space set represents the available regulatory actions. The trained environmental recognition model can intelligently select the optimal action based on the input state space set, outputting the corresponding target operating data. This design enables the power grid to automatically adjust operating parameters based on the current environmental mode and operating status to achieve optimization goals such as reducing losses and improving voltage quality. Furthermore, by constructing a state space set encompassing generator sets, balancing units, weak line load factors, weak node voltages, and single-step dynamic active power adjustment, it can comprehensively reflect the operational status of the power grid, providing the environmental recognition model with richer and more accurate information and further improving the model's recognition accuracy and predictive capabilities. This solution provides strong technical support for the intelligent operation and optimization of power grid operations, contributing to improved power system stability and security.

[0089] In this embodiment, by constructing an initial environmental recognition model that incorporates both reward and penalty mechanisms and training it based on sample data, we developed a model capable of accurately identifying power grid operating environment modes and determining corresponding target operating data. This model's design fully considers the operational objectives and optimization requirements of the power grid operation mode. Through reward and penalty mechanisms, the model learns the optimal operating strategy. During training, the model continuously adjusts and optimizes its parameters until it can accurately identify target operating data under different environmental modes. The successful implementation of this step provides strong support for subsequent intelligent adjustment and optimization. When the environmental recognition model is applied to actual power grid operation, it can output accurate target operating data based on the current environmental mode and preset optimization objectives. This data provides operators with a scientific basis for decision-making, enabling them to adjust the operating parameters of the power grid operation mode accordingly, thereby improving the stability and security of the power system. This innovative design not only improves the accuracy and reliability of the environmental recognition model but also opens up new avenues for intelligent operation and optimization of power grid operation modes, providing a strong guarantee for the stable operation of the power system.

[0090] In one embodiment, Figure 6 As shown, the historical data is input into a preset data generation model to obtain sample data, including:

[0091] Step S231 , performing standardization processing on the historical data to obtain standard data.

[0092] Step S232: input the standard data into a preset data generation model to obtain sample data.

[0093] In an exemplary embodiment, historical data is standardized to eliminate outliers and noise in the data so that it meets the input requirements of the data generation model. Standardization can include steps such as data cleaning, missing value filling, and outlier processing to ensure the accuracy and consistency of the input data. Through standardization, more reliable and stable standard data can be obtained, providing strong support for subsequent data generation. The standard data is input into a preset data generation model, which generates multiple sample data based on its internal algorithms and parameters. This data can simulate the actual operating conditions and environmental changes of the power grid, providing a rich dataset for subsequent environmental pattern recognition and determination of target operating data. During the sample data generation process, the parameters of the data generation model can be adjusted according to actual needs to generate sample data that more closely reflects the actual operating conditions and environmental changes of the power grid. The successful implementation of this step provides strong support for subsequent environmental pattern recognition and determination of target operating data, helping to improve the stability and security of the power system.

[0094] In an exemplary embodiment, standard data can be input into a data generation model comprising a first sub-model and a second sub-model. The first sub-model is used to generate random noise, and use the random noise and historical data to generate candidate sample data. The second sub-model is used to identify the data source of the candidate sample data, that is, when the proportion of candidate sample data predicted to be derived from historical data reaches a preset threshold, these candidate sample data are screened as valid sample data. In this way, we can ensure that the generated sample data is more real and effective, and provide stronger data support for subsequent environmental pattern recognition and determination of target operation data. After generating the sample data, we can also input these sample data into the trained environmental recognition model for verification to further evaluate the performance and accuracy of the data generation model. The successful implementation of this step will provide us with a more reliable and effective data foundation, and provide strong guarantees for the stable operation and optimization of the power system.

[0095] In this embodiment, by standardizing historical data, we ensure data accuracy and consistency, providing strong support for subsequent data generation. This measure not only improves the quality and diversity of sample data, but also provides more comprehensive and reliable data support for subsequent environmental pattern recognition and determination of target operation data.

[0096] In one embodiment, Figure 7 As shown, the method further includes:

[0097] Step S401: Acquire current data of the power grid operation mode.

[0098] Step S402 : determining the current environmental mode of the power grid operation mode according to the current data and reference data corresponding to each environmental mode of the power grid operation mode.

[0099] Step S403: Adjust the operation data of the power grid operation mode to the target operation data corresponding to the current environmental mode for operation.

[0100] In an exemplary embodiment, current grid operation data can be obtained by real-time monitoring of grid operation data, such as voltage, current, and power. This data reflects the current operating state and environmental conditions of the grid operation mode. Subsequently, the current grid operation data is compared and analyzed against pre-set reference data corresponding to each environmental mode, such as historical operating data and environmental parameters, to determine the current environmental mode of the grid operation mode. Once the current environmental mode of the grid operation mode is determined, the operating parameters of the grid operation mode can be adjusted accordingly based on the target operating data corresponding to the environmental mode. The purpose of this step is to match the operating state of the grid operation mode with the environmental mode to achieve optimal operating results. In this way, intelligent adjustment and optimization of the grid operation mode can be achieved, improving the stability and security of the power system. This design not only improves the operating efficiency of the grid operation mode but also provides a new approach for its intelligent operation and optimization, providing a strong guarantee for the stable operation of the power system.

[0101] In an exemplary embodiment, when the environmental pattern of the power grid operation mode is obtained through the environmental recognition model, corresponding reference data can also be generated, and the current data can be compared with the reference data to determine the current environmental pattern, etc.

[0102] In one exemplary embodiment, after obtaining reference data and current data, a similarity calculation model can be constructed to calculate the similarity between the current data and the reference data. The environmental mode corresponding to the reference data with the highest similarity is then determined as the current environmental mode. The implementation of this step depends on the accuracy and reliability of the similarity calculation model. Therefore, when constructing the similarity calculation model, it is necessary to fully consider the operating characteristics of the power grid operation mode and the changing patterns of the environmental mode, and select appropriate algorithms and parameters for training and optimization. Through similarity calculation, the current environmental mode of the power grid operation mode can be quickly and accurately determined, providing strong support for subsequent adjustment and optimization of operating data. After determining the current environmental mode, the operating parameters of the power grid operation mode can be adjusted accordingly based on the target operating data corresponding to the environmental mode, thereby improving the stability and security of the power system. This design not only improves the operating efficiency of the power grid operation mode, but also provides new ideas and methods for its intelligent operation and optimization, helping to promote the intelligent and automated progress of the power system. In another exemplary embodiment, the Euclidean distance between the current data and each reference data can be calculated, and the environmental mode corresponding to the reference data with the smallest distance is determined as the current environmental mode. Euclidean distance is a commonly used distance metric that reflects the degree of similarity between data. By calculating the Euclidean distance between the current data and various reference data, we can select the reference data with the smallest distance, assuming that its corresponding environmental pattern is closest to the current environmental pattern. This step also relies on the accuracy and consistency of the data, so before calculating the distance, the data must be thoroughly preprocessed and standardized. This method allows us to more accurately determine the current environmental pattern of the power grid's operation, providing a more reliable basis for subsequent adjustment and optimization of operating data. After determining the current environmental pattern, we can adjust the operating parameters of the power grid's operation mode accordingly based on the target operating data corresponding to that environmental pattern, thereby improving the stability and security of the power system. This design not only improves the operational efficiency of the power grid's operation mode but also provides new approaches and methods for its intelligent operation and optimization, helping to promote the intelligentization and automation of the power system.

[0103] In this embodiment, by real-time monitoring of the operating data of the power grid operation mode, combined with the reference data corresponding to each pre-set environmental mode, we can quickly and accurately determine the current environmental mode of the power grid operation mode. The implementation of this step depends on an in-depth understanding of the operating characteristics of the power grid operation mode and accurate data comparison and analysis. Once the current environmental mode is determined, action can be taken immediately to adjust the operating data of the power grid operation mode to the target operating data corresponding to the environmental mode to ensure that the power grid operation mode operates in the optimal state. This intelligent adjustment and optimization mechanism not only improves the operating efficiency of the power grid operation mode, but also significantly enhances the stability and security of the power system. In addition, by continuously accumulating and optimizing the algorithms and models for environmental pattern recognition and target operation data determination, we can further improve the intelligent operation level of the power grid operation mode.

[0104] In an exemplary embodiment, the acquisition of the data generation model may include statistically analyzing extreme scenario data, occurrence time, and weather conditions based on historical data; after the flow calculation, obtaining data such as grid line power, power plant active power, node voltage, and load, and performing data processing steps such as data cleaning, standardization, and data structure changes to output false data that can be used to compare the generator output; and establishing a data set containing multi-dimensional features:

[0105] (1)

[0106] (2)

[0107] (3)

[0108] (4)

[0109] (5)

[0110] Among them, these characteristics include meteorological factors such as temperature, humidity, wind speed, and corresponding power grid load, power generation and line power, balance machine output and other power operation parameters. is a complete set of weather data, for Time temperature, for Humidity at all times for Wind speed at any moment, The complete set of observed variables for all node loads, lines, generators, and balancing units. Nodes can include busbars or substations. The subscripts represent l for line, real for active power, cap for capacitors, and min and max for minimum and maximum values, respectively. Thus, lreal represents the active power of the corresponding line. The superscripts represent the generators, lines, and loads connected to different buses. Data normalization converts values ​​of different dimensions into dimensionless, standardized form. The formula is as follows:

[0111] (6)

[0112] in, is the value before conversion, is the maximum value.

[0113] GAN is trained to provide a diverse starting point for the generator, randomly generating a set of noise vectors drawn from a uniform distribution, as follows:

[0114] (7)

[0115] in, is a noise vector containing a set of all multidimensional features, Respectively represent the temperature, load, line, generator and balancing machine feature parts in the vector; the five variables can be independent of each other, representing random changes in different types of data, etc. After the generator receives these noise vectors, it uses its deep neural network architecture to map these noises to the data space to generate false data; the false data With real data The training set is input into the discriminator. The task of the discriminator is to distinguish whether the input data is a real sample or a generated sample. It outputs a probability value through a series of convolutional layers, activation function ReLU and the final Sigmoid layer. , where P represents the possibility that the input data is a real sample, and the output value is between 0 and 1. A value close to 1 indicates a high probability that the sample is real, while a value close to 0 indicates that it is a generated fake sample. The GAN network can be trained for about 20,000 iterations. At different stages of training, the gradient penalty, label smoothing, learning rate, etc. are adjusted to improve the stability of training and the quality of generated samples. Error analysis, distribution comparison, and discriminator feedback are performed on the generated samples and real samples to comprehensively evaluate the performance of the generator in generating samples in extreme power grid scenarios. Specifically, a set of noise vectors are randomly generated. The distribution of these vectors should be similar to that in the training phase, still using a multi-dimensional Gaussian distribution. The generated samples are compared with real power grid data. Real data usually comes from historical extreme scenarios or actual power grid monitoring data. These data are preprocessed and cleaned to ensure accuracy during comparison. To evaluate the quality of the generated samples, the mean square error (MSE) is calculated to quantify the difference between the generated samples and the real samples. The formula is as follows:

[0116] (8)

[0117] in, is the number of samples, It is The true value of the sample, is the estimated value of the i-th sample. The MSE metric is used to understand the generator's ability to simulate extreme grid scenarios. Generated samples are fed into a trained discriminator and the discriminator's output on these samples is obtained. The discriminator's output will indicate the probability that the generated sample is recognized as a real sample, further verifying the generator's effectiveness. If the discriminator has a high recognition rate for the generated samples, it indicates that the generator performs well in simulating extreme grid scenarios. Otherwise, further training may be required. After completing GAN network training, the generated data is verified to verify its compliance with the physical laws and constraints of actual grid operation.

[0118] In an exemplary embodiment, the environment recognition model may include constructing an extreme scenario agent optimization and adjustment Markov decision model, specifically as follows:

[0119] Construct a state space set including generator sets, balancing units, weak line load rate, weak node voltage, and single-step dynamic active power adjustment size:

[0120] (9)

[0121] (10)

[0122] (11)

[0123] (12)

[0124] in, and are the complete sets of all observed variables related to the generator and the balancing machine, respectively. is the power flow load rate and channel capacity of each weak branch of the power system, is the standard value of voltage at each node in the power system.

[0125] (13)

[0126] (14)

[0127] (15)

[0128] (16)

[0129] In the formula When the agent acts, the total active power adjustment power allocated to the generator set in a single step, n is the total number of generators in the system. Construct the action space set for the agent to adjust the real-time output of the generator set:

[0130] (17)

[0131] Where m represents the number of power plants, and the agent selects a power plant each time. , the total output of all generators in the power plant will change.

[0132] For the selected power plant, power distribution needs to be completed among the different generators in the power plant:

[0133] (18)

[0134] (19)

[0135] In the formula is the total amount of active power that the power plant needs to regulate, is the change in output power allocated to each generator, and n is the number of generators in the power station.

[0136] A three-category reward and penalty function framework is established to better improve the agent's search direction, including power flow convergence rewards, balancing machine output rewards, and safety constraint rewards. The details are as follows:

[0137] Establish a power flow convergence reward function:

[0138] (20)

[0139] Establish the balancing machine output reward function:

[0140] (twenty one)

[0141] Among them are the balancing machines Before the action, the difference between the real-time output of the balancing machine and its lower power limit, The difference after the corresponding action.

[0142] Establish a safety constraint reward function:

[0143] (twenty two)

[0144] (twenty three)

[0145] (twenty four)

[0146] (25)

[0147] (26)

[0148] (27)

[0149] in, is the number of nodes with voltage exceeding the limit, is the number of overloaded lines.

[0150] Building a compound reward function:

[0151] (28)

[0152] in, The final composite reward function of the agent is obtained by dividing the dataset generated by the scenario into training and test sets for effective model training and evaluation.

[0153] In an exemplary embodiment, a scene generation agent can be trained. The specific training method is as follows: in each training iteration, the discriminator is first trained. The loss of the discriminator on real samples and fake samples is calculated by the cross entropy loss function, and the loss function of the generator is And the loss function of the discriminator The formula is as follows:

[0154] (29)

[0155] Loss function of the discriminator The formula is as follows:

[0156] (30)

[0157] represents the real sample data, represents the probability distribution of the real sample x, Represents the discriminator's response to the real sample The discrimination result output is is the noise sampled from a multidimensional Gaussian distribution, is the distribution in the latent space. is the discriminator’s judgment result on the real sample, is the sample generated by the generator, Indicates the generated sample Is the probability true, Indicates the generated sample The probability that it is false. The label is 1, and the generated sample The label is 0. Through these labels, the gradient information of the discriminator can be obtained, and the back propagation algorithm is used to update the weight of the discriminator to improve the discriminator's ability to distinguish between real and fake data. Through multiple iterations, the discriminator can gradually learn the characteristic distribution of real data. The generator will generate new fake samples , these fake samples are fed into the discriminator. The discriminator's output is used to calculate the generator's loss, with the goal of making the generated fake samples "deceive" the discriminator as much as possible. The generator's loss is finally calculated using the probability values ​​output by the discriminator, and the generator's parameters are updated using the backpropagation algorithm. This alternating training process places the generator and discriminator in a state of competition, where they continuously optimize and minimize the objective function. The objective function formula is as follows:

[0158] (31)

[0159] in, is a real data sample, is the distribution of the real data, is the noise sampled from the latent space, is the distribution in the latent space. is the discriminator’s judgment result on the real sample, are samples generated by the generator. The entire training process typically requires 20,000 iterations. At different stages of training, adjustments to gradient penalties, label smoothing, and the learning rate are necessary to improve training stability and the quality of generated samples. In this way, the GAN can effectively learn the underlying characteristics of the data and generate new data samples with practical application value, ultimately enabling the generator to produce high-quality samples that approximate the real-world data distribution. The dataset used to train the extreme scenario intelligent adjustment PPO agent employed in this application is generated using the aforementioned method. This dataset is divided into a training set and a test set for effective model training and evaluation. Specifically, the training set contains sufficient extreme scenario cases and is trained over 80,000 rounds to ensure that the model can learn from diverse data. During training, the agent autonomously adjusts its active power, taking 8 steps per round, within the specified number of steps. If the power flow calculation converges and there are no overloaded lines or node voltages exceeding the limit, the round ends early. Otherwise, the agent is penalized for failed policy adjustment. The agent's reward function converges around 27,000 rounds of training. This phenomenon shows that the intelligent agent has found a set of effective strategies under this topology, which can effectively alleviate the impact of extreme power shortages in the power system.

[0160] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0161] Based on the same inventive concept, embodiments of the present application further provide a device for generating a power grid operation mode for implementing the aforementioned method for generating a power grid operation mode. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more embodiments of the device for generating a power grid operation mode provided below can be found in the aforementioned limitations of the method for generating a power grid operation mode, and will not be further elaborated here.

[0162] In one embodiment, Figure 8 As shown, a device 100 for generating a power grid operation mode is provided, comprising: a data acquisition module 101, a sample generation module 102 and a mode prediction module 103, wherein:

[0163] A data acquisition module is used to acquire historical data of the power grid operation mode; wherein the historical data includes environmental data and operation data;

[0164] A sample generation module is configured to input the historical data into a preset data generation model to obtain sample data; wherein the data generation model includes a first sub-model and a second sub-model, wherein the first sub-model is configured to generate candidate sample data using historical data and random noise; and the second sub-model is configured to identify the data sources of the historical data and the candidate sample data and determine the sample data;

[0165] The mode prediction module is used to input the sample data into a preset environment recognition model to obtain the environmental mode of the power grid operation mode and the corresponding target operation data.

[0166] In one embodiment, the sample generation model is further used to:

[0167] Using the historical data of the power grid operation mode, establishing first sample sub-data;

[0168] Inputting the historical data into a preset data generation model to obtain second sample sub-data;

[0169] The first sample sub-data and the second sample sub-data are determined as sample data.

[0170] In one embodiment, the second sub-model is further used to:

[0171] Inputting the candidate sample data into the second sub-model to predict the data source of the candidate sample data; wherein the data source includes historical data and model generation;

[0172] When the proportion of the predicted candidate sample data whose data source is historical data reaches a preset threshold, the predicted candidate sample data whose data source is historical data is determined as sample data.

[0173] In one embodiment, the apparatus further comprises a model building module for:

[0174] Establishing an initial first sub-model; wherein the first sub-model is used to generate random noise, and generate candidate sample data using the random noise and historical data;

[0175] Inputting the historical data into the initial first sub-model, and outputting candidate sample data;

[0176] Inputting the candidate sample data and historical data into the second sub-model, and adjusting the parameters of the initial first sub-model based on the prediction results;

[0177] When the proportion of historical data that is the source of candidate sample data and historical data predicted by the second sub-model reaches a preset threshold, the initial first sub-model is determined as the first sub-model.

[0178] In one embodiment, the model building module is further configured to include:

[0179] Establishing an initial environment recognition model; wherein the initial environment recognition model includes a reward mechanism and a penalty mechanism; determining target operating data under each environment mode by establishing the mechanism and the penalty mechanism;

[0180] The sample data is input into the environment recognition model for training to obtain the environment recognition model.

[0181] In one embodiment, the standardization processing module is used to:

[0182] performing standardization processing on the historical data to obtain standard data;

[0183] The standard data is input into a preset data generation model to obtain sample data.

[0184] In one embodiment, the apparatus further comprises:

[0185] A data acquisition module is used to obtain current data on the operation mode of the power grid;

[0186] a mode determination module, configured to determine a current environmental mode of the power grid operation mode based on the current data and reference data corresponding to each environmental mode of the power grid operation mode;

[0187] The data adjustment module is used to adjust the operation data of the power grid operation mode to the target operation data corresponding to the current environmental mode for operation.

[0188] Each module in the aforementioned power grid operation mode generation device may be implemented in whole or in part via software, hardware, or a combination thereof. Each module may be embedded in or independent of a processor in a computer device in the form of hardware, or may be stored in a memory in the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0189] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 9As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store historical data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for generating a power grid operation mode is implemented.

[0190] Those skilled in the art will understand that Figure 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0191] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions.

[0192] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), programmable logic devices (PLDs), data processing logic devices based on quantum computing, and the like.

[0193] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0194] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for generating a power grid operation mode, characterized in that: The method comprises: Acquiring historical data on power grid operation modes; wherein the historical data includes environmental data and operation data; Inputting the historical data into a preset data generation model to obtain sample data; wherein the data generation model includes a first sub-model and a second sub-model, wherein the first sub-model is used to generate candidate sample data using historical data and random noise; and the second sub-model is used to distinguish the data sources of the historical data and the candidate sample data and determine the sample data; The sample data is input into a preset environment recognition model to obtain the environment model of the power grid operation mode and the corresponding target operation data.

2. The method according to claim 1, characterized in that The step of inputting the historical data into a preset data generation model to obtain sample data includes: Using the power grid operation mode historical data, establishing first sample sub-data; Inputting the historical data into a preset data generation model to obtain second sample sub-data; The first sample sub-data and the second sample sub-data are determined as sample data.

3. The method according to claim 1, characterized in that The second sub-model is further used to: Inputting the candidate sample data into the second sub-model to predict the data source of the candidate sample data; wherein the data source includes historical data and model generation; When the proportion of the predicted candidate sample data whose data source is historical data reaches a preset threshold, the predicted candidate sample data whose data source is historical data is determined as sample data.

4. The method according to claim 1, wherein The first sub-model is obtained by: Establishing an initial first sub-model; wherein the first sub-model is used to generate random noise, and generate candidate sample data using the random noise and historical data; Inputting the historical data into the initial first sub-model, and outputting candidate sample data; Inputting the candidate sample data and historical data into the second sub-model, and adjusting the parameters of the initial first sub-model based on the prediction results; When the proportion of historical data that is the source of candidate sample data and historical data predicted by the second sub-model reaches a preset threshold, the initial first sub-model is determined as the first sub-model.

5. The method according to claim 1, wherein The acquisition of the environment recognition model includes: Establishing an initial environment recognition model; wherein the initial environment recognition model includes a reward mechanism and a penalty mechanism; determining target operating data under each environment mode by establishing the mechanism and the penalty mechanism; The sample data is input into the environment recognition model for training to obtain the environment recognition model.

6. The method according to claim 1, wherein The step of inputting the historical data into a preset data generation model to obtain sample data includes: performing standardization processing on the historical data to obtain standard data; The standard data is input into a preset data generation model to obtain sample data.

7. The method according to claim 1, characterized in that The method further comprises: Get current data on how the grid is operating; determining a current environmental mode of the power grid operation mode based on the current data and reference data corresponding to each environmental mode of the power grid operation mode; The operation data of the power grid operation mode is adjusted to the target operation data corresponding to the current environmental mode for operation.

8. A device for generating a power grid operation mode, characterized in that: The device comprises: A data acquisition module is used to acquire historical data of the power grid operation mode; wherein the historical data includes environmental data and operation data; A sample generation module is configured to input the historical data into a preset data generation model to obtain sample data; wherein the data generation model includes a first sub-model and a second sub-model, wherein the first sub-model is configured to generate candidate sample data using historical data and random noise; and the second sub-model is configured to identify the data sources of the historical data and the candidate sample data and determine the sample data; The mode prediction module is used to input the sample data into a preset environment recognition model to obtain the environmental mode of the power grid operation mode and the corresponding target operation data.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images