Generation method and device of power grid regulation and control scene, electronic equipment and storage medium

Abstract

The invention discloses a power grid regulation and control scene generation method and device, electronic equipment and a storage medium, and relates to the field of smart power grids or other related technical fields, and the method comprises the steps: collecting power grid load data, and carrying out the preprocessing of the power grid load data, and obtaining the preprocessed power grid load data; performing feature extraction and coding on the preprocessed power grid load data to obtain a load feature vector; inputting the load feature vector into an unsupervised generation model, and outputting a power grid regulation and control scene through the unsupervised generation model; and calculating the similarity between the power grid regulation and control scene and a pre-constructed reference regulation and control scene, configuring a type label for the generated power grid regulation and control scene based on the similarity, and generating a target power grid regulation and control scene based on the power grid regulation and control scene and the type label. According to the invention, the technical problem of low efficiency of a power grid regulation and control scene generation method based on historical data and an expert strategy in related technologies is solved.

Description

Technical Field

[0001] The present invention relates to the field of smart grids or other related technical fields. Specifically, it relates to a method and device for generating a power grid regulation scenario, an electronic device, and a storage medium. Background Art

[0002] In the efficient operation and stability management of the power system, the generation of power grid regulation scenarios plays a crucial role. As the power system becomes increasingly complex, especially with the large-scale access of renewable energy and the increase in adjustable loads, how to accurately monitor and predict the operating state of the power grid has become a key technical challenge in power grid dispatching and demand-side management. The generation of power grid regulation scenarios not only helps to optimize power grid dispatching strategies, ensure the balance between power supply and demand, but also improves the flexibility and reliability of the power system, and has significant positive effects on the stable operation of the power market, the efficient consumption of renewable energy, and the response to sudden faults.

[0003] Specifically, the core goal of generating power grid regulation scenarios is to provide a scientific basis for power grid dispatching decisions. By simulating future load changes, power generation fluctuations, and external influencing factors (such as weather conditions, economic activities, etc.), a series of possible operating scenarios are generated to help dispatchers plan dispatching schemes in advance and ensure the safe and stable operation of the power system. In addition, the generated regulation scenarios can also be used to evaluate the rationality of power market mechanisms, optimize demand response strategies, and improve the power system's ability to respond to extreme weather and emergencies, thereby reducing power outages and improving power service quality.

[0004] In related technologies, the methods for generating power grid regulation scenarios mainly include generating based on historical data and expert strategies. This method has a strong dependence on historical data and expert experience, and due to the high-dimensionality and complexity of power system load data, the computational efficiency is often low during data processing, making it difficult to meet the requirements of real-time power grid regulation.

[0005] In view of the above problems, no effective solution has been proposed yet. Summary of the Invention

[0006] Embodiments of the present invention provide a method and device for generating a power grid regulation scenario, an electronic device, and a storage medium, so as to at least solve the technical problem of low efficiency in the method for generating a power grid regulation scenario based on historical data and expert strategies in related technologies.

[0007] According to one aspect of an embodiment of the present invention, there is provided a method for generating a power grid regulation scenario, including: collecting power grid load data, and preprocessing the power grid load data to obtain preprocessed power grid load data; extracting features and encoding the preprocessed power grid load data to obtain a load feature vector; inputting the load feature vector into an unsupervised generation model, and outputting a power grid regulation scenario through the unsupervised generation model, where the unsupervised generation model is a pre-constructed machine learning model; calculating the similarity between the power grid regulation scenario and a pre-constructed reference regulation scenario, configuring a type label for the generated power grid regulation scenario based on the similarity, and generating a target power grid regulation scenario based on the power grid regulation scenario and the type label.

[0008] Further, the step of collecting power grid load data includes: accessing a multi-source data platform, and collecting historical load data at different time scales from the multi-source data platform; collecting real-time load data of different types of loads through real-time monitoring of the power grid system, where the load types include at least one of the following: industrial load, commercial load, residential load, electric vehicle charging load; obtaining power grid load data based on the historical load data and the real-time load data.

[0009] Further, the step of preprocessing the power grid load data to obtain preprocessed load data includes: performing data cleaning on the power grid load data to obtain cleaned power grid load data, where data cleaning includes at least one of the following: missing value processing, outlier processing, noise removal; performing standardization and normalization processing on the cleaned power grid load data to obtain preprocessed power grid load data.

[0010] Further, the step of preprocessing the power grid load data to obtain preprocessed load data further includes: inputting the power grid load data into a data generation model, and outputting synthetic load data similar to the power grid load data, where the data generation model is a pre-constructed machine learning model for generating similar data based on input data; expanding the power grid load data through data transformation techniques to obtain expanded load data; obtaining the preprocessed power grid load data based on the synthetic load data and the expanded load data.

[0011] Further, the steps of extracting features from the preprocessed power grid load data include: extracting time series features from the power grid load data, where the time series features include at least one of the following: mean value, variance, peak value, valley value, and fluctuation feature; extracting load features of different types of users from the power grid load data, where the user types include: industrial users, commercial users, and residential users, and the load features include at least one of the following: user power consumption distribution feature, user power consumption time distribution feature; combining the external environment data in the power grid load data to extract the influence features of the external environment on the load; extracting the regulation potential features of adjustable loads from the power grid load data, where the adjustable loads include at least one of the following: industrial load, temperature control load, electric vehicle load, energy storage load, and the regulation potential features are used to characterize the regulation ability and regulation range of the adjustable loads.

[0012] Further, the steps of constructing the unsupervised generation model include: obtaining historical power grid load data in a historical time period, and constructing a training sample based on the historical power grid load data; selecting a target machine learning model as the initial architecture of the unsupervised generation model, determining the hyperparameters and optimization objectives of the target machine learning model, and obtaining the initial unsupervised generation model; based on the training sample, iteratively training the model through a forward diffusion process and a reverse generation process until the number of iterations reaches the iteration number threshold, and obtaining the trained unsupervised generation model.

[0013] Further, the steps of iteratively training the model through a forward diffusion process and a reverse generation process based on the training sample include: in the forward diffusion process, adding Gaussian noise to the training sample at each iteration, and inputting the training sample with added noise into the initial unsupervised generation model for learning until the training sample is transformed into a random noise vector, and stopping the forward diffusion process; in the reverse generation process, training the initial unsupervised generation model to denoise the training sample at each iteration until the training sample is restored to the original data distribution, and stopping the reverse generation process; where, in each round of forward diffusion process and reverse generation process, the hyperparameters of the initial unsupervised generation model are updated through the optimization objective, and the next round of forward diffusion process and reverse generation process is executed based on the updated hyperparameters until the initial unsupervised generation model generates a power grid regulation scenario that meets the preset conditions.

[0014] According to another aspect of the embodiments of the present invention, there is also provided a device for generating a power grid regulation scenario, including: an acquisition unit, configured to acquire power grid load data and perform preprocessing on the power grid load data to obtain preprocessed power grid load data; an extraction unit, configured to perform feature extraction and encoding on the preprocessed power grid load data to obtain a load feature vector; an output unit, configured to input the load feature vector into an unsupervised generation model, and output a power grid regulation scenario through the unsupervised generation model, where the unsupervised generation model is a pre-constructed machine learning model; a generation unit, configured to calculate the similarity between the power grid regulation scenario and a pre-constructed reference regulation scenario, configure a type label for the generated power grid regulation scenario based on the similarity, and generate a target power grid regulation scenario based on the power grid regulation scenario and the type label.

[0015] Further, the acquisition unit includes: a first acquisition module, configured to access a multi-source data platform and acquire historical load data of different time scales from the multi-source data platform; a second acquisition module, configured to acquire real-time load data of different types of loads by real-time monitoring a power grid system, where the load types include at least one of the following: industrial load, commercial load, residential load, electric vehicle charging load; a first obtaining module, configured to obtain power grid load data based on the historical load data and the real-time load data.

[0016] Further, the acquisition unit further includes: a first cleaning module, configured to perform data cleaning on the power grid load data to obtain cleaned power grid load data, where the data cleaning includes at least one of the following: missing value processing, outlier processing, noise removal; a first processing module, configured to perform standardization and normalization processing on the cleaned power grid load data to obtain preprocessed power grid load data.

[0017] Further, the acquisition unit further includes: a first output module, configured to input the power grid load data into a data generation model and output synthetic load data similar to the power grid load data, where the data generation model is a pre-constructed machine learning model for generating similar data based on input data; a first expansion module, configured to expand the power grid load data through data transformation techniques to obtain expanded load data; a second obtaining module, configured to obtain the preprocessed power grid load data based on the synthetic load data and the expanded load data.

[0018] Further, the extraction unit includes: a first extraction module, configured to extract time series features from the grid load data, where the time series features include at least one of the following: mean, variance, peak value, valley value, and fluctuation feature; a second extraction module, configured to extract load features of different types of users from the grid load data, where the user types include: industrial users, commercial users, and residential users, and the load features include at least one of the following: user power consumption distribution feature, user power consumption time distribution feature; a third extraction module, configured to extract the influence features of the external environment on the load by combining the external environment data in the grid load data; a fourth extraction module, configured to extract the regulation potential features of the adjustable load from the grid load data, where the adjustable load includes at least one of the following: industrial load, temperature control load, electric vehicle load, energy storage load, and the regulation potential features are used to characterize the regulation ability and regulation range of the adjustable load.

[0019] Further, the generating device for the grid regulation scenario further includes a construction unit, and the construction unit includes: a first construction module, configured to obtain historical grid load data within a historical time period and construct a training sample based on the historical grid load data; a first determination module, configured to select a target machine learning model as the initial architecture of the unsupervised generation model, determine the hyperparameters and optimization objectives of the target machine learning model, and obtain an initial unsupervised generation model; a first training module, configured to iteratively train the model based on the training sample through a forward diffusion process and a reverse generation process until the number of iterations reaches an iteration number threshold, and obtain the trained unsupervised generation model.

[0020] Further, the first training module includes: a first training sub-module, configured to add Gaussian noise to the training sample at each iteration during the forward diffusion process, and input the training sample with added noise into the initial unsupervised generation model for learning until the training sample is transformed into a random noise vector, and stop the forward diffusion process; a second training sub-module, configured to train the initial unsupervised generation model to denoise the training sample at each iteration during the reverse generation process until the training sample is restored to the original data distribution, and stop the reverse generation process; where, in each round of the forward diffusion process and the reverse generation process, the hyperparameters of the initial unsupervised generation model are updated through the optimization objective, and the next round of the forward diffusion process and the reverse generation process are executed based on the updated hyperparameters until the initial unsupervised generation model generates a grid regulation scenario that meets the preset conditions.

[0021] According to another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium, which includes a stored computer program. When the computer program runs, it controls the device where the computer-readable storage medium is located to execute the generation method of any one of the above grid regulation scenarios.

[0022] According to another aspect of the embodiments of the present invention, there is also provided an electronic device, including one or more processors and a memory. The memory is used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the generation method of any one of the above grid regulation scenarios.

[0023] In this application, through the following steps: first, collect grid load data, preprocess the grid load data to obtain preprocessed grid load data, perform feature extraction and encoding on the preprocessed grid load data to obtain a load feature vector, then input the load feature vector into an unsupervised generation model, and output a grid regulation scenario through the unsupervised generation model. The unsupervised generation model is a pre-constructed machine learning model. Finally, calculate the similarity between the grid regulation scenario and a pre-constructed reference regulation scenario, configure a type label for the generated grid regulation scenario based on the similarity, and generate a target grid regulation scenario based on the grid regulation scenario and the type label.

[0024] In this application, collect grid load data, where the grid load data includes historical data and real-time data, preprocess and perform feature extraction on the grid load data to ensure data quality. For the extracted load feature vector, generate a grid regulation scenario based on an unsupervised generation model. Finally, calculate the similarity between the generated grid regulation scenario and a pre-constructed typical scenario to clarify the scenario type to which the generated grid regulation scenario belongs, thereby generating a target grid regulation scenario, achieving the purpose of automatically generating accurate grid regulation scenarios, meeting the real-time requirements of grid regulation, obtaining the technical effect of improving the generation efficiency of grid regulation scenarios, and further solving the technical problem of low efficiency in the method for generating grid regulation scenarios based on historical data and expert strategies in the related art. Description of the Drawings

[0025] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0026] Figure 1 is a flowchart of an alternative method for generating a grid regulation scenario according to an embodiment of the present invention;

[0027] Figure 2It is a schematic diagram of an optional generation process of a power grid regulation scenario according to an embodiment of the present invention;

[0028] Figure 3 It is a schematic diagram of an optional generation device of a power grid regulation scenario according to an embodiment of the present invention;

[0029] Figure 4 It is a hardware structure block diagram of an electronic device (or mobile device) for executing a method for generating a power grid regulation scenario according to an embodiment of the present invention. Detailed implementation manners

[0030] In order to enable those skilled in the art of the present technology to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0031] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0032] It should be noted that the method and device for generating a power grid regulation scenario in the present application can be used in the field of smart grids. In the case of generating a power grid regulation scenario based on artificial intelligence, it can also be used in any field other than the field of smart grids. In the case of generating a power grid regulation scenario based on artificial intelligence, the application field of the method and device for generating a power grid regulation scenario in the present application is not limited.

[0033] It should be noted that the relevant information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are information and data authorized by the user or fully authorized by all parties. Moreover, the processing of the relevant data, such as collection, storage, use, processing, transmission, provision, disclosure, and application, complies with the relevant laws, regulations, and standards in the relevant regions, adopts necessary confidentiality measures, does not violate public order and good customs, and provides corresponding operation entrances for users to choose to authorize or refuse. For example, there is an interface between this system and relevant users or institutions. Before obtaining relevant information, a request for acquisition needs to be sent to the aforementioned users or institutions through the interface, and after receiving the consent information feedback from the aforementioned users or institutions, the relevant information is obtained.

[0034] It should be noted that when collecting and analyzing customer information in this application, corresponding operation entrances are provided for users to choose to agree or refuse the results of automated decision-making; if the user chooses to refuse, the expert decision-making process will be entered.

[0035] The following embodiments of the present invention can be applied to various power grid regulation scenario generation systems / applications / devices. The present invention proposes a method for generating power grid regulation scenarios based on an unsupervised generation model, which solves problems such as low computational efficiency when dealing with massive high-dimensional load data by existing algorithms, the need to occupy a large amount of computing resources, and poor effects of generated regulation scenarios when the data quality is poor, and provides basic support for subsequent power grid regulation optimization, new energy consumption, source-network-load-storage collaborative interaction, and other links.

[0036] The present invention will be described in detail below in conjunction with each embodiment.

[0037] Embodiment 1

[0038] According to an embodiment of the present invention, an embodiment of a method for generating a power grid regulation scenario is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0039] Figure 1 is a flowchart of an optional method for generating a power grid regulation scenario according to an embodiment of the present invention, as Figure 1 shown, the method includes the following steps:

[0040] Step S101, collect power grid load data and preprocess the power grid load data to obtain preprocessed power grid load data.

[0041] The power system needs to balance power production and consumption in real time. With the wide application of renewable energy sources such as solar and wind energy, the power supply of the power grid has become fluctuating. At the same time, the power demands of adjustable loads such as industrial, commercial, residential, and electric vehicles are also becoming increasingly diverse and have a certain degree of flexibility. Generating power grid regulation scenarios can simulate power demand patterns under different conditions (such as weather changes and economic activity levels), helping power grid dispatchers predict load changes, formulate reasonable power generation plans and demand-side management strategies, ensure the balance of power supply and demand, avoid power overage or shortage, reduce power waste, and improve system efficiency.

[0042] Traditional power dispatch strategies often rely on fixed load forecasting. However, in modern power systems, the adjustability of loads brings new opportunities for power dispatch. By generating regulation scenarios with adjustable loads, it is possible to predict the responses of different load types (such as temperature control devices and energy storage systems) under specific conditions, thereby optimizing the dispatch strategy, achieving more flexible and accurate power dispatch, enabling the power system to better cope with load peaks, reducing dispatch costs, and improving the economy and stability of power system operation.

[0043] The generated power grid regulation scenarios can be used to evaluate the rationality of power market mechanisms and the effectiveness of demand response strategies. For example, by simulating the responses of users under different electricity price mechanisms, the electricity price design can be optimized to encourage users to reduce electricity consumption during peak load periods or increase electricity consumption during off-peak periods, thereby balancing the power grid load, improving the flexibility and economy of the power system. At the same time, the demand response strategy based on the regulation scenario can assist in managing power demand, reducing the operation cost of the power grid, and improving the efficiency of the power market.

[0044] Regarding the generation of regulation scenarios with adjustable loads, the embodiments of the present invention aim to solve problems such as low generation efficiency, high data quality requirements, and insufficient scenario diversity in the current technology, and ensure that the generated scenarios can play an important role in actual power grid regulation.

[0045] In the above step S101, in order to generate a power grid regulation scenario according to the real-time operation situation of the power grid system, it is first necessary to collect power grid load data from a multi-source data system. The power grid load data includes the historical operation data of the power grid system within the target time period and the real-time operation data of the power grid system at the current moment. Through the integrated analysis of historical data and real-time data, the dependence on historical data is reduced, the real-time performance and effectiveness of the scenario generation result are improved, and the accuracy of the generation result is guaranteed.

[0046] Furthermore, the collected power grid load data may contain noise, missing values, and anomalies. To improve the accuracy of subsequent generation results and enhance data quality, it is necessary to preprocess the collected data. The preprocessing includes data cleaning, data standardization, and normalization. In some scenarios with scarce data, it is also necessary to expand the power grid load data to enrich the dataset and improve data quality.

[0047] Furthermore, the steps for collecting power grid load data include: accessing a multi-source data platform to collect historical load data at different time scales from the multi-source data platform; collecting real-time load data of different types of loads through real-time monitoring of the power grid system, where the load types include at least one of the following: industrial load, commercial load, residential load, electric vehicle charging load; obtaining power grid load data based on the historical load data and the real-time load data.

[0048] In some embodiments, the power grid load data is obtained from the real-time operating status of a multi-source data platform and the power grid system. Specifically, first, access multi-source data platforms such as the power grid dispatching center, user-side smart meters, and distributed energy management systems to ensure the diversity and comprehensiveness of the data. These data platforms not only provide historical load data but also cover the electricity consumption at different time scales (such as minute-level, hour-level, daily-level), ensuring that the data covers different time periods (minutes, hours, days), which helps the model learn the short-term fluctuations and long-term trends of the load, obtain the dynamic change rules of the load, and improve the accuracy and practicality of scenario generation. Based on accessing the multi-source data platform, collect historical load data of the power grid at different time scales. The historical data contains the actual performance of the power grid load under various conditions in the past period and is an important basis for the model to analyze the load pattern and predict the future load trend. The diversity of time scales can cover the operating characteristics of the power system in different time cycles, making the generated scenarios more realistic and covering various operating states.

[0049] Through the data collection steps, the embodiments of the present invention can ensure that the data read by the power grid regulation scenario generation model is both comprehensive and accurate, not only covering historical load patterns but also reflecting the current load dynamics, making the generated scenarios closer to the actual power grid operating status and having high representativeness and practicality.

[0050] Furthermore, by real-time monitoring the power grid system, continuously collect real-time data of different types of loads (such as industrial, commercial, residential, electric vehicle charging) to capture the dynamic change trend of the load. The acquisition of real-time data helps the model update and learn, ensuring that the scenario generation can reflect the latest power grid operating status, improving the timeliness and flexibility of the scenario, and being crucial for coping with rapidly changing power demands.

[0051] Further, the steps for preprocessing the grid load data to obtain the preprocessed load data include: performing data cleaning on the grid load data to obtain the cleaned grid load data, where the data cleaning includes at least one of the following: missing value processing, outlier processing, and noise removal; performing standardization and normalization processing on the cleaned grid load data to obtain the preprocessed grid load data.

[0052] In some embodiments, to further ensure the quality of the input data, it is necessary to preprocess the grid load data. The preprocessing first includes data cleaning, and the data cleaning is further divided into missing value processing, outlier processing, and noise removal. The missing values in the grid load data may stem from equipment failures, network interruptions, or data recording errors. For missing values, interpolation methods (such as linear interpolation, spline interpolation) or machine learning-based methods (such as K-nearest neighbor imputation, time series-based prediction imputation) can be used to fill in the missing values. Outliers, that is, data points outside the normal range, may be caused by extreme weather, power system failures, or abnormal data records. To prevent outliers from interfering with the generated results, statistical methods or machine learning algorithms (such as isolation forest) can be used for outlier detection. Once outliers are detected, median replacement, boundary value replacement, or model-based predicted values can be used to correct the outliers to ensure the reliability of the data. In addition, the grid load data may also contain random noise, which comes from the accuracy limitations of measurement devices, data transmission errors, or minor fluctuations in the power system. Filtering techniques (such as moving average filtering, wavelet transform) can be used to remove the noise, improve the smoothness of the data, ensure that the model can accurately capture the patterns of load changes, and avoid the negative impact of noise on the generated results.

[0053] Further, the standardization processing converts the grid load data into a standard normal distribution with a mean of 0 and a variance of 1. Thereby eliminating the dimensional differences in the data, enabling the model to treat each feature fairly during the identification and analysis process, and improving the calculation speed and accuracy. The normalization processing ensures that all feature value ranges are consistent by using min-max scaling or converting the data into values within the range of 0-1. This helps the model handle non-linear relationships and accelerates the model calculation process.

[0054] Further, the steps for preprocessing the grid load data to obtain the preprocessed load data also include: inputting the grid load data into a data generation model to output synthetic load data similar to the grid load data, where the data generation model is a pre-constructed machine learning model for generating similar data based on the input data; expanding the grid load data through data transformation techniques to obtain expanded load data; and obtaining the preprocessed load data based on the synthetic load data and the expanded load data.

[0055] Step S102: Extract features and encode the preprocessed power grid load data to obtain a load feature vector.

[0056] In the above Step S102, after obtaining the preprocessed power grid load data, in order to convert the text data into data that can be processed by the model, it is necessary to extract features and perform vector encoding on the original data, so as to convert the text data into feature vectors. The construction of the load feature vector provides high-quality input data for the generation of power grid regulation scenarios based on unsupervised generation models, ensuring that the model can learn the multi-dimensional complex characteristics of power grid load data and generate more diverse and high-quality power grid regulation scenarios.

[0057] Further, the steps for extracting features from the preprocessed power grid load data include: extracting time series features from the power grid load data, where the time series features include at least one of the following: mean, variance, peak value, valley value, and fluctuation feature; extracting load features of different types of users from the power grid load data, where the user types include: industrial users, commercial users, and residential users, and the load features include at least one of the following: user power consumption distribution feature, user power consumption time distribution feature; extracting the influence features of the external environment on the load by combining the external environment data in the power grid load data; extracting the regulation potential features of adjustable loads from the power grid load data, where the adjustable loads include at least one of the following: industrial load, temperature control load, electric vehicle load, energy storage load, and the regulation potential features are used to characterize the regulation ability and regulation range of the adjustable loads.

[0058] Specifically, the extracted features include: time series features, user type features, external environment features, and adjustable load regulation potential features. Extracting time series features from the preprocessed load data includes short-term fluctuations, seasonal variations, trends, etc. of the load. Time series features can capture the patterns of load changes over time and are crucial for predicting peak and valley load periods and evaluating the output characteristics of renewable energy. According to the user types (industrial users, commercial users, residential users, electric vehicle charging users, etc.) of the data sources, extract load consumption features, such as average power consumption, power consumption peak value, power consumption pattern, etc. User type features help to understand the demand patterns of different types of loads and play an important role in formulating refined demand response strategies. Combine external data sources (such as meteorological data, economic activity data) to extract environmental impact features, such as temperature, humidity, holiday information, economic indicators, etc. These features can quantify the impact of the external environment on the load and are of great significance for evaluating the external factors affecting load changes. For adjustable loads in the power grid system, such as electric vehicle charging, user-side energy storage, etc., extract their regulation ability, response speed, regulation range and other features, so as to consider the flexibility and adjustability of the load in scenario generation, which is crucial for optimizing the scheduling plan, balancing supply and demand fluctuations, and increasing the participation of adjustable loads.

[0059] Further, after obtaining the load characteristics, the extracted characteristics are converted into a numerical vector representation for easy computer processing and model training. The vectorization process ensures the digitization and structuring of the characteristics, providing a standard input format for the learning of the unsupervised generation model.

[0060] In some embodiments, deep learning techniques can be used to encode the load characteristics to enhance the semantic information of the characteristics and capture the dynamic change rules of the load, which is crucial for generating scenarios closer to the actual situation.

[0061] Step S103: Input the load feature vector into the unsupervised generation model, and output the power grid regulation scenario through the unsupervised generation model.

[0062] In this embodiment, an unsupervised generation model is used to generate the required power grid regulation scenario. Among them, the unsupervised generation model is a pre-constructed machine learning model that does not require label information but automatically learns the internal structure and distribution of the data from the input data. In the field of power systems, the unsupervised generation model can capture the probability distribution of complex load data and generate regulation scenarios similar to the actual power grid operation state.

[0063] In the above step S103, the load feature vectors extracted and encoded in the above steps are used as input data. These vectors contain time series features, user type features, external environment features, and adjustable load regulation potential features, and can comprehensively describe various aspects of the power grid load. The unsupervised generation model obtained through pre-training captures the load distribution and operation conditions in the power grid scenario and establishes the power grid regulation scenario.

[0064] Further, the steps of constructing the unsupervised generation model include: obtaining the historical power grid load data within a historical time period, constructing training samples based on the historical power grid load data; selecting a target machine learning model as the initial architecture of the unsupervised generation model, determining the hyperparameters and optimization objectives of the target machine learning model to obtain the initial unsupervised generation model; based on the training samples, iteratively train the model through the forward diffusion process and the reverse generation process until the number of iterations reaches the iteration number threshold to obtain the trained unsupervised generation model.

[0065] It should be noted that when constructing an unsupervised model, a large amount of historical power grid load data within a historical time period is used as the training set to iteratively train the initial machine learning model, resulting in an unsupervised generation model that can generate power grid regulation scenarios. The specific model construction process includes: collecting historical power grid load data of the power grid system within a historical time period. This historical time period should be representative and cover the peak, trough, and transition periods of the power grid load to ensure that the model can learn the complete change pattern of the load. The historical power grid load data includes the load amounts of the power grid system at different time scales (such as minute-level, hour-level, daily-level), as well as the power consumption information of different types of loads (such as industrial, commercial, residential, electric vehicle charging). Then, preprocess the training samples, perform feature extraction and vector encoding to obtain the training samples.

[0066] Furthermore, according to the characteristics of the power grid load data, select a machine learning model as the initial architecture of the unsupervised generation model. Determine the hyperparameters of the model, such as the learning rate, batch size, noise steps, etc. These parameters will affect the training efficiency and generation effect of the model. Set the optimization objective of the model, such as minimizing the distance between the generated data and the real data (usually using maximum likelihood estimation) to ensure that the model can accurately learn and reconstruct the real distribution of the data.

[0067] When performing iterative training on the model, set multiple rounds of iterative training. In each round of iterative training process, gradually train the model to learn the complex characteristics of the historical power grid load data in the training samples through the forward diffusion process and the reverse generation process to generate power grid regulation scenarios. In each round of the model's iterative process, adjust the model parameters through the optimization algorithm until the predetermined iteration number threshold is reached. The iteration number threshold is an important parameter for model training, used to control the depth and complexity of training to ensure that the model can both learn the complex characteristics of the data and avoid overfitting.

[0068] Furthermore, the steps of iteratively training the model through the forward diffusion process and the reverse generation process based on the training samples include: in the forward diffusion process, add Gaussian noise to the training samples at each iteration and input the training samples with added noise into the initial unsupervised generation model for learning until the training samples are transformed into a random noise vector and stop the forward diffusion process; in the reverse generation process, train the initial unsupervised generation model to denoise the training samples at each iteration until the training samples are restored to the original data distribution and stop the reverse generation process; where, in each round of the forward diffusion process and the reverse generation process, update the hyperparameters of the initial unsupervised generation model through the optimization objective and perform the next round of the forward diffusion process and the reverse generation process based on the updated hyperparameters until the initial unsupervised generation model generates a power grid regulation scenario that meets the preset conditions.

[0069] In some embodiments, the purpose of the forward diffusion process is to gradually transform the training samples into a noise distribution. The Gaussian noise added in this process helps the model learn the distribution of the data at different noise levels, thereby enhancing the robustness of the model so that it can handle various forms of noise and incompleteness. During forward diffusion, first, the original training samples are used as input, and a certain amount of Gaussian noise is gradually added in each iteration. The addition of noise can be controlled by adjusting the number of noise steps and the noise level to ensure that the diffusion process of the data is both stable and sufficient. As the number of iterations increases, the training samples are gradually transformed into a random noise vector, which simulates the evolution of the data from clear to blurred and helps train the denoising ability and data recovery ability of the model.

[0070] Furthermore, the reverse generation process is a crucial step for the model to learn to recover the true distribution of the data. By gradually removing the noise added during the forward diffusion process, the model can learn and understand the internal structure and patterns of the data, thus acquiring the ability to generate scenes similar to the real data. After the training samples are transformed into a random noise vector, the model starts to execute the reverse generation process and gradually denoises. In each iteration, the model attempts to restore the true distribution of the data by adjusting its parameters, which usually relies on maximum likelihood estimation or minimizing the difference between the prediction and the real data. As the denoising process progresses, the training samples gradually return to the original data distribution, and the learning effect of the model is optimized and improved in this process.

[0071] In addition, in each round of the forward diffusion and reverse generation processes, the hyperparameters of the model also need to be updated through optimization objectives (such as maximum likelihood estimation, minimum loss function) to ensure that the generation ability of the model is continuously improved, and the finally generated scenes can meet the preset quality and diversity conditions. The update of the hyperparameters is usually based on the performance of the model during the forward diffusion and reverse generation processes. Through optimization algorithms such as backpropagation and gradient descent, key parameters such as the learning rate, batch size, and number of noise steps are adjusted so that the model can better fit the data distribution and produce high-quality outputs during the generation process. After each round of iteration, according to the generation effect of the model and the optimization objective, the hyperparameters are adjusted automatically or manually, and then the next round of forward diffusion and reverse generation processes are continued based on the updated hyperparameters until the scenes generated by the model meet the preset conditions, including but not limited to the diversity, representativeness of the scenes, and the similarity to the real data.

[0072] Step S104, calculate the similarity between the power grid regulation scenario and the pre-constructed reference regulation scenario, configure a type label for the generated power grid regulation scenario based on the similarity, and generate a target power grid regulation scenario based on the power grid regulation scenario and the type label.

[0073] In some embodiments, after generating the power grid regulation scenario through an unsupervised generation model, it is necessary to determine the scenario type of the power grid regulation scenario. The scenario type is determined according to the adjustable load type and the regulation scenario. The adjustable load is divided into four categories: industrial load, temperature control load, electric vehicle load, and energy storage load. The power grid regulation scenario is divided into three major categories: low load period, normal period, and peak load period. Then, considering the situation where four different types of adjustable loads are dominant, the scenario type is divided into industrial load - low load scenario, industrial load - normal scenario, industrial load - peak load scenario, temperature control load - low load scenario, temperature control load - normal scenario, temperature control load - peak load scenario, electric vehicle - low load scenario, electric vehicle - normal scenario, electric vehicle - peak load scenario, energy storage load - low load scenario, energy storage load - normal scenario, and energy storage load - peak load scenario.

[0074] In the above step S104, the scenario type to which the output power grid regulation scenario belongs is determined by calculating the similarity between the output power grid regulation scenario of the model and the preset reference regulation scenario, obtaining a type label, and generating the final target power grid regulation scenario based on the power grid regulation scenario and the type label.

[0075] In some optional embodiments, after generating the target power grid regulation scenario, the generated target power grid regulation scenario is applied to actual power grid regulation to test the effectiveness and practicability of the target power grid regulation scenario. The specific steps include: First, input the generated typical regulation scenario into the power grid dispatching system to simulate the power grid behavior under different operating conditions and obtain the prediction result of the generated scenario; Second, evaluate the accuracy and reliability of the scenario by comparing the actual operation data with the prediction result of the generated scenario to obtain an evaluation result; Finally, optimize the regulation strategy (such as demand response, load dispatching) according to the evaluation result to verify the practicability and effect of the scenario in actual application, providing a scientific basis and data support for the efficient, reliable, green, and intelligent operation of the power grid.

[0076] Through the above steps, first collect the power grid load data, preprocess the power grid load data to obtain the preprocessed power grid load data, extract and encode the features of the preprocessed power grid load data to obtain the load feature vector, then input the load feature vector into the unsupervised generation model, and output the power grid regulation scenario through the unsupervised generation model. Among them, the unsupervised generation model is a pre - constructed machine learning model. Finally, calculate the similarity between the power grid regulation scenario and the pre - constructed reference regulation scenario, configure a type label for the generated power grid regulation scenario based on the similarity, and generate the target power grid regulation scenario based on the power grid regulation scenario and the type label.

[0077] In this embodiment, power grid load data is collected. The power grid load data includes historical data and real-time data. Then, the power grid load data is preprocessed and feature extracted to ensure data quality. For the extracted load feature vectors, an unsupervised generation model is used to generate power grid regulation scenarios. Finally, the similarity between the generated power grid regulation scenarios and the pre-constructed typical scenarios is calculated to clarify the scenario type to which the generated power grid regulation scenarios belong, thereby generating target power grid regulation scenarios, achieving the purpose of automatically generating accurate power grid regulation scenarios, meeting the real-time requirements of power grid regulation, obtaining the technical effect of improving the generation efficiency of power grid regulation scenarios, and further solving the technical problem of low efficiency in the method for generating power grid regulation scenarios based on historical data and expert strategies in the related art.

[0078] The following is a detailed description in combination with another optional specific implementation manner.

[0079] Figure 2 It is a schematic diagram of an optional generation process of a power grid regulation scenario according to an embodiment of the present invention, as Figure 2 shown. The generation process of the power grid regulation scenario includes:

[0080] Step 1: Classify and define the adjustable loads and typical regulation scenarios in the power grid;

[0081] Adjustable loads refer to loads that can adjust the power consumption or power consumption time according to the power grid demand or external instructions, including interruptible loads (such as industrial loads), shiftable loads (such as electric vehicle charging), adjustable loads (such as air conditioners), and user-side energy storage, etc. Their core characteristics are flexibility and controllability. In a new power system, classifying adjustable loads is of great significance: First, it can improve the flexibility and accuracy of power grid regulation. By accurately identifying adjustable load resources and formulating targeted strategies, the dispatching decision-making can be optimized; Second, the classification supports the access of a high proportion of renewable energy. By using adjustable loads to balance the supply and demand fluctuations, the renewable energy consumption capacity can be improved; In addition, the classification enhances the stability and reliability of the power grid, helps identify risks and supports fault recovery, and at the same time promotes the development of the power market and demand response, providing data support for market mechanism design and demand response optimization. The classification of adjustable loads also promotes the construction of intelligent and digital power grids, provides support for decision-making based on big data and artificial intelligence, and helps with energy transformation and low-carbon development, optimizes renewable energy consumption strategies, and promotes the application of low-carbon technologies. Finally, the classification improves the research efficiency and the reliability of engineering applications, laying an important foundation for the efficient, reliable, green, and intelligent operation of the new power system.

[0082] In a new power system, classifying and defining typical power grid regulation scenarios is of great significance. First, it can improve the accuracy and efficiency of power grid regulation by formulating precise strategies for different scenarios and quickly responding to operating requirements. Second, classification and definition support the high-proportion access of renewable energy, optimize the supply-demand balance and enhance the consumption capacity of renewable energy, while enhancing the stability and reliability of the power grid, helping to identify risks and support fault recovery. In addition, the classification of typical regulation scenarios promotes the development of the power market and demand response, provides data support for market mechanism design and demand response optimization, and promotes the construction of intelligent and digital power grids, providing structured data support for decision-making based on big data and artificial intelligence.

[0083] In this embodiment, adjustable loads can be divided into four categories: industrial loads, temperature control loads, electric vehicle loads, and energy storage loads; power grid regulation scenarios are divided into three major categories: low-load periods, normal periods, and high-load periods. Then, considering the cases where four different types of adjustable loads are dominant, twelve typical regulation scenarios are defined.

[0084] Step 2: Collect historical load data of the power grid at different time scales and real-time monitoring data of different types of loads.

[0085] First, obtain historical load data from multi-source data platforms such as power grid dispatch centers, user-side smart meters, and distributed energy management systems, covering time scales of minutes, hours, and days to ensure the comprehensiveness and continuity of the data. Second, collect real-time data of different types of loads (such as industrial loads, commercial loads, residential loads, electric vehicle charging loads, etc.) by real-time monitoring the power system to help capture the dynamic change characteristics of the loads.

[0086] Step 3: Clean the data, handle missing values, outliers, and remove noise.

[0087] After obtaining the data, it is necessary to clean the data, handle missing values, outliers, and remove noise. Specifically, first, identify and delete duplicate or invalid data through data cleaning techniques to ensure the integrity of the data. Second, use interpolation methods (such as linear interpolation or spline interpolation) or machine learning methods (such as KNN imputation) to fill in missing values to ensure the continuity of the data. Then, detect and correct outliers through statistical methods (such as the 3σ principle) or machine learning algorithms (such as Isolation Forest) to avoid interference with model training. Finally, use filtering techniques (such as moving average filtering or wavelet transform) to remove noise in the data, improve the smoothness and quality of the data, and provide a clean and reliable data basis for subsequent model training and scenario generation.

[0088] Step 4: Standardize and normalize the data.

[0089] Normalize and standardize the data to improve the convergence speed of the model and avoid certain features from dominating the model training due to different dimensions. Specifically, it includes: First, standardize the collected load data, usually using the Z-score standardization method to convert the data into a standard normal distribution with a mean of 0 and a variance of 1. Second, normalize the data, using the min-max normalization method to scale the data to a specific range (such as [0,1]). Finally, through standardization and normalization, eliminate the dimensional differences between different features, ensure that each feature has the same weight in model training, thereby accelerating model convergence and improving training efficiency, and providing high-quality data input for subsequent scenario generation.

[0090] Step Five, extract features such as time series, user type, external environment, and adjustable potential from the load data;

[0091] First, extract key features (such as mean, variance, peak value, valley value, volatility, etc.) from the time series data to capture the dynamic change law of the load. Second, extract features such as electricity consumption and electricity consumption time distribution according to the user type (such as industrial, commercial, residential) to distinguish the load characteristics of different types of users. Then, combine the external environment data (such as temperature, humidity, holiday information) to extract features that have a significant impact on the load, such as the impact of temperature on air conditioning load. Finally, extract the regulation potential features (such as interruptible capacity, transferable capacity) for adjustable loads (such as electric vehicles, user-side energy storage) to provide multi-dimensional feature support for subsequent scenario generation and ensure the accuracy and practicality of the generated scenarios.

[0092] Step Six, when there is a shortage of data in certain scenarios, synthesize and augment the data;

[0093] Use an unsupervised generation model to generate synthetic data highly similar to the real load data to fill the data gap and enhance the diversity of the dataset. Second, augment the existing data through data transformation techniques (such as translation, scaling, adding noise) to generate more diverse load scenarios. Finally, combine the synthesized and augmented data with the original data to form a more comprehensive and representative dataset, providing sufficient data support for model training, thereby improving the accuracy and generalization ability of the generated scenarios and ensuring the generation of high-quality regulation scenarios even in the case of data scarcity.

[0094] Step Seven, store and manage the processed data;

[0095] First, store the data after cleaning, standardization, feature extraction, and synthetic augmentation in a high-performance database or distributed file system to ensure data security and accessibility. Secondly, establish a data management mechanism to classify, index, and control the version of the data for quick retrieval and update. Finally, achieve data traceability and reusability through a data management platform to ensure the efficient utilization of data in subsequent model training and scenario generation, providing reliable data support for power grid regulation.

[0096] Step eight, construct an unsupervised generation model with missing model architecture and model parameters.

[0097] Select a suitable unsupervised generation model architecture and design the network structure of the model (such as U-Net) to capture the complex characteristics of load data. Secondly, determine the hyperparameters of the model (such as the number of noise steps, learning rate, batch size) and the optimization objective (such as maximum likelihood estimation) to ensure that the model can efficiently learn the probability distribution of load data. Finally, determine the best parameter combination through experiments and tuning to lay the foundation for subsequent model training and scenario generation, ensuring the high quality and diversity of the generated scenarios.

[0098] Step nine, conduct model training through the iteration of the forward diffusion process and the reverse generation model.

[0099] First, in the forward diffusion process, gradually add Gaussian noise to the load data to transform the data from the real distribution to the noise distribution. Secondly, in the reverse generation process, train the model to recover the original data distribution from the noise distribution by gradually denoising, and optimize the model parameters using maximum likelihood estimation. Finally, through multiple iterations of training, enable the model to accurately learn the complex characteristics of load data, generate high-quality and diverse regulation scenarios, and provide reliable data support for power grid regulation.

[0100] Step ten, generate power grid regulation scenarios through an unsupervised generation model and determine the scenario type by comparing with typical scenarios.

[0101] First, use the trained unsupervised generation model to generate diverse load scenarios covering different operating conditions and boundary situations. Secondly, compare the generated scenarios with predefined typical regulation scenarios (such as peak load scenarios, valley load scenarios), and determine the type of the generated scenarios through similarity metrics (such as Euclidean distance, cosine similarity) or clustering analysis. Finally, classify and label according to the scenario type to provide structured scenario support for power grid regulation, facilitating the efficient, reliable, green, and intelligent operation of the power grid.

[0102] Step eleven, apply the generated scenarios to actual power grid regulation to test the effectiveness and practicality of the scenarios.

[0103] First, input the generated regulation scenarios into the power grid dispatching system to simulate the power grid behavior under different operating conditions and obtain the prediction results of the generated scenarios. Secondly, evaluate the accuracy and reliability of the scenarios by comparing the actual operation data with the prediction results of the generated scenarios. Finally, optimize the regulation strategies (such as demand response, load dispatching) according to the evaluation results, verify the practicality and effectiveness of the scenarios in actual applications, and provide scientific basis and data support for the efficient, reliable, green, and intelligent operation of the power grid.

[0104] Step Twelve: Adjust the parameters of the unsupervised generation model according to the verification results.

[0105] First, analyze the application effects of the generated scenarios in actual power grid regulation, and identify the deficiencies of the model in terms of accuracy, diversity, or real-time performance. Secondly, adjust the key parameters of the model (such as the number of noise steps, learning rate, network structure) according to the verification results to optimize the generation ability of the model. Finally, through iterative training and verification, gradually improve the model performance to ensure that the generated scenarios are more in line with the actual operation requirements and provide high-quality and diverse scenario support for power grid regulation.

[0106] Step Thirteen: Regularly update the training dataset and retrain the unsupervised generation model using the updated data.

[0107] Collect the latest load data (such as real-time monitoring data, user behavior data) and external data (such as weather, economic data), and combine them with the original dataset. Secondly, clean, standardize, and extract features from the updated data to ensure data quality. Finally, retrain the unsupervised generation model using the updated dataset, optimize the model parameters and generation ability, and ensure that the model can adapt to the changes in the power grid operation environment and continuously generate high-quality and diverse regulation scenarios to provide reliable data support for power grid regulation.

[0108] In the embodiments of the present invention, a method for generating power grid regulation scenarios based on an unsupervised generation model is formed to solve problems such as low computational efficiency, high computational resource consumption when dealing with high-dimensional load data by existing algorithms, and insufficient generation effects when data quality is poor, and provide basic support for subsequent power grid regulation optimization, new energy consumption, source-network-load-storage collaborative interaction, and other links.

[0109] The following is a detailed description in combination with another embodiment.

[0110] Embodiment Two

[0111] A device for generating power grid regulation scenarios provided in this embodiment includes multiple implementation units, and each implementation unit corresponds to each implementation step in Embodiment One above. The specific implementation manners and beneficial effects can refer to the foregoing method embodiments and will not be elaborated here.

[0112] Figure 3 is a schematic diagram of a generating device for an optional power grid regulation scenario according to an embodiment of the present invention. As Figure 3 shown, the generating device for the power grid regulation scenario may include: an acquisition unit 31, an extraction unit 32, an output unit 33, and a generation unit 34, where

[0113] The acquisition unit 31 is configured to acquire power grid load data and preprocess the power grid load data to obtain preprocessed power grid load data;

[0114] The extraction unit 32 is configured to perform feature extraction and encoding on the preprocessed power grid load data to obtain a load feature vector;

[0115] The output unit 33 is configured to input the load feature vector into an unsupervised generation model, and output a power grid regulation scenario through the unsupervised generation model, where the unsupervised generation model is a pre-constructed machine learning model;

[0116] The generation unit 34 is configured to calculate the similarity between the power grid regulation scenario and a pre-constructed reference regulation scenario, configure a type label for the generated power grid regulation scenario based on the similarity, and generate a target power grid regulation scenario based on the power grid regulation scenario and the type label.

[0117] For the above-mentioned generating device for the power grid regulation scenario, the acquisition unit 31 acquires power grid load data and preprocesses the power grid load data to obtain preprocessed power grid load data; the extraction unit 32 performs feature extraction and encoding on the preprocessed power grid load data to obtain a load feature vector; the output unit 33 inputs the load feature vector into an unsupervised generation model, and outputs a power grid regulation scenario through the unsupervised generation model, where the unsupervised generation model is a pre-constructed machine learning model; the generation unit 34 calculates the similarity between the power grid regulation scenario and a pre-constructed reference regulation scenario, configures a type label for the generated power grid regulation scenario based on the similarity, and generates a target power grid regulation scenario based on the power grid regulation scenario and the type label.

[0118] In this embodiment, power grid load data is collected. The power grid load data includes historical data and real-time data. Then, preprocessing and feature extraction are performed on the power grid load data to ensure data quality. For the extracted load feature vectors, an unsupervised generation model is used to generate power grid regulation scenarios. Finally, similarity calculation is performed between the generated power grid regulation scenarios and pre-constructed typical scenarios to clarify the scenario types to which the generated power grid regulation scenarios belong, thereby generating target power grid regulation scenarios, achieving the purpose of automatically generating accurate power grid regulation scenarios, meeting the real-time requirements of power grid regulation, obtaining the technical effect of improving the generation efficiency of power grid regulation scenarios, and further solving the technical problem of low efficiency in the method for generating power grid regulation scenarios based on historical data and expert strategies in the related art.

[0119] Further, the acquisition unit 31 includes: a first acquisition module for accessing a multi-source data platform to acquire historical load data of different time scales from the multi-source data platform; a second acquisition module for acquiring real-time load data of different types of loads by real-time monitoring the power grid system, where the load types include at least one of the following: industrial load, commercial load, residential load, electric vehicle charging load; a first obtaining module for obtaining power grid load data based on the historical load data and the real-time load data.

[0120] Further, the acquisition unit 31 further includes: a first cleaning module for cleaning the power grid load data to obtain the cleaned power grid load data, where the data cleaning includes at least one of the following: missing value processing, outlier processing, noise removal; a first processing module for performing standardization and normalization processing on the cleaned power grid load data to obtain the preprocessed power grid load data.

[0121] Further, the acquisition unit 31 further includes: a first output module for inputting the power grid load data into a data generation model and outputting synthetic load data similar to the power grid load data, where the data generation model is a pre-constructed machine learning model for generating similar data based on input data; a first expansion module for expanding the power grid load data through data transformation technology to obtain expanded load data; a second obtaining module for obtaining the preprocessed power grid load data based on the synthetic load data and the expanded load data.

[0122] Further, the extraction unit 32 includes: a first extraction module for extracting time series features from the grid load data, where the time series features include at least one of the following: mean, variance, peak value, valley value, and fluctuation feature; a second extraction module for extracting load features of different types of users from the grid load data, where the user types include: industrial users, commercial users, and residential users, and the load features include at least one of the following: user power consumption distribution feature, user power consumption time distribution feature; a third extraction module for combining the external environment data in the grid load data to extract the influence features of the external environment on the load; a fourth extraction module for extracting the regulation potential features of the adjustable load from the grid load data, where the adjustable load includes at least one of the following: industrial load, temperature control load, electric vehicle load, energy storage load, and the regulation potential features are used to characterize the regulation ability and regulation range of the adjustable load.

[0123] Further, the generating device for the grid regulation scenario further includes a construction unit, and the construction unit includes: a first construction module for obtaining the historical grid load data within a historical time period and constructing a training sample based on the historical grid load data; a first determination module for selecting a target machine learning model as the initial architecture of the unsupervised generation model, determining the hyperparameters and optimization objectives of the target machine learning model, and obtaining the initial unsupervised generation model; a first training module for iteratively training the model based on the training sample through a forward diffusion process and a reverse generation process until the number of iterations reaches the iteration number threshold, and obtaining the trained unsupervised generation model.

[0124] Further, the said first training module includes: a first training sub-module for adding Gaussian noise to the training sample at each iteration during the forward diffusion process, and inputting the training sample with added noise into the initial unsupervised generation model for learning until the training sample is transformed into a random noise vector and the forward diffusion process stops; a second training sub-module for training the initial unsupervised generation model to denoise the training sample at each iteration during the reverse generation process until the training sample is restored to the original data distribution and the reverse generation process stops; where, in each round of the forward diffusion process and the reverse generation process, the hyperparameters of the initial unsupervised generation model are updated through the optimization objective, and the next round of the forward diffusion process and the reverse generation process are executed based on the updated hyperparameters until the initial unsupervised generation model generates a grid regulation scenario that meets the preset conditions.

[0125] The above-mentioned generating device for the grid regulation scenario may further include a processor and a memory. The above-mentioned acquisition unit 31, extraction unit 32, output unit 33, generation unit 34, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to implement the corresponding functions.

[0126] The above-mentioned processor includes a kernel, which retrieves the corresponding program unit from the memory. One or more kernels can be set, and the grid regulation scenario is generated by adjusting the kernel parameters.

[0127] The above-mentioned memory may include non-permanent memory in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. The memory includes at least one memory chip.

[0128] According to another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium, which includes a stored computer program. When the computer program runs, it controls the device where the computer-readable storage medium is located to execute the method for generating any one of the above grid regulation scenarios.

[0129] According to another aspect of the embodiments of the present invention, there is also provided an electronic device, which includes one or more processors and a memory. The memory is used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the method for generating any one of the above grid regulation scenarios.

[0130] According to another aspect of the embodiments of the present invention, there is also provided a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the method for generating any one of the above grid regulation scenarios.

[0131] This application also provides a computer program product, which is suitable for executing a program initialized with the following method steps when executed on a data processing device: collecting grid load data, preprocessing the grid load data to obtain preprocessed grid load data; extracting and encoding features from the preprocessed grid load data to obtain a load feature vector; inputting the load feature vector into an unsupervised generation model, and outputting a grid regulation scenario through the unsupervised generation model, where the unsupervised generation model is a pre-constructed machine learning model; calculating the similarity between the grid regulation scenario and a pre-constructed reference regulation scenario, configuring a type label for the generated grid regulation scenario based on the similarity, and generating a target grid regulation scenario based on the grid regulation scenario and the type label.

[0132] The present application also provides a computer program product, which, when executed on a data processing device, is further adapted to execute a program initialized with the following method steps: The steps of collecting grid load data include: accessing a multi-source data platform and collecting historical load data at different time scales from the multi-source data platform; collecting real-time load data of different types of loads through real-time monitoring of the power grid system, where the load types include at least one of the following: industrial load, commercial load, residential load, and electric vehicle charging load; obtaining grid load data based on the historical load data and the real-time load data.

[0133] The present application also provides a computer program product, which, when executed on a data processing device, is further adapted to execute a program initialized with the following method steps: The steps of preprocessing the grid load data to obtain preprocessed load data include: performing data cleaning on the grid load data to obtain the cleaned grid load data, where the data cleaning includes at least one of the following: missing value processing, outlier processing, and noise removal; performing standardization and normalization processing on the cleaned grid load data to obtain the preprocessed load data.

[0134] The present application also provides a computer program product, which, when executed on a data processing device, is further adapted to execute a program initialized with the following method steps: The steps of preprocessing the grid load data to obtain preprocessed load data further include: inputting the grid load data into a data generation model to output synthetic load data similar to the grid load data, where the data generation model is a machine learning model pre-constructed for generating similar data based on input data; expanding the grid load data through a data transformation technique to obtain expanded load data; obtaining the preprocessed load data based on the synthetic load data and the expanded load data.

[0135] The present application also provides a computer program product, which, when executed on a data processing device, is further adapted to execute a program initialized with the following method steps: The steps of extracting features from the preprocessed grid load data include: extracting time series features from the grid load data, where the time series features include at least one of the following: mean, variance, peak value, valley value, and fluctuation feature; extracting load features of different types of users from the grid load data, where the user types include: industrial users, commercial users, and residential users, and the load features include at least one of the following: user power consumption distribution feature, user power consumption time distribution feature; combining the external environment data in the grid load data to extract the influence features of the external environment on the load; extracting the regulation potential features of adjustable loads from the grid load data, where the adjustable loads include at least one of the following: industrial load, temperature control load, electric vehicle load, and energy storage load, and the regulation potential features are used to characterize the regulation ability and regulation range of the adjustable loads.

[0136] The present application also provides a computer program product, which, when executed on a data processing device, is also adapted to execute a program initialized with the following method steps: The steps of constructing an unsupervised generation model include: obtaining historical power grid load data within a historical time period, and constructing training samples based on the historical power grid load data; selecting a target machine learning model as the initial architecture of the unsupervised generation model, determining the hyperparameters and optimization objectives of the target machine learning model, to obtain an initial unsupervised generation model; based on the training samples, iteratively training the model through a forward diffusion process and a reverse generation process until the number of iterations reaches an iteration number threshold, to obtain a trained unsupervised generation model.

[0137] The present application also provides a computer program product, which, when executed on a data processing device, is also adapted to execute a program initialized with the following method steps: The steps of iteratively training the model through a forward diffusion process and a reverse generation process based on the training samples include: in the forward diffusion process, at each iteration, adding Gaussian noise to the training samples, and inputting the training samples with added noise into the initial unsupervised generation model for learning until the training samples are transformed into a random noise vector, and stopping the forward diffusion process; in the reverse generation process, training the initial unsupervised generation model to denoise the training samples at each iteration until the training samples are restored to the original data distribution, and stopping the reverse generation process; wherein, in each round of the forward diffusion process and the reverse generation process, the hyperparameters of the initial unsupervised generation model are updated through the optimization objective, and the next round of the forward diffusion process and the reverse generation process are executed based on the updated hyperparameters until the initial unsupervised generation model generates a power grid regulation scenario that meets the preset conditions.

[0138] Figure 4 is a hardware structure block diagram of an electronic device (or mobile device) for executing a method for generating a power grid regulation scenario according to an embodiment of the present invention. As Figure 4 shown, the electronic device may include one or more processors ( Figure 4 denoted as 402a, 402b,..., 402n in , and the processor may include, but is not limited to, a processing device such as a microprocessor MCU or a field programmable gate array FPGA), a memory 404 for storing data. In addition, it may further include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, a keyboard, a power supply, and / or a camera. Those of ordinary skill in the art can understand that Figure 4 the structure shown Figure 4 is only schematic and does not limit the structure of the above-mentioned electronic device. For example, the electronic device may further include more or fewer components than Figure 4 shown, or have a different configuration from

[0139] The serial numbers of the above embodiments of the present invention are only for description and do not represent the superiority or inferiority of the embodiments.

[0140] In the above embodiments of the present invention, the descriptions of the various embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0141] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only illustrative. For example, the division of the units can be a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces. The indirect coupling or communication connection of the units or modules can be in an electrical or other form.

[0142] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0143] In addition, the functional units in the various embodiments of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0144] If the above integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. And the foregoing storage medium includes: USB flash drives, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), mobile hard disks, magnetic disks or optical disks and other various media that can store program codes.

[0145] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for generating a power grid regulation scenario, characterized in that Including: Collecting grid load data, preprocessing the grid load data to obtain preprocessed grid load data; Performing feature extraction and encoding on the preprocessed grid load data to obtain a load feature vector; Inputting the load feature vector into an unsupervised generation model, and outputting a grid regulation scenario through the unsupervised generation model, where the unsupervised generation model is a pre-constructed machine learning model; Calculating the similarity between the grid regulation scenario and a pre-constructed reference regulation scenario, configuring a type label for the generated grid regulation scenario based on the similarity, and generating a target grid regulation scenario based on the grid regulation scenario and the type label.

2. The method according to claim 1, wherein The steps of collecting grid load data include: Accessing a multi-source data platform and collecting historical load data at different time scales from the multi-source data platform; Collecting real-time load data of different types of loads through real-time monitoring of the power grid system, where the load types include at least one of the following: industrial load, commercial load, residential load, electric vehicle charging load; Obtaining grid load data based on the historical load data and the real-time load data.

3. The method according to claim 1, characterized in that The steps of preprocessing the grid load data to obtain preprocessed load data include: Performing data cleaning on the grid load data to obtain cleaned grid load data, where data cleaning includes at least one of the following: missing value processing, outlier processing, noise removal; Performing standardization and normalization processing on the cleaned grid load data to obtain preprocessed grid load data.

4. The method according to claim 1, characterized in that, The steps of preprocessing the grid load data to obtain preprocessed load data further include: Inputting the grid load data into a data generation model to output synthetic load data similar to the grid load data, where the data generation model is a pre-constructed machine learning model for generating similar data based on input data; Expanding the grid load data through data transformation techniques to obtain expanded load data; Obtaining the preprocessed grid load data based on the synthetic load data and the expanded load data.

5. The method according to claim 1, characterized in that The steps of performing feature extraction on the preprocessed grid load data include: Extracting time series features from the grid load data, where the time series features include at least one of the following: mean, variance, peak value, valley value, and fluctuation feature; Extracting load features of different types of users from the grid load data, where the user types include: industrial users, commercial users, and residential users, and the load features include at least one of the following: user power consumption distribution feature, user power consumption time distribution feature; Combining the external environment data in the grid load data to extract the influence features of the external environment on the load; Extracting the regulation potential features of adjustable loads from the grid load data, where the adjustable loads include at least one of the following: industrial load, temperature control load, electric vehicle load, energy storage load, and the regulation potential features are used to characterize the regulation ability and regulation range of the adjustable loads.

6. The method according to claim 1, characterized in that, The steps of constructing the unsupervised generation model include: Obtain historical power grid load data within a historical time period, and construct training samples based on the historical power grid load data; Select a target machine learning model as the initial architecture of the unsupervised generation model, determine the hyperparameters and optimization objectives of the target machine learning model, and obtain an initial unsupervised generation model; Based on the training samples, iteratively train the model through the forward diffusion process and the reverse generation process until the number of iterations reaches the iteration number threshold, and obtain the trained unsupervised generation model.

7. The method according to claim 6, characterized in that, The steps of iteratively training the model through the forward diffusion process and the reverse generation process based on the training samples include: In the forward diffusion process, at each iteration, add Gaussian noise to the training samples, and input the training samples with added noise into the initial unsupervised generation model for learning until the training samples are transformed into a random noise vector, and stop the forward diffusion process; In the reverse generation process, train the initial unsupervised generation model to denoise the training samples at each iteration until the training samples are restored to the original data distribution, and stop the reverse generation process; Among them, in each round of the forward diffusion process and the reverse generation process, update the hyperparameters of the initial unsupervised generation model through the optimization objective, and execute the next round of the forward diffusion process and the reverse generation process based on the updated hyperparameters until the initial unsupervised generation model generates a power grid regulation scenario that meets the preset conditions.

8. A generating device for a power grid regulation scenario, characterized in that, Include: An acquisition unit, configured to acquire power grid load data, and preprocess the power grid load data to obtain preprocessed power grid load data; An extraction unit, configured to extract and encode features of the preprocessed power grid load data to obtain a load feature vector; An output unit, configured to input the load feature vector into an unsupervised generation model, and output a power grid regulation scenario through the unsupervised generation model, where the unsupervised generation model is a pre-constructed machine learning model; A generation unit, configured to calculate the similarity between the power grid regulation scenario and a pre-constructed reference regulation scenario, configure a type label for the generated power grid regulation scenario based on the similarity, and generate a target power grid regulation scenario based on the power grid regulation scenario and the type label.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein when the computer program runs, it controls the device where the computer-readable storage medium is located to execute the method for generating a power grid regulation scenario according to any one of claims 1 to 7.

10. An electronic device, characterized in that, Include one or more processors and a memory, the memory is configured to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the method for generating a power grid regulation scenario according to any one of claims 1 to 7.

Images