Intelligent power grid load dynamic monitoring method and system

Abstract

The invention provides an intelligent power grid load dynamic monitoring method and system, and the method comprises the steps: collecting real-time power consumption data, and generating a real-time power consumption data sequence; constructing a time sequence model based on historical synchronous data, generating an expected power load data sequence, and comparing the real-time power consumption data sequence with the expected power load data sequence to obtain a load trend prediction result; analyzing by applying a self-adaptive dynamic clustering algorithm based on a graph neural network, generating a configuration scheme of a clustering center and a clustering number, determining an optimal power distribution scheme in combination with a bilevel programming model and a robust optimization algorithm, and evaluating the risk of the optimal power distribution scheme to generate a power distribution strategy; based on a power distribution strategy, a demand side management strategy based on the game theory is formulated to generate an intelligent load response strategy so as to realize dynamic load monitoring; according to the invention, the management level of the smart power grid is improved, and a solid foundation is provided for future energy management and optimal scheduling.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of smart grid technology, and in particular to a method and system for dynamically monitoring loads in a smart grid. Background Art

[0002] With the transformation of the global energy structure and the development of smart grids, the complexity and dynamics of power systems have increased significantly. Dynamic load monitoring in smart grids is an important part of ensuring the stable and safe operation of power systems. With the popularity of distributed energy resources, electric vehicles, and smart home devices, power demand has become increasingly complex and unpredictable. To adapt to this change, the power grid needs to monitor and manage power consumption in real time to optimize resource allocation, improve efficiency, and ensure a balance between supply and demand.

[0003] Currently, most power companies use traditional methods based on SCADA systems to monitor loads. SCADA systems collect information through sensors installed at key locations such as substations and power plants, and then transmit the data to a central control system for analysis. Existing methods have low sampling frequencies and cannot capture fast-changing load fluctuations, which can lead to delayed or inaccurate decisions. Summary of the invention

[0004] The embodiments of the present invention provide a method and system for dynamically monitoring loads in a smart grid, so as to solve the problem that the prior art cannot capture rapidly changing load fluctuations, which may lead to delayed or inaccurate decision making.

[0005] In a first aspect, an embodiment of the present invention provides a method for dynamically monitoring loads in a smart grid, comprising:

[0006] Collecting real-time power consumption data and generating a real-time power consumption data sequence;

[0007] Based on the historical data of the same period, a time series model is constructed, based on the time series model, an expected power load data sequence is generated, and the real-time power consumption data sequence and the expected power load data sequence are compared to determine the load trend prediction result;

[0008] Based on the load trend prediction results, an adaptive dynamic clustering algorithm based on a graph neural network is applied to analyze the spatial distribution characteristics and time variation characteristics of the power load to generate a configuration scheme of cluster centers and the number of clusters; based on the configuration scheme, an optimal power distribution scheme is determined in combination with a bi-level programming model and a robust optimization algorithm; the risk of the optimal power distribution scheme is evaluated to generate a power distribution strategy;

[0009] Based on the power distribution strategy, a demand-side management strategy based on game theory is formulated. Based on the demand-side management strategy, the interaction behaviors among different user groups are simulated to generate an intelligent load response strategy, and the intelligent load response strategy is used to achieve dynamic load monitoring.

[0010] Optionally, based on the load trend prediction result, an adaptive dynamic clustering algorithm based on graph neural network is applied to analyze the spatial distribution characteristics and time-varying characteristics of power load, so as to generate a configuration plan for the clustering center and the number of clusters. Based on the configuration plan, combined with a bilevel programming model and a robust optimization algorithm, an optimal power distribution plan is determined, and the risk of the optimal power distribution plan is evaluated to generate a power distribution strategy, including:

[0011] Based on the load trend prediction result, an adaptive dynamic clustering algorithm based on graph neural network is applied to analyze the connection relationship among grid components in the grid topology structure, so as to obtain the spatial distribution characteristics and time-varying characteristics of power load. Based on the spatial distribution characteristics and the time-varying characteristics, a configuration plan for the clustering center location and the number of clusters is generated;

[0012] Based on the configuration plan, a bilevel programming model is used to handle the power distribution and scheduling problems to determine a preliminary power distribution plan. A robust optimization algorithm is introduced to handle the uncertainty and risk factors in the preliminary power distribution plan to determine the optimal power distribution plan;

[0013] Bayesian network technology is applied to construct a causal relationship model. Based on the causal relationship model, the occurrence probability and influence degree of known risk points of the optimal power distribution plan are evaluated to obtain a first risk assessment result;

[0014] The Monte Carlo simulation algorithm is used to evaluate the occurrence probability and influence degree of different risk scenarios to obtain a second risk assessment result. Combining the first risk assessment result, a target risk assessment result is determined. A reinforcement learning algorithm is applied to optimize the optimal power distribution plan and the target risk assessment result to generate a power distribution strategy.

[0015] Optionally, based on the load trend prediction result, an adaptive dynamic clustering algorithm based on graph neural network is applied to analyze the connection relationship among grid components in the grid topology structure, so as to obtain the spatial distribution characteristics and time-varying characteristics of power load. Based on the spatial distribution characteristics and the time-varying characteristics, a configuration plan for the clustering center location and the number of clusters is generated, including:

[0016] Based on the load trend prediction result, the connection relationship among grid components in the grid topology structure is modeled to obtain a connection relationship model among grid components;

[0017] Based on the connection relationship model, combined with the spatio-temporal attention mechanism, obtain the spatio-temporal characteristics of the power load to generate a spatio-temporal feature representation, where the spatio-temporal feature representation includes the spatial distribution characteristics and time variation characteristics of the power load;

[0018] Apply a variational autoencoder to perform anomaly detection and load fluctuation pattern recognition processing on the spatio-temporal feature representation to obtain a load recognition result;

[0019] Based on the load recognition result, use an adaptive dynamic clustering algorithm and hierarchical clustering technology to configure the clustering center position and the number of clusters, and generate a configuration scheme for the clustering center position and the number of clusters, so that each cluster reflects the load patterns in different regions or time periods of the smart grid.

[0020] Optionally, based on the power distribution strategy, formulate a demand-side management strategy based on game theory. Based on the demand-side management strategy, simulate the interaction behaviors between different user groups to generate an intelligent load response strategy, and the intelligent load response strategy is used to achieve load dynamic monitoring, including:

[0021] Based on the power distribution strategy, use game theory to analyze the interaction behaviors among multiple participants to obtain an interaction behavior analysis result, and at the same time set multiple incentive measures. Based on the incentive measures and the interaction behavior analysis result, formulate a demand-side management strategy, and the participants include power users, power generation stations, and grid operators;

[0022] Based on the demand-side management strategy, construct a simulation environment to simulate the interaction behaviors between different user groups to obtain a target interaction behavior simulation result. Based on the target interaction behavior simulation result, predict the reactions of different user groups under different incentive measures to obtain a user reaction prediction result, and the user groups include residential users, commercial users, and industrial users;

[0023] Based on the user reaction prediction result, generate a preliminary load response strategy, and the preliminary load response strategy includes a real-time pricing mechanism, a demand response plan, and an automatic control strategy;

[0024] Apply the preliminary load response strategy, and monitor the load change situation in real time to obtain real-time monitoring data. Combine the historical data and the real-time monitoring data to predict the load trend. Based on the load trend, adjust the power distribution strategy to optimize the preliminary load response strategy to obtain an intelligent load response strategy, and perform load dynamic monitoring based on the intelligent load response strategy.

[0025] Optionally, based on the demand-side management strategy, a simulation environment is constructed to simulate the interaction behaviors among different user groups, obtaining a target interaction behavior simulation result. Based on the target interaction behavior simulation result, the reactions of different user groups under different incentive measures are predicted, obtaining a user reaction prediction result, including:

[0026] Based on the demand-side management strategy, a simulation environment is constructed using a multi-agent reinforcement learning framework to simulate the interaction behaviors among different user groups, obtaining a preliminary interaction behavior simulation result;

[0027] Based on the preliminary interaction behavior simulation result, a generative adversarial network is applied to generate multiple extreme load scenarios, simulating the interaction behaviors of different user groups under different extreme load scenarios to optimize the preliminary interaction behavior simulation result, obtaining a target interaction behavior simulation result;

[0028] Based on the target interaction behavior simulation result, the impact of each user group on the smart grid is analyzed, obtaining a user impact result. The actual reaction patterns of different user groups under different incentive measures are evaluated, obtaining a user reaction evaluation result. Combining the user impact result and the user reaction evaluation result, a user impact evaluation result is obtained;

[0029] Based on the user impact evaluation result, a long short-term memory network is used to predict the change trend of the electricity consumption behaviors of different user groups in different time periods, and a graph neural network is used to analyze the spatial correlation among different user groups. Combining the change trend of the electricity consumption behaviors and the spatial correlation, a user reaction prediction result is generated.

[0030] Optionally, based on historical data for the same period, a time series model is constructed. Based on the time series model, an expected power load data sequence is generated. The real-time power consumption data sequence and the expected power load data sequence are compared to determine a load trend prediction result, including:

[0031] Based on historical data for the same period, a time series model is constructed. Using the time series model, the expected power load situation in a specific time period is predicted to generate an expected power load data sequence;

[0032] Based on historical data for the same period, the expected power load data sequence and the real-time power consumption data sequence are compared to identify a verified load fluctuation abnormal pattern, and a cause analysis is performed on the verified load fluctuation abnormal pattern to obtain a cause analysis result;

[0033] Based on the cause analysis results, the key parameters of the time series model are adjusted to obtain an optimized time series model. Based on the optimized time series model, the power load trend is predicted to obtain a load trend prediction result, which includes a movement prediction result during peak and valley periods and a rise and fall prediction result of the average load level.

[0034] Optionally, comparing the expected power load data sequence with the real-time power consumption data sequence, identifying a verified abnormal load fluctuation pattern, and performing a cause analysis on the verified abnormal load fluctuation pattern to obtain a cause analysis result, including:

[0035] Comparing the expected power load data sequence with the real-time power consumption data sequence to obtain a comparison result, performing difference analysis on the comparison result using a statistical method, and identifying a preliminary load fluctuation abnormal pattern;

[0036] Combining the preliminary load fluctuation abnormal pattern with the target features affecting load changes extracted from historical data, and building an abnormality detection model based on the target features using a local outlier factor algorithm, the target features including temperature and humidity;

[0037] Applying the anomaly detection model to detect the real-time power consumption data sequence to identify a verified load fluctuation anomaly pattern;

[0038] Combined with environmental and social dynamic information, external factor evaluation and processing are performed on the verified abnormal load fluctuation pattern to obtain the external factor impact result;

[0039] Detect the internal situation of the smart grid, perform internal factor evaluation on the verified abnormal load fluctuation pattern to obtain internal factor impact results, perform cause analysis on the verified abnormal load fluctuation pattern based on the external factor impact results and the internal factor impact results to obtain cause analysis results, wherein the internal situation includes equipment status, maintenance records, and user behavior history.

[0040] In a second aspect, an embodiment of the present invention provides a smart grid load dynamic monitoring system, including:

[0041] A collection module, used to collect real-time power consumption data and generate a real-time power consumption data sequence;

[0042] A construction module is used to construct a time series model based on historical data of the same period, generate an expected power load data sequence based on the time series model, and compare the real-time power consumption data sequence with the expected power load data sequence to determine a load trend prediction result;

[0043] An evaluation module, configured to apply an adaptive dynamic clustering algorithm based on a graph neural network to analyze the spatial distribution characteristics and temporal variation characteristics of the power load based on the load trend prediction result, so as to generate a configuration scheme for the cluster center and the number of clusters. Based on the configuration scheme, combining a bilevel programming model and a robust optimization algorithm, determine an optimal power distribution scheme, and evaluate the risk of the optimal power distribution scheme to generate a power distribution strategy;

[0044] A generation module, configured to formulate a demand-side management strategy based on game theory based on the power distribution strategy, and simulate the interaction behaviors among different user groups based on the demand-side management strategy to generate an intelligent load response strategy, where the intelligent load response strategy is used to implement load dynamic monitoring.

[0045] In a third aspect, an embodiment of the present invention provides a computing device, including a processor and a memory, where a computer program is stored in the memory, and the processor is configured to run the computer program to execute any one of the methods for dynamically monitoring the load of an intelligent power grid in the first aspect.

[0046] In a fourth aspect, an embodiment of the present invention provides a computer storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the methods for dynamically monitoring the load of an intelligent power grid described in any one of the first aspects are implemented.

[0047] In the embodiments of the present invention, real-time power consumption data is collected to generate a real-time power consumption data sequence; based on historical data in the same period, a time series model is constructed, and based on the time series model, an expected power load data sequence is generated. The real-time power consumption data sequence and the expected power load data sequence are compared to determine the load trend prediction result; based on the load trend prediction result, an adaptive dynamic clustering algorithm based on a graph neural network is applied to analyze the spatial distribution characteristics and time variation characteristics of the power load to generate a configuration scheme for the clustering center and the number of clusters. Based on the configuration scheme, combined with a bilevel programming model and a robust optimization algorithm, an optimal power distribution scheme is determined, and the risk of the optimal power distribution scheme is evaluated to generate a power distribution strategy; based on the power distribution strategy, a demand-side management strategy based on game theory is formulated, and based on the demand-side management strategy, the interaction behaviors between different user groups are simulated to generate an intelligent load response strategy, and the intelligent load response strategy is used to achieve load dynamic monitoring. The technical solution provided by the present invention improves the accuracy of load prediction through a time series model and anomaly detection; ensures the optimal allocation of power resources and improves the operating efficiency of the system through a bilevel programming model and a robust optimization algorithm; improves the user participation and response efficiency and promotes the supply-demand balance based on the demand-side management strategy based on game theory; ensures the safety and reliability of the power distribution strategy and reduces potential risks through risk assessment; the intelligent load response strategy is used to achieve load dynamic monitoring to ensure that the power grid can operate efficiently under various conditions. Among them, through a spatio-temporal attention mechanism and a variational autoencoder, the spatial distribution characteristics and time variation characteristics of the power load are accurately captured, improving the accuracy of load prediction; applying an adaptive dynamic clustering algorithm and hierarchical clustering technology to ensure the flexibility and accuracy of the clustering results, enhancing the adaptability and robustness of the system; based on the detailed load identification results and the clustering configuration scheme, the optimal allocation of power resources is realized, improving the operating efficiency and economic benefits of the system; anomaly detection is performed through a variational autoencoder to timely discover and process abnormal patterns in the power load, ensuring the safe and stable operation of the system; based on the clustering results, more intelligent power scheduling and user response strategies can be formulated, improving the flexibility and response efficiency of the system.

[0048] These aspects or other aspects of the present invention will be more clearly understood in the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0050] Figure 1 A flowchart of an intelligent power grid load dynamic monitoring method provided by an embodiment of the present invention;

[0051] Figure 2 A schematic structural diagram of an intelligent power grid load dynamic monitoring system provided by an embodiment of the present invention;

[0052] Figure 3 A schematic structural diagram of a computing device provided by an embodiment of the present invention. Detailed implementation manners

[0053] In order to enable those skilled in the art to better understand the solution 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.

[0054] In some processes described in the specification, claims and above-mentioned drawings of the present invention, a plurality of operations that appear in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions such as "first" and "second" in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit that "first" and "second" are of different types.

[0055] 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 skilled in the art without creative efforts belong to the scope of protection of the present invention.

[0056] Figure 1 An embodiment of the present invention provides a flowchart of an intelligent power grid load dynamic monitoring method. As Figure 1 shown, the method includes:

[0057] Facing the increasingly complex power system and the growing energy demand, traditional load monitoring and management methods are difficult to meet the requirements of modern intelligent power grids. In order to improve the stability and reliability of power supply, optimize the resource utilization efficiency, and achieve more accurate and flexible power dispatching, there is an urgent need for a load dynamic monitoring method that can monitor in real time, predict accurately, and respond intelligently. Based on this, the present invention provides an intelligent power grid load dynamic monitoring method. As Figure 1 shown, it includes:

[0058] Step 101: Collect real-time power consumption data and generate a real-time power consumption data sequence;

[0059] In this step, real-time power consumption data refers to the instantaneous power usage of each user or area in the power grid within a certain time period. These data are usually collected every minute or hour through devices such as smart meters and sensors. The real-time power consumption data sequence refers to arranging the collected real-time power consumption data in chronological order to form a time series for subsequent analysis and processing.

[0060] Suppose in a city power grid, there are multiple smart meters distributed in different areas (such as residential areas, commercial areas, industrial areas). Each meter records the power consumption of its area every 15 minutes. By collecting these data, a data sequence containing timestamps and corresponding power consumption values is generated. In this way, the system can continuously obtain and update real-time power consumption data, providing basic data support for subsequent steps.

[0061] Step 102: Based on historical data for the same period, construct a time series model. Based on the time series model, generate an expected power load data sequence, and compare the real-time power consumption data sequence with the expected power load data sequence to determine the load trend prediction result;

[0062] In this step, historical data for the same period refers to power consumption data in the same time period in the past, such as data for the same month or the same day last year. The expected power load data sequence refers to the power load data sequence predicted for a certain future period through a time series model. The abnormal pattern of load fluctuations refers to the significant difference between the actual power consumption data and the expected power load data, which may indicate the influence of unforeseen external factors or internal changes in the system.

[0063] Taking a typical summer weekday as an example, suppose the power consumption data for weekdays in July of each of the past three years is collected. These historical data for the same period are used to construct a time series model to predict the power load for a weekday in July this year. Then, this expected power load data sequence is compared with the real-time power consumption data sequence to identify data points that deviate significantly from the expected values. For example, at 08:15, the actual power consumption is 520 kW while the expected value is 510 kW, which indicates possible abnormal fluctuations. By analyzing these abnormal patterns, future load trends can be further predicted, such as finding that the peak period arrives earlier or the trough period extends.

[0064] Step 103: Based on the load trend prediction results, apply the adaptive dynamic clustering algorithm based on graph neural network to analyze the spatial distribution characteristics and temporal variation characteristics of the power load, so as to generate a configuration plan for the clustering centers and the number of clusters. Based on the configuration plan, combine the bilevel programming model and the robust optimization algorithm to determine the optimal power distribution plan, and evaluate the risk of the optimal power distribution plan to generate a power distribution strategy;

[0065] In this step, the spatial distribution characteristics refer to the distribution of the power load at different geographical locations. The temporal variation characteristics refer to the variation law of the power load over time.

[0066] This step is based on the load trend prediction results and applies the adaptive dynamic clustering algorithm based on graph neural network to analyze the spatial distribution characteristics and temporal variation characteristics of the power load. For example, suppose there is a city power grid divided into multiple power supply areas (such as Area A, Area B, Area C, etc.). Through the analysis of the adaptive dynamic clustering algorithm of the graph neural network, it is found that the power loads in some areas have strong spatial correlation (such as between Area A and Area B), and at the same time, the load variation laws of each area at different time periods are identified. Based on these analysis results, a configuration plan for the clustering centers and the number of clusters is generated. For example, Area A and Area B are combined into one cluster, while Area C is taken as a separate cluster. Then, combine the bilevel programming model and the robust optimization algorithm to determine the optimal power distribution plan. For example, through the bilevel programming model, optimize the power generation of power plants and the scheduling of transmission lines to ensure minimizing costs while meeting the demand; through the robust optimization algorithm, evaluate the risks of different plans and select the plan with the lowest risk and the best performance as the final power distribution strategy.

[0067] Step 104: Based on the power distribution strategy, formulate a demand-side management strategy based on game theory. Based on the demand-side management strategy, simulate the interaction behaviors among different user groups to generate an intelligent load response strategy, and the intelligent load response strategy is used to achieve dynamic load monitoring;

[0068] In this step, the demand-side management strategy refers to using various technologies and strategies to influence the electricity consumption behaviors of users to achieve the balance between power supply and demand and the optimal allocation of resources.

[0069] This step formulates a set of demand - side management strategies based on game theory, which is based on the power distribution strategy. For example, assume there are three user groups (residential users, commercial users, and industrial users), and their electricity consumption behaviors affect each other. Through the game theory model, the interaction behaviors among these user groups are simulated. For example, residential users reduce air - conditioner usage during peak hours, commercial users increase electricity consumption during off - peak hours, and industrial users adjust their electricity - using times according to production plans. Based on these simulation results, intelligent load response strategies are generated. For example, electricity - fee discounts are provided to residential users to encourage them to reduce electricity consumption during peak hours, and subsidies are provided to commercial users to encourage them to increase electricity consumption during off - peak hours. These strategies not only improve users' participation and response efficiency but also achieve dynamic balance and optimized management of the power load.

[0070] In the embodiments of the present invention, through the time - series model and anomaly detection, the accuracy of load forecasting is significantly improved, enhancing the scientific nature and reliability of power dispatching; through the graph neural network and the adaptive dynamic clustering algorithm, the spatial distribution characteristics and time - varying characteristics of the power load are comprehensively captured, improving the adaptability and robustness of the system; by combining the bilevel programming model and the robust optimization algorithm, the optimal allocation of power resources is ensured, improving the operation efficiency and economic benefits of the system; the demand - side management strategy based on game theory fully considers the interaction behaviors among users, improves users' participation and response efficiency, and promotes the balance between supply and demand; through risk assessment, the security and reliability of the power distribution strategy are ensured, potential risks are reduced, and the safe and stable operation of the power grid is guaranteed.

[0071] After completing the load trend forecasting and anomaly pattern recognition, in order to further optimize the allocation and management of power resources, based on this, the present invention provides a specific embodiment, step 103. Based on the load trend forecasting results, apply the adaptive dynamic clustering algorithm based on the graph neural network to analyze the spatial distribution characteristics and time - varying characteristics of the power load, so as to generate a configuration plan for the clustering center and the number of clusters. Based on the configuration plan, combine the bilevel programming model and the robust optimization algorithm to determine the optimal power allocation plan, and evaluate the risk of the optimal power allocation plan to generate a power distribution strategy, which specifically includes the following steps:

[0072] Step 301: Based on the load trend forecasting results, apply the adaptive dynamic clustering algorithm based on the graph neural network to analyze the connection relationships among the power grid components in the power grid topology structure, so as to obtain the spatial distribution characteristics and time - varying characteristics of the power load. Based on the spatial distribution characteristics and the time - varying characteristics, generate a configuration plan for the clustering center location and the number of clusters;

[0073] Assume that the load trend prediction results of a certain urban power grid in the next week have been obtained. Next, an adaptive dynamic clustering algorithm based on graph neural networks is applied to analyze the connection relationships among various components (such as power plants, substations, transmission lines, etc.) in the power grid topology. For example, through the adaptive dynamic clustering algorithm of graph neural networks, it can be found that the power loads in certain areas have strong spatial correlations (such as between Area A and Area B), and at the same time, the load change rules in each area at different time periods can be identified. Based on these analysis results, a configuration plan for the clustering center location and the number of clusters is generated. For example, Area A and Area B are merged into one cluster, while Area C is taken as a separate cluster. This configuration plan not only considers the spatial distribution characteristics of the power load but also captures the time-varying characteristics, providing a basis for subsequent power distribution optimization.

[0074] Step 302: Based on the configuration plan, use a bilevel programming model to handle the power distribution and scheduling problems to determine a preliminary power distribution plan, and introduce a robust optimization algorithm to handle the uncertainty and risk factors in the preliminary power distribution plan to determine the optimal power distribution plan;

[0075] This step is based on the configuration plan of the clustering center location and the number of clusters, and uses a bilevel programming model to handle the power distribution and scheduling problems. For example, through the bilevel programming model, the power generation of power plants and the scheduling of transmission lines are optimized to ensure minimizing costs while meeting the demand. This step generates a preliminary power distribution plan. However, there are many uncertainties and risk factors in the actual operating environment (such as weather changes, equipment failures, etc.). Therefore, a robust optimization algorithm is introduced to handle the preliminary power distribution plan. For example, through the robust optimization algorithm, the risks in different scenarios are evaluated, and the plan with the lowest risk and the best performance is selected as the final optimal power distribution plan. This can ensure that the system can still operate efficiently under various uncertain conditions.

[0076] Step 303: Apply Bayesian network technology to construct a causal relationship model, and based on the causal relationship model, evaluate the occurrence probability and impact degree of the known risk points of the optimal power distribution plan to obtain the first risk assessment result;

[0077] In this step, the known risk points refer to the known potential risk factors in the system, such as equipment aging, weather changes, etc.

[0078] To further evaluate the risks of the optimal power distribution scheme, Bayesian network technology is applied to construct a causal relationship model. For example, assuming that certain risk points are known (such as the aging of a certain generator, specific weather conditions, etc.), the occurrence probability and impact degree of these risk points can be quantified through the Bayesian network. Based on the causal relationship model, the impacts of the known risk points in the optimal power distribution scheme are evaluated. For example, the aging of a certain generator may lead to a 5% probability of failure during high-load periods, and such a failure may cause a local power outage, affecting approximately 10% of users. In this way, the first risk assessment result is obtained, providing a scientific basis for subsequent risk management.

[0079] Step 304: Use the Monte Carlo simulation algorithm to evaluate the occurrence probability and impact degree of different risk scenarios, obtain the second risk assessment result, combine the first risk assessment result, determine the target risk assessment result, and apply the reinforcement learning algorithm to optimize the optimal power distribution scheme and the target risk assessment result to generate a power distribution strategy;

[0080] In this step, different risk scenarios refer to various possible risk events and their combinations.

[0081] To further evaluate the occurrence probability and impact degree of different risk scenarios, the Monte Carlo simulation algorithm is used for simulation. For example, assume that multiple risk scenarios (such as extreme weather, equipment failures, etc.) are simulated, and the occurrence probability and impact degree of each scenario are evaluated. Through a large number of random simulations, the second risk assessment result is obtained. Combining the first and second risk assessment results, the target risk assessment result is determined. For example, through comprehensive analysis, it is found that the occurrence probability of a certain extreme weather scenario is 2%, but its impact is very serious, possibly causing a large-scale power outage. Based on these assessment results, the reinforcement learning algorithm is applied to generate a power distribution strategy. For example, through the reinforcement learning algorithm, the dispatching of power plants and the electricity consumption behavior of users are optimized to ensure the stable operation of the system under high-risk scenarios.

[0082] In the embodiments of the present invention, through the graph neural network and the adaptive dynamic clustering algorithm, the spatial distribution characteristics and temporal variation characteristics of the power load are comprehensively captured, and an optimal configuration scheme for the clustering center position and the number of clusters is generated, improving the optimal allocation ability of power resources; combined with the bilevel programming model and the robust optimization algorithm, the best performance of the power distribution scheme under uncertainty and risk factors is ensured, improving the stability and reliability of the system; applying the Bayesian network technology and the Monte Carlo simulation algorithm, the occurrence probabilities and influence degrees of known and unknown risk points are comprehensively evaluated, providing a scientific basis for risk management; based on the target risk assessment results, applying the reinforcement learning algorithm to generate the power distribution strategy, realizing intelligent power dispatching and user response, and improving the flexibility and response efficiency of the system; through multi-level risk assessment and optimization strategies, the safe and stable operation of the power system is ensured, potential risks are reduced, and the security and reliability of the power grid are guaranteed.

[0083] To deeply analyze the spatial distribution characteristics and temporal variation characteristics of the power load and provide a scientific basis for the subsequent optimization of power distribution, based on this, the present invention provides a specific embodiment. Step 301: Based on the load trend prediction result, apply the adaptive dynamic clustering algorithm based on the graph neural network to analyze the connection relationship between the power grid components in the power grid topology structure, so as to obtain the spatial distribution characteristics and temporal variation characteristics of the power load. Based on the spatial distribution characteristics and the temporal variation characteristics, generate a configuration scheme for the clustering center position and the number of clusters, which specifically includes the following steps:

[0084] Step 311: Based on the load trend prediction result, model the connection relationship between the power grid components in the power grid topology structure to obtain a connection relationship model between the power grid components;

[0085] In this step, the power grid topology structure refers to the physical connection relationship between the components in the power grid (such as power plants, substations, and transmission lines, etc.).

[0086] Suppose the load trend prediction result of a certain urban power grid in the next week has been obtained. Next, model the connection relationship between the power grid components in the power grid topology structure. For example, the power grid includes multiple power plants, substations, and transmission lines, and there are complex connection relationships between these components. Represented by a network diagram, each node represents a power grid component, and each edge represents the connection relationship between the components. Among them, an adjacency matrix or an adjacency list can be used to represent this connection relationship. Through this connection relationship model, the connection relationship between each component can be clearly seen, providing a basis for subsequent analysis.

[0087] More specifically, the present invention provides a calculation formula for the connection relationship model to more comprehensively capture the complex relationship between the power grid components. The specific calculation formula is as follows:

[0088] G = (V, E, W = f(H, R));

[0089] W = [ w i j ] |V|×|V| ;

[0090]

[0091] Among them, G represents the connection relationship model; the node set V represents the set containing all power grid components, and each node v i ∈ V can be represented by a set of feature vectors x i , and these features include location information, device type, historical load data, etc.; the edge set E represents the set containing the connection relationships between all power grid components, and each edge e ij ∈ E connects two nodes v i and v j , and can be represented by a binary tuple ( v i ,v j ) ; the weight matrix W represents the set of connection strengths between nodes; w ij represents the connection weight between nodes v i and v j ; |V| represents the size of the node set V; α represents the degree of influence of the control weight coefficient spatial distance on the weight; β represents the degree of influence of the control historical load correlation on the weight; γ represents the degree of influence of the control real-time load difference on the weight; ρ ( H i ,H j ) represents the correlation between the historical load data between nodes v i and v j ; ΔR ( v i ,v j ) represents the real-time load difference; d ( v i ,v j ) represents the spatial distance between nodes v i and v j ; σ d represents the spatial distance attenuation parameter, which is used to adjust the degree of influence of the spatial distance on the weight; represents the exponential decay function.

[0092] The above connection relationship model formula comprehensively calculates the connection weights between nodes by introducing multiple factors such as spatial distance, historical load correlation, and real-time load difference. It not only considers the static spatial distance but also combines the dynamic historical and real-time load changes. The model can more comprehensively capture the complex dependencies between power grid components. In addition, by introducing a time decay factor and an adaptive learning mechanism, the weight matrix can be dynamically adjusted to adapt to the real-time changes of the power system. Combining with graph neural networks for feature extraction further enhances the model's understanding of the power grid topology structure, improving prediction accuracy and robustness. Generally speaking, this calculation method ensures the stability and reliability of the connection relationship model under different times and scenarios, providing a solid foundation for power load forecasting and anomaly detection in smart grids.

[0093] Step 312: Based on the connection relationship model, combine the spatio-temporal attention mechanism to obtain the spatio-temporal features of the power load to generate a spatio-temporal feature representation, where the spatio-temporal feature representation includes the spatial distribution characteristics and temporal variation characteristics of the power load.

[0094] This step is based on the connection relationship model and combines the spatio-temporal attention mechanism to obtain the spatio-temporal features of the power load. The spatio-temporal attention mechanism can dynamically focus on the importance of different time and space positions when processing time-series data, thereby more accurately capturing the change patterns of the power load. For example, in a smart grid with multiple regions, each region has different power consumption patterns. Through the spatio-temporal attention mechanism, it can be found that the power consumption in some regions increases significantly during peak hours (such as 7 pm to 9 pm), while other regions are relatively stable. At the same time, the mechanism can also capture the spatial correlation between different regions. For example, the power loads of adjacent regions may fluctuate synchronously. This spatio-temporal feature representation not only reflects the spatial distribution characteristics of the power load but also captures its temporal variation characteristics, providing rich information for subsequent anomaly detection and clustering analysis.

[0095] Step 313: Apply a variational autoencoder to perform anomaly detection and load fluctuation pattern recognition processing on the spatio-temporal feature representation to obtain a load recognition result.

[0096] In this step, a variational autoencoder is applied to perform anomaly detection and load fluctuation pattern recognition on the above spatio-temporal feature representation. The variational autoencoder can automatically learn the latent representation of power load data and identify anomaly patterns and fluctuation patterns. For example, by training the variational autoencoder, it can be found that the power load in certain time periods significantly deviates from the normal range, which may indicate the influence of unforeseen external factors or internal changes in the system. Suppose that during the period from 2 pm to 4 pm on a certain weekday, it is found that the power load in a certain area suddenly increases significantly while remaining stable in other areas. The variational autoencoder can identify this anomaly pattern and mark it as an object that needs further investigation. In addition, the variational autoencoder can also identify common load fluctuation patterns, such as load surges during peak hours and load decreases during off-peak hours. These load recognition results provide important inputs for subsequent clustering analysis.

[0097] Step 314: Based on the load recognition results, use an adaptive dynamic clustering algorithm and hierarchical clustering technology to configure the clustering center positions and the number of clusters, and generate a configuration plan for the clustering center positions and the number of clusters, so that each cluster reflects the load patterns in different regions or time periods of the smart grid;

[0098] In this step, based on the load recognition results, an adaptive dynamic clustering algorithm and hierarchical clustering technology are used to configure the clustering center positions and the number of clusters. For example, some anomaly patterns and fluctuation patterns are identified. The clustering centers and the number can be dynamically adjusted through the adaptive dynamic clustering algorithm to ensure that each cluster reflects the load patterns in different regions or time periods of the smart grid. Specifically, suppose there are three main regions (Region A, Region B, Region C). Through the adaptive dynamic clustering algorithm, it can be found that the loads in some regions are similar during peak hours (such as from 7 pm to 9 pm), while the load patterns are different during other time periods. Therefore, these similar load patterns are grouped into the same cluster. For example, Cluster 1 has similar load patterns in Region A and Region B during peak hours; Cluster 2 has a relatively stable load pattern in Region C during off-peak hours. Through hierarchical clustering technology, these clusters can be further refined. For example, Region A can be further divided into two sub-clusters to respectively reflect its load changes under different weather conditions. This configuration plan not only considers the spatial distribution characteristics of the power load but also captures the time-varying characteristics, providing a scientific basis for subsequent power distribution optimization.

[0099] In the embodiments of the present invention, through the spatio-temporal attention mechanism and the variational autoencoder, the spatial distribution characteristics and temporal variation characteristics of the power load are accurately captured, improving the accuracy of load forecasting; the adaptive dynamic clustering algorithm and hierarchical clustering technology are applied to ensure the flexibility and accuracy of the clustering results, enhancing the adaptability and robustness of the system; based on the detailed load identification results and clustering configuration scheme, the optimal allocation of power resources is achieved, improving the operating efficiency and economic benefits of the system; anomaly detection is performed through the variational autoencoder to timely discover and handle abnormal patterns in the power load, ensuring the safe and stable operation of the system; based on the clustering results, more intelligent power dispatching and user response strategies can be formulated, improving the flexibility and response efficiency of the system.

[0100] After the accurate prediction, anomaly detection, and optimal clustering of the power load are completed, in order to further improve the flexibility of the power system and user participation, based on this, the present invention provides a specific embodiment. Step 104: Based on the power distribution strategy, formulate a demand-side management strategy based on game theory. Based on the demand-side management strategy, simulate the interaction behaviors among different user groups to generate an intelligent load response strategy, and the intelligent load response strategy is used to achieve load dynamic monitoring, which specifically includes the following steps:

[0101] Step 401: Based on the power distribution strategy, use game theory to analyze the interaction behaviors among multiple participants to obtain the interaction behavior analysis result, and at the same time set multiple incentive measures. Based on the incentive measures and the interaction behavior analysis result, formulate a demand-side management strategy, and the participants include power users, power generation stations, and grid operators;

[0102] In this step, the incentive measures refer to the means used to guide users to change their electricity consumption behaviors, such as electricity bill discounts, subsidies, etc.

[0103] Suppose a power distribution strategy has been formulated, which stipulates the power supply quantities in different regions at different times. Next, game theory is used to analyze the interaction behaviors among power users, power generation stations, and power grid operators. For example, residential users may reduce electricity consumption during peak hours to obtain electricity bill discounts, while power generation stations may adjust power generation according to electricity prices, and power grid operators are responsible for coordinating the interests of all parties to ensure the stable operation of the system. By constructing a game theory model, these interaction behaviors are simulated, and the analysis results of the interaction behaviors are obtained. For example, it is found that residential users are very sensitive to electricity bill discounts, commercial users are more concerned about production efficiency, and industrial users tend to have long-term stable power supply. Based on these analysis results, a variety of incentive measures are set. For example, for residential users, electricity bill discounts during peak hours are provided to encourage them to use energy storage devices; for commercial users, electricity consumption subsidies during off-peak hours are provided to encourage them to increase electricity consumption during non-peak hours; for industrial users, long-term power supply contracts are signed to provide stable and preferential electricity prices. Combining these incentive measures and the analysis results of the interaction behaviors, a demand-side management strategy is formulated, aiming to guide users to change their electricity consumption behaviors through a reasonable incentive mechanism to achieve the balance between power supply and demand.

[0104] Step 402: Based on the demand-side management strategy, construct a simulation environment to simulate the interaction behaviors among different user groups, obtain the target interaction behavior simulation results, and based on the target interaction behavior simulation results, predict the responses of different user groups under different incentive measures to obtain the user response prediction results, where the user groups include residential users, commercial users, and industrial users;

[0105] In this step, the simulation environment refers to a virtual simulation system used to simulate the interaction behaviors among different user groups and their responses to incentive measures.

[0106] In this step, based on the above demand-side management strategy, a simulation environment is constructed to simulate the interaction behaviors among residential users, commercial users, and industrial users. For example, a virtual urban power grid is set up, which includes multiple residential areas, shopping malls, and factories. In this simulation environment, the impacts of different incentive measures on user behaviors are tested. Through multiple simulation experiments, the target interaction behavior simulation results are obtained. For example, it is found that when residential users obtain electricity bill discounts during peak hours, they significantly reduce the use of air conditioners and water heaters; when commercial users obtain electricity consumption subsidies during off-peak hours, they will increase the operating time of production equipment; and when industrial users sign long-term power supply contracts, they will be more dependent on the stable power provided by the power grid. Based on these simulation results, the responses of different user groups under different incentive measures are predicted to obtain the user response prediction results, providing an important basis for the subsequent load response strategy.

[0107] Step 403: Based on the predicted results of the user response, generate a preliminary load response strategy, which includes a real-time pricing mechanism, a demand response plan, and an automated control strategy;

[0108] In this step, the real-time pricing mechanism refers to dynamically adjusting the electricity price according to the real-time power load situation to guide users to use electricity rationally. The demand response plan refers to a pre-established user participation plan, which guides users to adjust their electricity consumption behaviors during specific time periods through incentive measures. The automated control strategy refers to automatically adjusting users' electrical equipment through an intelligent control system to achieve load optimization.

[0109] This step generates a preliminary load response strategy based on the above predicted results of the user response, including the following real-time pricing mechanism, demand response plan, and automated control strategy. For example, increase the electricity price during peak hours to encourage users to reduce electricity consumption; lower the electricity price during off-peak hours to encourage users to increase electricity consumption. Provide electricity bill discounts for residential users during peak hours, provide electricity subsidies for commercial users during off-peak hours, and provide long-term stable electricity price contracts for industrial users. Smart meters can automatically adjust the air conditioner temperature according to the real-time electricity price, and smart sockets can automatically start high-power devices such as washing machines during off-peak hours. These preliminary load response strategies can not only effectively guide users to use electricity rationally, but also optimize the load distribution of the power system.

[0110] Step 404: Apply the preliminary load response strategy, and monitor the load change situation in real time to obtain real-time monitoring data. Combine the historical data and the real-time monitoring data to predict the load trend. Based on the load trend, adjust the power distribution strategy to optimize the preliminary load response strategy and obtain an intelligent load response strategy. Conduct load dynamic monitoring based on the intelligent load response strategy;

[0111] This step applies the preliminary load response strategy, monitors the load changes in real time through smart meters and other sensors, and obtains real-time monitoring data. For example, the power consumption of each area is recorded every 15 minutes. Combine these real-time monitoring data with historical data, and use time series models to predict future load trends. Based on the predicted load trends, adjust the power distribution strategy to optimize the preliminary load response strategy. For example, if it is predicted that the load in a certain area will increase significantly during peak hours, arrange for power generation stations to increase power generation in advance, or guide users in other areas to increase power consumption during off-peak hours. Finally, the intelligent load response strategy is obtained. This strategy not only considers real-time load changes, but also combines historical data and prediction results to ensure the efficient operation of the power system and the satisfaction of user needs. For example, during peak hours, through real-time pricing mechanisms and demand response programs, residential users are guided to reduce power consumption. During off-peak hours, through automated control strategies, the production equipment of commercial users is automatically started. For industrial users, through long-term stable electricity price contracts, ensure that their production is not affected. Through this intelligent load response strategy, dynamic monitoring and optimized management of the load can be achieved, ensuring the safe and stable operation of the power system.

[0112] In the embodiment of the present invention, through reasonable incentive measures and interactive behavior analysis, the participation and response efficiency of users are significantly improved, promoting the balance between supply and demand; based on the user response prediction results, a scientific and reasonable load response strategy is generated, achieving the optimal allocation of power resources; through real-time monitoring and dynamic adjustment, it is ensured that the power system can flexibly respond to various load changes, improving the adaptability and robustness of the system; by combining historical data and real-time monitoring data, the load trend is accurately predicted, ensuring the scientificity and reliability of the power distribution strategy; through the intelligent load response strategy, the dynamic changes of the load are accurately monitored, ensuring the safe and stable operation of the power system, reducing potential risks, and guaranteeing the security and reliability of the power grid.

[0113] After formulating the demand-side management strategy, to further verify and optimize the effects of these strategies, based on this, the present invention provides a specific embodiment, step 402. Based on the demand-side management strategy, construct a simulation environment to simulate the interactive behaviors among different user groups, obtain the target interactive behavior simulation results, and based on the target interactive behavior simulation results, predict the reactions of different user groups under different incentive measures to obtain the user response prediction results, which specifically include the following steps:

[0114] Step 411: Based on the demand-side management strategy, use a multi-agent reinforcement learning framework to construct a simulation environment to simulate the interactive behaviors among different user groups and obtain the preliminary interactive behavior simulation results;

[0115] Suppose that demand-side management strategies have been formulated, including incentive measures such as electricity bill discounts and subsidies. Next, a simulation environment is constructed using a multi-agent reinforcement learning framework to simulate the interaction behaviors among residential users, commercial users, and industrial users. For example, each user group is modeled as an agent, and these agents adjust their electricity consumption behaviors according to different incentive measures. Specifically, the following parameters are set. Residential users are mainly concerned about electricity costs and reduce electricity consumption during peak hours; commercial users pay more attention to production efficiency and increase electricity consumption during off-peak hours; industrial users require a stable power supply, and long-term contracts provide stable electricity prices. Through multiple simulation experiments, preliminary simulation results of interaction behaviors are obtained. For example, it is found that residential users significantly reduce the use of air conditioners and water heaters during peak hours, while commercial users increase the operating time of production equipment during off-peak hours. These preliminary results provide a basis for subsequent optimization.

[0116] Step 412: Based on the preliminary simulation results of interaction behaviors, apply a generative adversarial network to generate multiple extreme load scenarios, simulate the interaction behaviors of different user groups under different extreme load scenarios, so as to optimize the preliminary simulation results of interaction behaviors and obtain the target simulation results of interaction behaviors;

[0117] In this step, the extreme load scenario refers to the possible extreme power load conditions, such as high load caused by high temperature weather or low load caused by equipment failure. The target simulation results of interaction behaviors refer to the final interaction behavior patterns after optimization, reflecting the actual responses of different user groups under extreme load scenarios.

[0118] This step applies a generative adversarial network to generate multiple extreme load scenarios based on the preliminary simulation results of interaction behaviors. For example, generate extreme situations such as high temperature weather, cold snap, and sudden power outages, and simulate the interaction behaviors of different user groups under these scenarios. Specifically, assume that the following extreme load scenarios are generated. In the high temperature weather scenario, residential users use a large amount of air conditioners, resulting in a sharp increase in power load; in the cold snap scenario, residential users increase the use of heating equipment, and the power load rises significantly; in the sudden power outage scenario: some users rely on backup power supplies, while other users reduce non-essential electricity consumption. Through the simulation of these extreme load scenarios, the preliminary simulation results of interaction behaviors are optimized. For example, it is found that residential users are difficult to significantly reduce electricity consumption even with electricity bill discounts during high temperature weather, so other incentive measures need to be considered, such as providing subsidies for energy storage equipment. Finally, the target simulation results of interaction behaviors are obtained, and these results more comprehensively reflect the actual responses of different user groups in extreme situations.

[0119] Step 413: Based on the target interaction behavior simulation results, analyze the impact of each user group on the smart grid to obtain user impact results, evaluate the actual response patterns of different user groups under different incentive measures to obtain user response evaluation results, and combine the user impact results and the user response evaluation results to obtain user impact assessment results;

[0120] This step analyzes the impact of each user group on the smart grid based on the target interaction behavior simulation results. For example, it is found that the electricity consumption fluctuations of residential users during peak hours have a greater impact on the grid stability, while the long-term stable electricity consumption of industrial users helps to smooth the load curve. At the same time, the actual response patterns of different user groups under different incentive measures are evaluated. For example, residential users are very sensitive to electricity bill discounts, commercial users are more inclined to electricity subsidies during off-peak hours, and industrial users rely on long-term stable electricity price contracts. Combining these analysis results, user impact assessment results are obtained, which provide important basis for subsequent prediction and strategy optimization.

[0121] Step 414: Based on the user impact assessment results, use long short-term memory networks to predict the changing trends of electricity consumption behaviors of different user groups at different time periods, use graph neural networks to analyze the spatial correlations between different user groups, and combine the changing trends of electricity consumption behaviors and the spatial correlations to generate user response prediction results;

[0122] This step uses long short-term memory networks to predict the changing trends of electricity consumption behaviors of different user groups at different time periods based on the above user impact assessment results. For example, through historical data and real-time monitoring data, long short-term memory networks can predict the air conditioner usage of residential users during high-temperature periods in summer, the electricity consumption changes of commercial users during holidays, and the electricity demand of industrial users in different seasons. At the same time, graph neural networks are used to analyze the spatial correlations between different user groups. For example, it is found that there is a strong spatial correlation between residential users and commercial users in certain areas, that is, the reduction of electricity consumption by residential users in one area may drive the increase of electricity consumption by nearby commercial users. Combining the changing trends of electricity consumption behaviors and the spatial correlations, user response prediction results are generated, which not only help us better understand user behavior patterns, but also provide a scientific basis for optimizing power distribution and scheduling.

[0123] In the embodiments of the present invention, through a multi-agent reinforcement learning framework, the interaction behaviors between different user groups are simulated, improving user participation and response efficiency; a generative adversarial network is applied to generate extreme load scenarios, optimizing the preliminary simulation results of interaction behaviors and enhancing the system's adaptability and robustness; combined with the user impact assessment results, a long short-term memory network and a graph neural network are used to accurately predict the changing trend and spatial correlation of users' electricity consumption behaviors, ensuring the scientificity and reliability of the power distribution strategy; based on the detailed user response prediction results, the optimal allocation of power resources is achieved, improving the system's operation efficiency and economic benefits; by simulating and predicting the behaviors of different user groups, the safe and stable operation of the power system is ensured, potential risks are reduced, and the security and reliability of the power grid are guaranteed.

[0124] After the real-time collection and preliminary analysis of power consumption data, to further improve the accuracy and reliability of load forecasting, based on this, the present invention provides a specific embodiment. Step 102: Based on historical data of the same period, construct a time series model. Based on the time series model, generate an expected power load data sequence. Compare the real-time power consumption data sequence with the expected power load data sequence to determine the load trend forecasting result, which specifically includes the following steps:

[0125] Step 201: Based on historical data of the same period, construct a time series model. Use the time series model to predict the expected power load situation within a specific time period to generate an expected power load data sequence;

[0126] In this step, the specific time period refers to a specific future time interval selected according to actual needs, such as several hours, days, or weeks, etc.

[0127] Suppose there is historical power consumption data of a city power grid, covering the working day data of each July in the past three years. To predict the power load situation of a working day in July this year, use this historical data of the same period to construct a time series model. The specific steps are as follows: collect and organize the power consumption data of each working day in July of the past three years. Select a suitable time series model and use the historical data for training and parameter optimization. Based on the trained model, predict the power load situation of a working day in July this year to generate an expected power load data sequence, and this expected power load data sequence will be used as the basis for subsequent comparison and analysis.

[0128] Step 202: Based on historical data of the same period, compare the expected power load data sequence with the real-time power consumption data sequence, identify the verified load fluctuation abnormal patterns, and conduct a cause analysis on the verified load fluctuation abnormal patterns to obtain a cause analysis result;

[0129] In this step, the abnormal load fluctuation pattern refers to the significant difference between the actual power consumption data and the expected power load data, which may indicate the existence of unforeseen external factors or internal system changes. The cause analysis result refers to finding out the causes and influencing factors leading to the abnormality through the analysis of the abnormal pattern.

[0130] In this step, the expected power load data sequence is compared with the real-time power consumption data sequence to identify the abnormal load fluctuation pattern. For example, assume that the real-time power consumption data of a working day in July this year has been collected. Through comparison, we can find that the actual power consumption in some time periods deviates significantly from the expected value. For example, at 08:15, the actual power consumption is 520 kW, while the expected value is 510 kW, indicating that there may be abnormal fluctuations. To further understand the reasons for these abnormal fluctuations, cause analysis is carried out. For example, it is found that the temperature suddenly rises in the morning of that day, resulting in an increase in the air conditioner usage frequency of residential users; or an industrial user undertakes additional production tasks during this period, increasing the power consumption. Through the detailed analysis of these causes, the cause analysis result is obtained, providing a basis for the subsequent model optimization.

[0131] Step 203: Based on the cause analysis result, adjust the key parameters of the time series model to obtain an optimized time series model. Based on the optimized time series model, predict the power load trend to obtain the load trend prediction result, where the prediction result includes the moving prediction result of peak and trough periods and the rising and falling prediction result of the average load level.

[0132] In this step, the key parameters refer to the important parameters in the time series model, such as seasonal components, trend components, etc., which directly affect the prediction accuracy of the model.

[0133] This step adjusts the key parameters of the time series model based on the results of the cause analysis to improve the prediction accuracy. For example, if it is found that the temperature has a significant impact on the electricity load, the temperature variable is introduced as an exogenous variable in the model. In addition, if certain special behavior patterns of users are found, the trend component and seasonal component of the model can be adjusted. Specifically, assume the following key parameters are adjusted: the temperature is introduced as an exogenous variable to capture the impact of weather on the electricity load. The seasonal component is adjusted to better reflect the weekly and daily cyclic variations. The trend component is modified to consider the long-term growth or decline trend. Through these adjustments, an optimized time series model is obtained and used to predict the future electricity load trend. For example, the prediction results show that the moving prediction results for peak and trough periods are that the predicted peak period is advanced from the original 18:00 - 20:00 to 17:30 - 19:30, and the trough period is extended to 3:00 - 6:00 in the early morning. The prediction results for the rise and fall of the average load level are that the average load level is expected to increase by about 5% during high-temperature weather and decrease by about 3% during cold weather. These load trend prediction results can better reflect the future changes in electricity load and also provide a scientific basis for optimizing power distribution and scheduling.

[0134] In the embodiment of the present invention, a time series model is constructed through historical data of the same period, and the model parameters are optimized by combining cause analysis, significantly improving the accuracy of load prediction; through the identification and cause analysis of abnormal patterns of load fluctuations, the adaptability and robustness of the system are enhanced, ensuring reliable predictions in various situations; based on the optimized time series model, the movement of peak and trough periods and the changes in the average load level are accurately predicted, providing a scientific basis for the optimal allocation of power resources; through detailed load trend prediction results, the efficient utilization of power resources is achieved, improving the operating efficiency and economic benefits of the system; by timely identifying and handling abnormal patterns of load fluctuations, potential risks are reduced, ensuring the safe and stable operation of the power grid.

[0135] After generating the expected electricity load data sequence, to further ensure the accuracy and reliability of the prediction, based on this, the present invention provides a specific embodiment, step 202, comparing the expected electricity load data sequence and the real-time electricity consumption data sequence, identifying the verified abnormal patterns of load fluctuations, and performing cause analysis on the verified abnormal patterns of load fluctuations to obtain the cause analysis results, which specifically include the following steps:

[0136] Step 211: Comparing the expected electricity load data sequence and the real-time electricity consumption data sequence to obtain a comparison result, and using statistical methods to perform differential analysis on the comparison result to identify the preliminary abnormal patterns of load fluctuations;

[0137] Suppose that the expected power load data sequence of a certain urban power grid on a working day in July has been generated, and at the same time, the real-time power consumption data of that day has been collected. Compare these two groups of data point by point to obtain the comparison result. Use statistical methods (such as mean, standard deviation) to quantitatively analyze these differences. For example, calculate the mean and standard deviation of the differences in all time periods to determine whether there are significant deviations. If it is found that the differences in some time periods are significantly beyond the normal range (such as exceeding two standard deviations), then identify them as preliminary load fluctuation abnormal patterns. For example, at 08:15, the difference is +10kW, while the average difference is +5kW, and the standard deviation is ±3kW. Therefore, the data point at 08:15 is identified as a preliminary load fluctuation abnormal pattern.

[0138] Step 212: Combine the preliminary load fluctuation abnormal patterns and the target features affecting load changes extracted from historical data. Based on the target features, use the Local Outlier Factor (LOF) algorithm to construct an anomaly detection model. The target features include temperature and humidity.

[0139] In this step, the target features refer to the key factors affecting power load changes, such as temperature, humidity, etc.

[0140] To further verify the preliminarily identified load fluctuation abnormal patterns, it is necessary to combine the target features (such as temperature and humidity) affecting load changes extracted from historical data. Suppose the temperature, humidity, and power consumption data of each day in the past three years have been collected, and relevant features are extracted from them. The specific steps are as follows: Extract temperature and humidity from historical data as target features; Use the Local Outlier Factor (LOF) algorithm to construct an anomaly detection model. The Local Outlier Factor (LOF) algorithm can identify outliers in the data and help us more accurately identify abnormal patterns. Through this algorithm, it can be identified which data points are outliers. For example, it is found that the combination of temperature and humidity at 08:15 on that day led to a significant increase in power consumption, which may be due to residents increasing the air conditioner usage frequency.

[0141] Step 213: Apply the anomaly detection model to detect the real-time power consumption data sequence and identify the verified load fluctuation abnormal patterns.

[0142] In this step, the anomaly detection model is applied to detect the real-time power consumption data sequence and identify the verified load fluctuation abnormal patterns. For example, use the anomaly detection model to detect the real-time power consumption data of that day and find that there is indeed an abnormal fluctuation in power consumption at 08:15. Specifically, the anomaly detection model outputs an anomaly score, indicating the degree of outlier of each data point. According to the set threshold (such as an anomaly score above 2.0), it is confirmed that the power consumption at 08:15 is a verified load fluctuation abnormal pattern.

[0143] Step 214: Combine environmental and social dynamic information to perform an external factor evaluation process on the verified abnormal load fluctuation pattern, and obtain an external factor influence result;

[0144] In this step, environmental and social dynamic information refers to factors that may affect the power load, such as weather forecasts and social event arrangements.

[0145] This step combines the environmental and social dynamic information of the day to conduct an external factor evaluation on the verified abnormal load fluctuation pattern. For example, checking the weather forecast of the day and finding that the temperature has suddenly risen; or looking at the social event arrangements and finding that a large shopping mall is holding a promotional event, attracting a large number of people. By comprehensively analyzing these external factors, an external factor influence result is obtained. For example, the increase in temperature has led to an increase in the air conditioner usage frequency of residential users, and the promotional event in the large shopping mall has increased the electricity consumption of commercial users. The combined effect of these external factors has led to the abnormal load fluctuation at 08:15.

[0146] Step 215: Detect the internal situation of the smart grid, perform an internal factor evaluation process on the verified abnormal load fluctuation pattern, obtain an internal factor influence result, and based on the external factor influence result and the internal factor influence result, conduct a cause analysis on the verified abnormal load fluctuation pattern to obtain a cause analysis result. The internal situation includes equipment status, maintenance records, and user behavior history;

[0147] In this step, the internal situation refers to the state inside the smart grid, such as equipment status, maintenance records, and user behavior history.

[0148] This step evaluates the internal factors of the verified abnormal load fluctuation patterns by detecting the internal situation of the smart grid. For example, by checking the equipment status of the power station, it is found that a generator has been urgently repaired during this period; by checking the maintenance records, it is found that a transmission line has been overhauled the day before; by analyzing the user behavior history, it is found that some industrial users have increased production tasks during this period. Through the analysis of these internal factors, the internal factor impact results are obtained. For example, the emergency repair of a generator may lead to insufficient power supply, the overhaul of the transmission line affects the power transmission efficiency, and the increase of production tasks by industrial users increases the power demand. Combining the external factor impact results and the internal factor impact results, a comprehensive cause analysis is conducted on the verified abnormal load fluctuation patterns to obtain the cause analysis results. For example, the external factor is that the rising temperature causes residential users to increase the frequency of air conditioning use, and the promotion activities of large shopping malls increase the power consumption of commercial users. The internal factors are that the emergency repair of a generator affects the power supply, the overhaul of the transmission line reduces the transmission efficiency, and the increase of production tasks by industrial users increases the power demand. Finally, the cause analysis result showed that the abnormal load fluctuation at 08:15 was mainly caused by the combined effects of external environmental factors (increased temperature, social activities) and internal power grid conditions (equipment maintenance, line maintenance, and user behavior).

[0149] The embodiments of the present invention use statistical methods and local outlier factor algorithms to accurately identify abnormal fluctuation patterns in power loads, thereby improving detection accuracy; combine environmental and social dynamic information to comprehensively evaluate the impact of external factors, thereby enhancing the adaptability and robustness of the system; through a detailed evaluation of the internal power grid conditions, ensure the stable operation of the power system and improve resource utilization efficiency; based on detailed cause analysis results, achieve optimal allocation of power resources and improve the operating efficiency and economic benefits of the system; timely identify and process abnormal load fluctuation patterns, reduce potential risks, and ensure the safe and stable operation of the power grid.

[0150] Figure 2 A schematic diagram of a smart grid load dynamic monitoring system is provided in accordance with an embodiment of the present invention. Figure 2 As shown, the system includes:

[0151] The collection module 21 is used to collect real-time power consumption data and generate a real-time power consumption data sequence;

[0152] A construction module 22 is used to construct a time series model based on historical data of the same period, generate an expected power load data sequence based on the time series model, and compare the real-time power consumption data sequence with the expected power load data sequence to determine a load trend prediction result;

[0153] An evaluation module 23, configured to apply an adaptive dynamic clustering algorithm based on a graph neural network to analyze the spatial distribution characteristics and temporal variation characteristics of the power load based on the load trend prediction result, so as to generate a configuration scheme for the clustering center and the number of clusters. Based on the configuration scheme, in combination with a bilevel programming model and a robust optimization algorithm, determine an optimal power distribution scheme, and evaluate the risk of the optimal power distribution scheme to generate a power distribution strategy;

[0154] A generation module 24, configured to formulate a demand-side management strategy based on game theory based on the power distribution strategy, and simulate the interaction behaviors among different user groups based on the demand-side management strategy to generate an intelligent load response strategy, where the intelligent load response strategy is used to implement load dynamic monitoring.

[0155] Figure 2 The described intelligent power grid load dynamic monitoring system can execute Figure 1 The intelligent power grid load dynamic monitoring method described in the illustrated embodiment, and its implementation principle and technical effects will not be elaborated. For the intelligent power grid load dynamic monitoring system in the above embodiment, the specific manners in which each module and unit perform operations have been described in detail in the embodiment related to the method, and will not be elaborated here.

[0156] In a possible design, Figure 2 The intelligent power grid load dynamic monitoring system of the illustrated embodiment can be implemented as a computing device, such as Figure 3 shown, and the computing device can include a storage component 31 and a processing component 32;

[0157] The storage component 31 stores one or more computer instructions, where the one or more computer instructions are called and executed by the processing component 32.

[0158] The processing component 32 is configured to: collect real-time power consumption data and generate a real-time power consumption data sequence; construct a time series model based on historical data for the same period, generate an expected power load data sequence based on the time series model, compare the real-time power consumption data sequence with the expected power load data sequence to determine a load trend prediction result; based on the load trend prediction result, apply an adaptive dynamic clustering algorithm based on a graph neural network to analyze the spatial distribution characteristics and time variation characteristics of the power load to generate a configuration scheme for the clustering center and the number of clusters, and based on the configuration scheme, combine a bilevel programming model and a robust optimization algorithm to determine an optimal power distribution scheme, evaluate the risk of the optimal power distribution scheme to generate a power distribution strategy; based on the power distribution strategy, formulate a demand-side management strategy based on game theory, and based on the demand-side management strategy, simulate the interaction behaviors among different user groups to generate an intelligent load response strategy, and the intelligent load response strategy is used to achieve dynamic load monitoring.

[0159] Among them, the processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above method.

[0160] The storage component 31 is configured to store various types of data to support the operation of the terminal. The storage component may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, a magnetic disk or an optical disc.

[0161] Of course, the computing device may also necessarily include other components, such as an input / output interface, a display component, a communication component, etc.

[0162] The input / output interface provides an interface between the processing component and the peripheral interface module, and the peripheral interface module may be an output device, an input device, etc.

[0163] The communication component is configured to facilitate communication between the computing device and other devices in a wired or wireless manner, etc.

[0164] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform. In this case, the computing device can refer to a cloud server. The above processing components, storage components, etc. can be basic server resources rented or purchased from a cloud computing platform.

[0165] An embodiment of the present invention also provides a computer storage medium storing a computer program, which when executed by a computer can implement the above Figure 1 intelligent power grid load dynamic monitoring method shown in the embodiment.

[0166] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0167] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0168] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0169] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for dynamic monitoring of smart grid loads, characterized in that: include: Collecting real-time power consumption data and generating a real-time power consumption data sequence; Based on the historical data of the same period, a time series model is constructed, based on the time series model, an expected power load data sequence is generated, and the real-time power consumption data sequence and the expected power load data sequence are compared to determine the load trend prediction result; Based on the load trend prediction results, an adaptive dynamic clustering algorithm based on a graph neural network is applied to analyze the spatial distribution characteristics and time variation characteristics of the power load to generate a configuration scheme of cluster centers and the number of clusters; based on the configuration scheme, an optimal power distribution scheme is determined in combination with a bi-level programming model and a robust optimization algorithm; the risk of the optimal power distribution scheme is evaluated to generate a power distribution strategy; Based on the power distribution strategy, a demand side management strategy based on game theory is formulated. Based on the demand side management strategy, the interactive behaviors between different user groups are simulated to generate an intelligent load response strategy. The intelligent load response strategy is used to realize dynamic load monitoring.

2. The method according to claim 1, characterized in that Based on the load trend prediction results, an adaptive dynamic clustering algorithm based on a graph neural network is applied to analyze the spatial distribution characteristics and time variation characteristics of the power load to generate a configuration scheme of cluster centers and the number of clusters. Based on the configuration scheme, a bi-level programming model and a robust optimization algorithm are combined to determine an optimal power distribution scheme, and the risk of the optimal power distribution scheme is evaluated to generate a power distribution strategy, including: Based on the load trend prediction results, an adaptive dynamic clustering algorithm based on a graph neural network is applied to analyze the connection relationship between the power grid components in the power grid topology structure to obtain the spatial distribution characteristics and time variation characteristics of the power load, and based on the spatial distribution characteristics and the time variation characteristics, a configuration scheme of the cluster center position and the number of clusters is generated; Based on the configuration scheme, a bi-level programming model is used to process the power distribution and scheduling problem to determine a preliminary power distribution scheme, and a robust optimization algorithm is introduced to process the uncertainty and risk factors in the preliminary power distribution scheme to determine the optimal power distribution scheme; Applying Bayesian network technology to construct a causal relationship model, and evaluating the occurrence probability and impact degree of known risk points of the optimal power distribution plan based on the causal relationship model to obtain a first risk assessment result; The Monte Carlo simulation algorithm is used to evaluate the occurrence probability and impact degree of different risk scenarios to obtain a second risk assessment result. The first risk assessment result is combined to determine the target risk assessment result. The reinforcement learning algorithm is applied to optimize the optimal power distribution plan and the target risk assessment result to generate a power distribution strategy.

3. The method according to claim 2, characterized in that Based on the load trend prediction results, an adaptive dynamic clustering algorithm based on a graph neural network is applied to analyze the connection relationship between the power grid components in the power grid topology structure to obtain the spatial distribution characteristics and time variation characteristics of the power load, and based on the spatial distribution characteristics and the time variation characteristics, a configuration scheme of the cluster center position and the number of clusters is generated, including: Based on the load trend prediction result, modeling the connection relationship between the power grid components in the power grid topology structure to obtain a connection relationship model between the power grid components; Based on the connection relationship model, the spatiotemporal characteristics of the power load are obtained in combination with the spatiotemporal attention mechanism to generate a spatiotemporal feature representation, wherein the spatiotemporal feature representation includes the spatial distribution characteristics and time variation characteristics of the power load; Applying a variational autoencoder to perform anomaly detection and load fluctuation pattern recognition processing on the spatiotemporal feature representation to obtain a load recognition result; Based on the load identification result, an adaptive dynamic clustering algorithm and a hierarchical clustering technique are used to configure the cluster center position and the number of clusters, and a configuration scheme for the cluster center position and the number of clusters is generated so that each cluster reflects the load pattern in different areas or time periods of the smart grid.

4. The method according to claim 1, characterized in that: Based on the power distribution strategy, a demand side management strategy based on game theory is formulated. Based on the demand side management strategy, the interactive behaviors between different user groups are simulated to generate an intelligent load response strategy. The intelligent load response strategy is used to realize dynamic load monitoring, including: Based on the power distribution strategy, using game theory to analyze the interactive behaviors among multiple participants, obtaining interactive behavior analysis results, and setting multiple incentive measures. Based on the incentive measures and the interactive behavior analysis results, a demand side management strategy is formulated. The participants include power users, power stations, and power grid operators; Based on the demand-side management strategy, a simulation environment is constructed to simulate the interactive behaviors between different user groups to obtain target interactive behavior simulation results, and based on the target interactive behavior simulation results, the responses of different user groups under different incentives are predicted to obtain user response prediction results, wherein the user groups include residential users, commercial users, and industrial users; Based on the user response prediction results, generating a preliminary load response strategy, the preliminary load response strategy including a real-time pricing mechanism, a demand response plan, and an automated control strategy; Apply the preliminary load response strategy and monitor load changes in real time to obtain real-time monitoring data. Combine historical data with the real-time monitoring data to predict load trends. Adjust the power distribution strategy based on the load trends to optimize the preliminary load response strategy and obtain an intelligent load response strategy. Perform dynamic load monitoring based on the intelligent load response strategy.

5. The method according to claim 4, characterized in that Based on the demand-side management strategy, a simulation environment is constructed to simulate the interactive behaviors between different user groups to obtain target interactive behavior simulation results. Based on the target interactive behavior simulation results, the reactions of different user groups under different incentives are predicted to obtain user reaction prediction results, including: Based on the demand-side management strategy, a simulation environment is constructed using a multi-agent reinforcement learning framework to simulate the interactive behaviors between different user groups and obtain preliminary interactive behavior simulation results; Based on the preliminary interactive behavior simulation results, a generative adversarial network is applied to generate a variety of extreme load scenarios, and the interactive behaviors of different user groups under different extreme load scenarios are simulated to optimize the preliminary interactive behavior simulation results and obtain target interactive behavior simulation results; Based on the target interactive behavior simulation results, analyze the impact of each user group on the smart grid to obtain user impact results, evaluate the actual response patterns of different user groups under different incentives to obtain user reflection evaluation results, and combine the user impact results and the user reflection evaluation results to obtain user impact evaluation results; Based on the user impact assessment results, a long short-term memory network is used to predict the changing trends of electricity consumption behaviors of different user groups in different time periods. A graph neural network is used to analyze the spatial correlations between different user groups. The changing trends of electricity consumption behaviors and the spatial correlations are combined to generate user response prediction results.

6. The method according to claim 1, characterized in that Based on the historical data of the same period, a time series model is constructed, based on the time series model, an expected power load data sequence is generated, and the real-time power consumption data sequence and the expected power load data sequence are compared to determine the load trend prediction result, including: Based on historical data of the same period, a time series model is constructed, and the expected power load situation in a specific time period is predicted using the time series model to generate an expected power load data sequence; Based on historical data of the same period, the expected power load data sequence and the real-time power consumption data sequence are compared to identify a verified abnormal load fluctuation pattern, and a cause analysis is performed on the verified abnormal load fluctuation pattern to obtain a cause analysis result; Based on the cause analysis results, the key parameters of the time series model are adjusted to obtain an optimized time series model. Based on the optimized time series model, the power load trend is predicted to obtain a load trend prediction result, which includes a movement prediction result during peak and valley periods and a rise and fall prediction result of the average load level.

7. The method according to claim 6, characterized in that Comparing the expected power load data sequence with the real-time power consumption data sequence, identifying a verified abnormal load fluctuation pattern, and performing a cause analysis on the verified abnormal load fluctuation pattern to obtain a cause analysis result, including: Comparing the expected power load data sequence with the real-time power consumption data sequence to obtain a comparison result, performing difference analysis on the comparison result using a statistical method, and identifying a preliminary load fluctuation abnormal pattern; Combining the preliminary load fluctuation abnormal pattern with the target features affecting load changes extracted from historical data, and building an abnormality detection model based on the target features using a local outlier factor algorithm, the target features including temperature and humidity; Applying the anomaly detection model to detect the real-time power consumption data sequence to identify a verified load fluctuation anomaly pattern; Combined with environmental and social dynamic information, external factor evaluation and processing are performed on the verified abnormal load fluctuation pattern to obtain the external factor impact result; Detect the internal situation of the smart grid, perform internal factor evaluation on the verified abnormal load fluctuation pattern to obtain internal factor impact results, perform cause analysis on the verified abnormal load fluctuation pattern based on the external factor impact results and the internal factor impact results to obtain cause analysis results, wherein the internal situation includes equipment status, maintenance records, and user behavior history.

8. A smart grid load dynamic monitoring system, characterized in that: include: A collection module, used to collect real-time power consumption data and generate a real-time power consumption data sequence; A construction module is used to construct a time series model based on historical data of the same period, generate an expected power load data sequence based on the time series model, and compare the real-time power consumption data sequence with the expected power load data sequence to determine a load trend prediction result; An evaluation module is used to apply an adaptive dynamic clustering algorithm based on a graph neural network based on the load trend prediction result to analyze the spatial distribution characteristics and time variation characteristics of the power load to generate a configuration scheme of cluster centers and the number of clusters, and based on the configuration scheme, combined with a bi-level programming model and a robust optimization algorithm, determine an optimal power distribution scheme, evaluate the risk of the optimal power distribution scheme, and generate a power distribution strategy; A generation module is used to formulate a demand side management strategy based on game theory based on the power distribution strategy, simulate the interactive behavior between different user groups based on the demand side management strategy, and generate an intelligent load response strategy, which is used to realize dynamic load monitoring.

9. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a smart grid load dynamic monitoring method as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, a smart grid load dynamic monitoring method as described in any one of claims 1 to 7 is implemented.

Images