Power system load regulation method based on intelligent demand response optimization

Abstract

The invention discloses a power system load regulation method based on intelligent demand response optimization, and relates to the technical field of power grid load regulation, and the method comprises the steps: obtaining the operation data of a power system in real time; based on the collected operation data, external data are introduced and integrated, and load demand prediction of the power system is generated through an intelligent algorithm; constructing a dynamic optimization model in combination with the power grid dispatching instruction and the load prediction result; based on the dynamic optimization model, adopting a model prediction control and reinforcement learning hybrid optimization method to obtain an optimal load regulation instruction; transmitting the optimization instruction to a scheduling node, and adjusting load distribution and generating capacity of a power system and charging and discharging control of energy storage equipment; and based on adjustment of a control instruction, monitoring an adjustment effect in real time, and dynamically adjusting model parameters and optimization strategies of model prediction control through reinforcement learning. According to the method, the load regulation precision and response speed are improved, the dynamic optimization model comprehensively considers the system stability and the energy efficiency, and the automation level is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid load regulation, and specifically to a method for regulating the load of a power system based on intelligent demand response optimization. Background Art

[0002] With the gradual development of the power system towards intelligence and automation, traditional power load regulation methods can no longer meet the growing power demand and the increasingly complex operating conditions of the power system. Especially in the context of the access of large-scale renewable energy, the load fluctuations of the power grid have intensified, and traditional load regulation methods face many challenges, such as slow response speed, low regulation accuracy, and low energy utilization efficiency. Therefore, how to improve the intelligent level of power system load regulation, achieve rapid response, precise regulation, and efficient energy utilization has become a key issue in power system dispatching.

[0003] The load regulation method based on intelligent demand response combines advanced technologies such as big data, artificial intelligence, and the Internet of Things, and can obtain the operating data of the power system in real time. Through data analysis and intelligent algorithms, load demand prediction and optimization regulation are carried out. These technologies can help the power system optimize resource allocation, improve energy utilization efficiency, reduce energy waste, and ensure the safety and stability of system operation. Especially the hybrid optimization method of reinforcement learning and model predictive control can achieve adaptive load regulation, enabling the power system to maintain good regulation performance and stability under dynamically changing load demands and external conditions. Summary of the Invention

[0004] To solve the above technical problems, a method for regulating the load of a power system based on intelligent demand response optimization is provided, and the present technical solution solves the above problems.

[0005] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] A method for regulating the load of a power system based on intelligent demand response optimization includes:

[0007] Deploy distributed sensors at the nodes of the power system to form a distributed sensor network, and obtain the operating data of the power system in real time;

[0008] Based on the collected operating data, introduce and integrate external data, and generate a load demand prediction for the power system through intelligent algorithms;

[0009] Combined with the power grid dispatching instructions and the load prediction results, with the goal of the highest system operation stability, the shortest regulation response time, and the optimal energy utilization efficiency, construct a dynamic optimization model;

[0010] Based on a dynamic optimization model, a hybrid optimization method combining model predictive control and reinforcement learning is adopted to obtain the optimal load regulation command;

[0011] The optimization command is transmitted to the scheduling node to adjust the load distribution, power generation, and charge and discharge control of energy storage devices in the power system;

[0012] Based on the regulation of the control command, the regulation effect is monitored in real time, and the model parameters and optimization strategies of the model predictive control are dynamically adjusted through reinforcement learning to achieve adaptive control of the load regulation in the power system.

[0013] Preferably, it is characterized in that the specific steps of forming a distributed sensor network by deploying distributed sensors at the nodes of the power system and obtaining the operation data of the power system in real time include:

[0014] Receiving the data collected by the sensors, performing preliminary processing on the sensor data, including denoising and filtering, and caching the data;

[0015] Based on the wireless communication, each sensor node is connected to the data aggregation platform, and the data is uploaded to the central monitoring system in real time through the communication network for monitoring and management;

[0016] The data is stored through the database, and the data from different sensors is fused.

[0017] Preferably, the specific steps of introducing and integrating external data based on the collected operation data and generating the load demand prediction of the power system through intelligent algorithms include:

[0018] The external data includes meteorological data, date type, season factors, and holiday information;

[0019] Based on the correlation analysis method, the features that have an important impact on the load demand prediction are extracted from the above operation data and external data, and the features with high correlation with the load demand are screened out;

[0020] According to the above operation data, external data, and load demand features, a neural network model is constructed;

[0021] The neural network model is trained using historical data, and the real-time collected operation data and external data are input into the trained model. The model predicts the grid load demand in the short term in the future according to the learned patterns and rules.

[0022] Preferably, the specific steps of extracting the features that have an important impact on the load demand prediction from the above operation data and external data based on the correlation analysis method and screening out the features with high correlation with the load demand include:

[0023] Among them, the calculation formula of the correlation analysis method is:

[0024]

[0025] Where ρ zc is the Pearson correlation coefficient, representing the degree of linear correlation between the feature and the load demand. n is the number of samples, f v and c v are the v-th observed values of the feature z and the load demand c respectively. and are the sample means of the feature z and the load demand c respectively. The value of ρ zc ranges from -1 to 1. The closer the absolute value is to 1, the stronger the linear correlation between the two variables; the closer it is to 0, the weaker the linear correlation.

[0026] Preferably, combining the power grid dispatching instructions and the load forecasting results, with the goal of maximizing the system operation stability, minimizing the regulation response time, and optimizing the energy utilization efficiency, constructing the dynamic optimization model specifically includes:

[0027] Using the weighted summation method to construct the dynamic optimization model to balance the system stability, regulation response time, and energy utilization efficiency. The calculation formula of the dynamic optimization model is:

[0028]

[0029] Where J is the value of the dynamic optimization model, w1, w2, and w3 are the weight coefficients of the three objectives of system stability, regulation response time, and energy utilization efficiency respectively. S is the system stability, T is the regulation response time, E is the energy utilization efficiency, L t is the target load of the power grid dispatching instruction, L a is the actual value of the power grid load demand, H is the target load required by the power grid dispatching instruction, L a is the actual value of the power grid load demand, L is the energy loss generated during the system operation, and f is the error function.

[0030] Preferably, based on the dynamic optimization model, adopting a hybrid optimization method of model predictive control and reinforcement learning to obtain the optimal load regulation instruction specifically includes:

[0031] According to the operating characteristics of the power system, establish a state space model to describe the dynamic behavior of the power system. The state equation of the system is:

[0032] x(t + 1) = f(x(t), u(t))

[0033] Where f is a non-linear function, which can be estimated according to the historical operation data of the power system through the system identification method. u(t) is the state vector of the system at time t, x(t) is the control vector at time t, and x(t + 1) is the state of the system at time (t + 1).

[0034] Based on the state - space model, predict the system state at future moments. Among them, the prediction formula is:

[0035]

[0036] In the formula, represents the system state at the predicted moment at time t + k, is the state variable of the system at time step t + k - 1, is the predicted control input, and f is a linear function;

[0037] Take the previously constructed dynamic optimization model as the objective function of model predictive control, and consider the cumulative optimization effect in the next few moments:

[0038] is expressed as taking the dynamic optimization model as the objective function of model predictive control, is the actual load of the power system, is the target load required by the grid dispatching instruction, is the actual power output of the power system, and is the system energy efficiency loss index.

[0039] Preferably, based on the dynamic optimization model, a hybrid optimization method of model predictive control and reinforcement learning is adopted to obtain the optimal load regulation instruction, which specifically includes:

[0040] Among them, the calculation formula of the deep Q - network of the reinforcement learning algorithm is:

[0041]

[0042] In the formula, Q(s t ,a t ) is the Q - value when choosing action a t under state s t , s t is the state at the current moment, a t is the action taken at time t, α represents the learning rate, r t represents the immediate reward obtained at time t, γ represents the discount factor, represents the maximum Q - value corresponding to all possible actions a in the next state s t+1 ;

[0043] Use the mean - square error as the loss function to update the parameters of the Q - network.

[0044] Preferably, the transmission of the optimization instruction to each scheduling node to adjust the load distribution, power generation, and charge - discharge control of energy storage devices in the power system specifically includes:

[0045] The control instruction is transmitted from the control system to each scheduling node through the communication network;

[0046] Feedback regulation requires real-time monitoring of the operation data of each node in the power system during the regulation process to provide feedback information for the control system;

[0047] The control system adjusts the optimization instructions according to the feedback information to form a closed-loop control, continuously optimizing the control effect and improving the stability and reliability of the power system.

[0048] Preferably, the regulation based on control instructions and real-time monitoring of the regulation effect specifically includes:

[0049] Within each control cycle, the monitoring system collects the operation data of the power system and calculates the monitoring indicators of the regulation effect. The monitoring indicators include power system load indicators, energy consumption indicators, equipment operation status indicators, and environmental indicators.

[0050] Preferably, the dynamic adjustment of the model parameters and optimization strategy of model predictive control through reinforcement learning to achieve the adaptive regulation of the power system specifically includes:

[0051] Select an action according to the current state and adjust the model parameters and optimization strategy of model predictive control;

[0052] Model predictive control generates new load regulation instructions according to the adjusted model parameters and optimization strategy, transmits them to each scheduling node, and regulates the load distribution, power generation, and charge and discharge control of energy storage devices in the power system;

[0053] Repeat the steps to form a closed-loop adaptive regulation process, enabling the power system to maintain good load regulation performance and system stability under different operating conditions.

[0054] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0055] The present invention can improve the load regulation accuracy and response speed by obtaining power system data in real time through a distributed sensor network and analyzing it in combination with external factors. The dynamic optimization model comprehensively considers system stability and energy efficiency, improves energy utilization on the premise of ensuring stability, and realizes adaptive control through a hybrid optimization method of reinforcement learning and model predictive control, adjusts the regulation strategy in real time, enhances system stability and reliability. In addition, intelligent demand response optimization enhances the flexibility and adaptability of the system, reduces human intervention, improves the automation level, and reduces the operation complexity and error probability. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 is the step flow framework diagram of the present invention;

[0057] Figure 2 is the step flow framework diagram of load demand prediction in the present invention. DETAILED DESCRIPTION

[0058] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and those skilled in the art can think of other obvious variations.

[0059] Referring to Figure 1 As shown, a power system load regulation method based on intelligent demand response optimization includes:

[0060] Deploying distributed sensors at the nodes of the power system to form a distributed sensor network and obtaining the operation data of the power system in real time;

[0061] Based on the collected operation data, introducing and integrating external data, and generating a load demand prediction for the power system through intelligent algorithms;

[0062] Combining the grid dispatching instructions and the load prediction results, and constructing a dynamic optimization model with the goals of the highest system operation stability, the shortest regulation response time, and the optimal energy utilization efficiency;

[0063] Based on the dynamic optimization model, adopting a hybrid optimization method of model predictive control and reinforcement learning to obtain the optimal load regulation instruction;

[0064] Transmitting the optimization instruction to the dispatching node to adjust the load distribution, power generation amount, and charge and discharge control of energy storage devices in the power system;

[0065] Based on the adjustment of the control instruction, real-time monitoring of the adjustment effect, and dynamically adjusting the model parameters and optimization strategy of the model predictive control through reinforcement learning to achieve the adaptive control of the power system load regulation.

[0066] Deploying distributed sensors at the nodes of the power system to form a distributed sensor network and obtaining the operation data of the power system in real time specifically includes:

[0067] Receiving the data collected by the sensors, preliminarily processing the sensor data, including denoising and filtering, and caching the data;

[0068] Based on the wireless communication to connect each sensor node with the data aggregation platform, and the data is uploaded to the central monitoring system in real time through the communication network for monitoring and management;

[0069] Storing the data through the database and fusing the data from different sensors;

[0070] Adopting wireless communication and data fusion technologies can obtain the operation state data of the power system from multiple angles in real time, improve the accuracy and reliability of the data, and provide high-quality data support for subsequent load demand prediction and dynamic optimization.

[0071] Referring to Figure 2As shown, based on the collected operation data, external data is introduced and integrated, and the load demand prediction of the power system is generated through intelligent algorithms, specifically including:

[0072] The external data includes meteorological data, date types, seasonal factors, and holiday information;

[0073] Based on the correlation analysis method, features that have an important impact on load demand prediction are extracted from the above operation data and external data, and features with high correlation with load demand are screened out;

[0074] According to the above operation data, external data, and load demand characteristics, a neural network model is constructed;

[0075] The neural network model is trained using historical data, and the real-time collected operation data and external data are input into the trained model. The model predicts the grid load demand in the short term in the future according to the learned patterns and rules;

[0076] By integrating external data and making predictions based on a neural network, the limitations of traditional load prediction methods are broken through, and big data and artificial intelligence technologies are used to improve the prediction accuracy, thereby optimizing power dispatching and load distribution.

[0077] Based on the correlation analysis method, features that have an important impact on load demand prediction are extracted from the above operation data and external data, and features with high correlation with load demand are screened out, specifically including:

[0078] Among them, the calculation formula of the correlation analysis method is:

[0079]

[0080] In the formula, ρ zc is the Pearson correlation coefficient, which represents the linear correlation degree between the feature and the load demand. n is the number of samples, z v and c v are the v-th observed values of the feature z and the load demand c respectively, and are the sample means of the feature z and the load demand c respectively. The value of ρ zc ranges from -1 to 1. The closer the absolute value is to 1, the stronger the linear correlation between the two variables; the closer to 0, the weaker the linear correlation;

[0081] Innovatively, the Pearson correlation coefficient analysis method is adopted to screen out key features from a large amount of external and system data, improving the accuracy and credibility of load demand prediction, thereby providing accurate reference data for system dispatching.

[0082] Combined with the power grid dispatching instructions and the load forecasting results, a dynamic optimization model is constructed with the goals of maximizing the system operation stability, minimizing the regulation response time, and optimizing the energy utilization efficiency. Specifically, it includes:

[0083] A weighted summation method is used to construct a dynamic optimization model to balance the system stability, regulation response time, and energy utilization efficiency. The calculation formula of the dynamic optimization model is:

[0084]

[0085] In the formula, J is the value of the dynamic optimization model, w1, w2, and w3 are the weight coefficients of the three goals of system stability, regulation response time, and energy utilization efficiency respectively. S is the system stability, T is the regulation response time, E is the energy utilization efficiency, L t is the target load of the power grid dispatching instruction, L a is the actual value of the power grid load demand, H is the target load required by the power grid dispatching instruction, L a is the actual value of the power grid load demand, L is the energy loss generated during the system operation, f is the error function;

[0086] This method proposes a weighted summation optimization model that comprehensively considers the power grid dispatching instructions, load forecasting results, and dynamic optimization goals, providing a multi-objective balanced solution for power system load regulation.

[0087] Based on the dynamic optimization model, a hybrid optimization method of model predictive control and reinforcement learning is adopted to obtain the optimal load regulation instruction. Specifically, it includes:

[0088] According to the operating characteristics of the power system, a state space model is established to describe the dynamic behavior of the power system. The state equation of the system is:

[0089] x(t + 1) = f(x(t), u(t))

[0090] In the formula, f is a non-linear function, which can be estimated according to the historical operation data of the power system through system identification methods. u(t) is the state vector of the system at time t, x(t) is the control vector at time t, and x(t + 1) is the state of the system at time (t + 1);

[0091] Based on the state space model, the system state at future times is predicted. Among them, the prediction formula is:

[0092]

[0093] In the formula, represents the system state at the predicted time at time t + k, the state variable of the system at time step t + k - 1, is the predicted control input, and f is a linear function;

[0094] Take the previously constructed dynamic optimization model as the objective function of model predictive control, and consider the cumulative optimization effect in the next few moments:

[0095] It is expressed that the dynamic optimization model is used as the objective function of model predictive control. \(P_{act}\) is the actual load of the power system, \(P_{ref}\) is the target load required by the grid dispatching instruction, \(P_{out}\) is the actual power output of the power system, and \(\eta\) is the system energy efficiency loss index;

[0096] Innovatively combine model predictive control and reinforcement learning technology. Predict future load demands through a dynamic optimization model to achieve intelligent and adaptive control of load regulation commands, minimizing energy losses and system instability to the greatest extent.

[0097] Based on the dynamic optimization model, adopt a hybrid optimization method of model predictive control and reinforcement learning to obtain the optimal load regulation command, which specifically includes:

[0098] Among them, the calculation formula of the deep Q-network of the reinforcement learning algorithm is:

[0099]

[0100] In the formula, \(Q(s t ,a t ) is the Q value when choosing action a t in state s t . s t is the current state, a t is the action taken at time t, \(\alpha\) represents the learning rate, r t represents the immediate reward obtained at time t, \(\gamma\) represents the discount factor, represents the maximum Q value corresponding to all possible actions a in the next state s t+1 ;

[0101] Use the mean square error as the loss function to update the parameters of the Q network;

[0102] Propose the deep Q-network in the reinforcement learning algorithm to optimize the load regulation decision. This method can adjust the regulation strategy according to the dynamic state of the power system, overcome the limitations of traditional methods, and improve the learning ability and self-adaptability of the system.

[0103] Transmit the optimized instructions to each dispatching node to adjust the load distribution, power generation, and charge and discharge control of energy storage devices in the power system, which specifically includes:

[0104] The control instructions are transmitted from the control system to each dispatching node through a communication network;

[0105] Feedback regulation requires real-time monitoring of the operating data of each node in the power system during the regulation process to provide feedback information for the control system;

[0106] The control system adjusts the optimization instructions according to the feedback information to form a closed-loop control, continuously optimizing the control effect and improving the stability and reliability of the power system.

[0107] The regulation based on the control instructions and real-time monitoring of the regulation effect specifically includes:

[0108] In each control cycle, the monitoring system collects the operating data of the power system and calculates the monitoring indicators of the regulation effect. The monitoring indicators include power system load indicators, energy consumption indicators, equipment operating status indicators, and environmental indicators.

[0109] The dynamic adjustment of the model parameters and optimization strategy of model predictive control through reinforcement learning to achieve the adaptive regulation of the power system specifically includes:

[0110] Select an action according to the current state and adjust the model parameters and optimization strategy of model predictive control;

[0111] Model predictive control generates new load regulation instructions according to the adjusted model parameters and optimization strategy, transmits them to each dispatching node, and regulates the load distribution, power generation amount, and charge and discharge control of energy storage devices in the power system;

[0112] Repeat the steps to form a closed-loop adaptive regulation process, enabling the power system to maintain good load regulation performance and system stability under different operating conditions.

[0113] In summary, the advantages of the present invention are as follows:

[0114] By obtaining the operating data of the power system in real time through a distributed sensor network and comprehensively analyzing it in combination with external data, it is possible to more accurately predict the grid load demand. At the same time, the load regulation instructions optimized based on intelligent algorithms can effectively improve the regulation response speed, enabling the power system to respond more quickly to load fluctuations;

[0115] By constructing a dynamic optimization model and comprehensively considering multiple factors such as system stability, regulation response time, and energy utilization efficiency, it is possible to improve the energy utilization efficiency and reduce unnecessary energy waste on the premise of ensuring system stability;

[0116] The hybrid optimization method of reinforcement learning and model predictive control can real-time monitor the load regulation effect and dynamically adjust the model parameters and optimization strategy according to the actual situation to ensure that the power system can maintain the optimal load regulation performance in various operating environments;

[0117] By continuously optimizing the load regulation command and adopting a closed-loop control strategy, and constantly adjusting and optimizing the load distribution, power generation, and charge and discharge control of energy storage devices in the power system, the operation stability and reliability of the power system can be improved, and the overload problem that occurs in the power system during peak loads can be reduced;

[0118] Through the intelligent demand response optimization method, the power system can flexibly adjust its operation strategy according to changes in external conditions and power demand, making the system more flexible and adaptable, and capable of coping with different load demands and external disturbances;

[0119] This method can automatically optimize the power load regulation process by integrating advanced intelligent algorithms and automated control technologies, reduce human intervention, improve the automation level of the system, and reduce the complexity and error probability of manual operations.

[0120] The above has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art of this industry should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection required by the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for regulating the load of a power system based on intelligent demand response optimization, characterized in that, Including: Deploying distributed sensors at nodes of the power system to form a distributed sensor network, and acquiring the operation data of the power system in real time; Based on the acquired operation data, introducing and integrating external data, and generating a load demand prediction of the power system through intelligent algorithms; Combining grid dispatching instructions and load prediction results, and constructing a dynamic optimization model with the goals of the highest system operation stability, the shortest adjustment response time, and the optimal energy utilization efficiency; Based on the dynamic optimization model, adopting a hybrid optimization method of model predictive control and reinforcement learning to obtain the optimal load adjustment instruction; Transmitting the optimization instruction to the dispatching node to adjust the load distribution, power generation amount, and charge and discharge control of energy storage devices of the power system; Based on the adjustment of the control instruction, monitoring the adjustment effect in real time, and dynamically adjusting the model parameters and optimization strategies of the model predictive control through reinforcement learning to achieve the adaptive control of the load adjustment of the power system.

2. The method for regulating the load of a power system based on intelligent demand response optimization according to claim 1, wherein The specific steps of deploying distributed sensors at nodes of the power system to form a distributed sensor network and acquiring the operation data of the power system in real time include: Receiving the data collected by the sensors, performing preliminary processing on the sensor data, including denoising and filtering, and caching the data; Connecting each sensor node to the data aggregation platform based on wireless communication, and uploading the data to the central monitoring system through the communication network in real time for monitoring and management; Storing the data through a database and fusing the data from different sensors.

3. The method for regulating the load of a power system optimized based on intelligent demand response according to claim 2, wherein, The specific steps of introducing and integrating external data based on the acquired operation data and generating a load demand prediction of the power system through intelligent algorithms include: The external data includes meteorological data, date type, seasonal factors, and holiday information; Extracting the features that have an important impact on the load demand prediction from the above operation data and external data based on the correlation analysis method, and screening out the features with high correlation with the load demand; Constructing a neural network model according to the above operation data, external data, and load demand features; Training the neural network model with historical data, inputting the real-time acquired operation data and external data into the trained model, and the model predicts the grid load demand in the short term in the future according to the learned patterns and rules.

4. The method for regulating the power system load based on intelligent demand response optimization according to claim 3, wherein, The specific steps of extracting the features that have an important impact on the load demand prediction from the above operation data and external data based on the correlation analysis method and screening out the features with high correlation with the load demand include: Among them, the calculation formula of the correlation analysis method is: Where ρ zc is the Pearson correlation coefficient, representing the degree of linear correlation between the feature and the load demand. n is the number of samples, z v and c v are the v-th observed values of the feature z and the load demand c respectively. and are the sample means of the feature z and the load demand c respectively. The value of ρ zc ranges from -1 to 1. The closer the absolute value is to 1, the stronger the linear correlation between the two variables. The closer it is to 0, the weaker the linear correlation.

5. The load regulation method of the power system optimized based on intelligent demand response according to claim 4, characterized in that The specific steps of combining grid dispatching instructions and load prediction results and constructing a dynamic optimization model with the goals of the highest system operation stability, the shortest adjustment response time, and the optimal energy utilization efficiency include: Adopting the method of weighted summation to construct a dynamic optimization model to balance system stability, adjustment response time, and energy utilization efficiency, where the calculation formula of the dynamic optimization model is: Wherein, J is the value of the dynamic optimization model, w1, w2, and w3 are the weight coefficients of the three objectives of system stability, regulation response time, and energy utilization efficiency respectively, S is system stability, T is regulation response time, E is energy utilization efficiency, L t is the target load of the grid dispatching instruction, L a is the actual value of the grid load demand, H is the target load required by the grid dispatching instruction, L a is the actual value of the grid load demand, L is the energy loss generated during the operation of the system, and f is the error function.

6. The load regulation method for a power system optimized based on intelligent demand response according to claim 5, characterized in that The specific steps of adopting a hybrid optimization method of model predictive control and reinforcement learning based on the dynamic optimization model to obtain the optimal load adjustment instruction include: According to the operation characteristics of the power system, establishing a state space model to describe the dynamic behavior of the power system, and the state equation of the system is: x(t + 1) = f(x(t), u(t)) In the formula, f is a non-linear function, which can be estimated by system identification method according to the historical operation data of the power system. u(t) is the state vector of the system at time t, x(t) is the control vector at time t, and x(t + 1) is the state of the system at time (t + 1). Based on the state space model, predict the system state at future times. Among them, the prediction formula is: wherein, represents the system state at the predicted time at time t + k, is the state variable of the system at time step t + k - 1, is the predicted control input, and f is a linear function; Take the previously constructed dynamic optimization model as the objective function of model predictive control, and consider the cumulative optimization effect in the next few moments: It is expressed that the dynamic optimization model is used as the objective function of model predictive control. is the actual load of the power system, is the target load required by the grid dispatching instruction, is the actual power output of the power system, and is the system energy efficiency loss index.

7. The load regulation method for a power system optimized based on intelligent demand response according to claim 6, characterized in that, Based on the dynamic optimization model, adopt a hybrid optimization method of model predictive control and reinforcement learning to obtain the optimal load regulation instruction. Specifically Include: Among them, the calculation formula of the deep Q network of the reinforcement learning algorithm is: where Q(s t , a t ) is the Q-value when choosing action a t in state s t , s t is the state at the current moment, a t is the action taken at time t, α represents the learning rate, r t represents the immediate reward obtained at time t, γ represents the discount factor, represents the maximum Q-value corresponding to all possible actions a in the next state s t+1 ; Use the mean square error as the loss function to update the parameters of the Q network.

8. The method for regulating the load of a power system based on intelligent demand response optimization according to claim 7, wherein The transmission of the optimization instruction to each scheduling node to adjust the load distribution, power generation and charge-discharge control of energy storage devices of the power system specifically includes: The control instruction is transmitted from the control system to each scheduling node through the communication network; Feedback regulation. During the regulation process, it is necessary to monitor the operation data of each node of the power system in real time to provide feedback information for the control system; The control system adjusts the optimization instruction according to the feedback information to form a closed-loop control, continuously optimize the control effect, and improve the stability and reliability of the power system.

9. The method for regulating the load of a power system based on intelligent demand response optimization according to claim 8, wherein The real-time monitoring of the regulation effect based on the control instruction specifically includes: In each control cycle, the monitoring system collects the operation data of the power system and calculates the monitoring indicators of the regulation effect. The monitoring indicators include power system load indicators, energy consumption indicators, equipment operation status indicators, and environmental indicators.

10. The method for regulating the load of a power system based on intelligent demand response optimization according to claim 9, characterized in that, The dynamic adjustment of the model parameters and optimization strategy of model predictive control through reinforcement learning to achieve the adaptive regulation of the power system specifically includes: Select an action according to the current state and adjust the model parameters and optimization strategy of model predictive control; Model predictive control generates a new load regulation instruction according to the adjusted model parameters and optimization strategy, transmits it to each scheduling node, and adjusts the load distribution, power generation and charge-discharge control of energy storage devices of the power system; Repeat the steps to form a closed-loop adaptive regulation process, so that the power system can maintain good load regulation performance and system stability under different operating conditions.

Images