Virtual power plant dynamic aggregation method based on user power generation and utilization complementary characteristics

Abstract

The invention relates to the technical field of power demand response, and discloses a user power generation and utilization complementary characteristic-based virtual power plant dynamic aggregation method, which comprises the following steps of: predicting photovoltaic power generation capacity by using an LSTM (Long Short Term Memory) model to provide a basis for scheduling; air conditioner load response is optimized through reinforcement learning, the load is reduced in the peak period, and the user comfort is guaranteed; a genetic algorithm is adopted to optimize a charging and discharging strategy of the energy storage system, and peak load shifting and cost minimization are achieved. And the power generation and utilization complementary relationship among the users is identified through clustering analysis, and the aggregation relationship is dynamically adjusted, so that the optimal configuration of resources is achieved. The method improves the operation efficiency of the virtual power plant, has high reliability and economic benefits, and provides powerful support for load optimization scheduling.

Description

Technical Field

[0001] The present invention relates to the technical field of power demand response, and in particular to a virtual power plant dynamic aggregation method based on the complementary characteristics of power generation and consumption by users. Background Art

[0002] A virtual power plant is an energy aggregation platform based on the Internet. Through digital technology and the power market mechanism, various distributed energy resources are centrally managed and optimized to achieve flexible allocation and efficient use of resources. The core of virtual power plant technology is to establish a platform that aggregates various distributed energy devices and achieves centralized management and optimization of various energy devices through digital technology. This platform needs to have functions such as data collection, data processing, equipment control and optimization, and also needs to have the ability to interact with the power market.

[0003] With the rapid development of distributed energy in the power system, photovoltaic power generation, energy storage systems and user load control have gradually become key components in virtual power plant technology. However, photovoltaic power generation is uncertain due to the influence of weather and time, resulting in large fluctuations in its output power; energy storage systems can balance supply and demand fluctuations through charging and discharging, but they still face challenges in strategy optimization; intelligent response of air conditioning load can achieve load reduction during peak hours, but a balance needs to be struck between user comfort and energy saving goals.

[0004] Chinese patent document CN118396284A discloses a "dynamic aggregation method of virtual power plants based on dynamic clustering". It includes the following steps: clustering the historical time series data of energy nodes to obtain multiple initial clusters, the historical time series data includes multiple data points, and the data points include power generation, voltage and location; training the data points in the initial cluster with a hidden Markov chain to obtain the transition probability of the data points; according to the transition probability, calculating the dynamic optimization factor to optimize the dynamic clustering; obtaining the required voltage and required power of each preset power node, determining the classification cluster that matches the power supply information through the classification cluster center point in the fuzzy clustering model, traversing the energy nodes in the classification cluster, and determining the energy node that matches the power supply information of the power node and is closest to the power node as the matching energy node. The above technical solution has low matching efficiency and does not take into account the complementary relationship between power generation and power consumption between users, making it difficult to achieve dynamic scheduling and optimal configuration of resources. Summary of the invention

[0005] The present invention mainly solves the technical problems that the original matching efficiency is low and the complementary relationship between power generation and consumption among users is not considered, making it difficult to realize dynamic scheduling and optimal configuration of resources. It provides a virtual power plant dynamic aggregation method based on the complementary characteristics of power generation and consumption among users, and uses an AI prediction model in combination with weather and historical data to realize short-term prediction of photovoltaic power generation to provide support for scheduling; uses a reinforcement learning algorithm to control the air-conditioning load to ensure that the load is reduced during peak power periods while maintaining user comfort; uses a genetic algorithm to optimize the charging and discharging strategy of the energy storage system to achieve the goals of peak shaving and valley filling and cost minimization, and finally identifies the complementary relationship between power generation and consumption among users through cluster analysis to realize dynamic scheduling and optimal configuration of resources, thereby improving the overall efficiency and economy of the system.

[0006] The above technical problem of the present invention is mainly solved by the following technical solution: The present invention comprises the following steps: S1. Make short-term forecasts of photovoltaic power generation based on weather and historical power generation data; S2. Dynamically control air conditioning operation to reduce load during peak hours; S3. Dynamically adjust the charging and discharging strategy of the energy storage system to achieve peak load shifting and cost minimization; S4. Dynamically identify the complementary relationship between power generation and consumption among users, perform cluster analysis on the time series data of power generation and consumption of users, establish power generation sequence and power consumption sequence respectively, and for each user, By setting the threshold Compare and judge the complementary characteristics of power generation and consumption between users i and j; S5. Adjust the aggregation relationship between users in real time to achieve dynamic scheduling and optimal configuration of power generation and consumption resources.

[0007] Preferably, short-term prediction includes obtaining historical power generation data, weather forecast data and time characteristics of the photovoltaic system, performing preprocessing steps including standardization and normalization on the input data, and constructing effective time series features; using machine learning or deep learning models suitable for time series prediction to process nonlinear data and capture time dependencies.

[0008] As a preference, historical data is used to train the model and optimize parameters to improve the accuracy of the prediction. In actual operation, the latest weather data and time characteristics are used as input into the model to output the predicted value of photovoltaic power generation in the future in real time. .

[0009] Preferably, the dynamic control of air-conditioning operation specifically includes designing a reward function R to balance user comfort and energy-saving goals, and using a reinforcement learning model to model the air-conditioning demand response problem as a Markov decision process, defined as (S, A, P, R, γ), where S is the state space, including temperature, humidity and load information; A is the action space, including the adjustment operation of the air conditioner; P is the state transition probability, which is approximately learned through experience replay and training process in reinforcement learning; R is the reward function; and γ is the discount factor.

[0010] As a preference, a genetic algorithm is used to design a charging and discharging strategy for energy storage to balance supply and demand fluctuations and reduce electricity costs. The objective function is to minimize the cost of the energy storage system, and the constraints include energy storage capacity constraints and charging and discharging rate constraints.

[0011] Preferably, dynamically identifying the complementary relationship between power generation and consumption among users specifically includes, through cluster analysis, identifying the power generation and consumption patterns of different users, thereby determining which users' power generation and power consumption demands are complementary, identifying user groups with complementary characteristics, and achieving a dynamic balance between power generation and consumption among users.

[0012] As a preferred method, cluster analysis is performed on the user's power generation and power consumption time series data to establish power generation sequence and power consumption sequence respectively. Suppose the power generation of user i at time t is The electricity demand is , define the complementary characteristic index of user power generation and consumption Used to measure the balance between power generation and consumption among different users.

[0013] As a preferred method, complementary characteristics of user groups are identified. For each user, according to the complementary index Set a threshold ,like , it is considered that users i and j have complementary characteristics of power generation and consumption. Users in the period of surplus power generation supply power to users with power shortage during the peak period of power demand, thus optimizing the balance of power resources among users.

[0014] Preferably, resources among users are dynamically aggregated according to their power generation forecasts, the charging and discharging status of the energy storage system, and load demand, and the aggregation relationship among users is adjusted in real time to achieve load scheduling and optimal allocation of power resources. The multi-objective optimization functions implemented include minimizing peak load, maximizing user comfort, and maximizing energy storage benefits.

[0015] As a preferred option, it also includes real-time monitoring, dynamic monitoring of users' electricity load, photovoltaic power generation output and energy storage status, and adjusting the optimization parameters of power generation and consumption complementarity analysis and user dynamic aggregation based on the monitoring data to adapt to environmental changes and user demand fluctuations, thereby achieving adaptive optimization of the system, and using the monitoring results as feedback information input to recalculate the optimization plan in the next time period.

[0016] The beneficial effects of the present invention are: using AI prediction models, combined with weather and historical data, to achieve short-term prediction of photovoltaic power generation to provide support for scheduling; using reinforcement learning algorithms to control air-conditioning loads to ensure load reduction during peak power periods while maintaining user comfort; using genetic algorithms to optimize the charging and discharging strategies of energy storage systems to achieve the goals of peak load shaving and valley filling and cost minimization, and finally identifying the complementary relationship between power generation and consumption among users through cluster analysis, achieving dynamic scheduling and optimal configuration of resources, thereby improving the overall efficiency and economy of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of the present invention.

[0018] Figure 2 It is a flow chart of embodiment 2 of the present invention.

[0019] Figure 3 It is a dynamic aggregation line chart of photovoltaic power generation, air conditioning load and energy storage system of the present invention. DETAILED DESCRIPTION

[0020] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the technical solutions of the present application are further described in detail below through embodiments and in combination with the accompanying drawings. It should be understood that the specific implementation method described here is only an optimal embodiment of the present application, which is only used to explain the present application and does not limit the protection scope of the present application. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0021] The present invention proposes a method for dynamic aggregation of virtual power plants based on the complementary characteristics of power generation and consumption of users. Through AI prediction models, combined with weather and historical data, short-term prediction of photovoltaic power generation is achieved to provide support for scheduling; reinforcement learning algorithms are used to control air-conditioning loads to ensure that loads are reduced during peak power periods while maintaining user comfort; genetic algorithms are used to optimize the charging and discharging strategies of energy storage systems to achieve the goals of peak shaving and valley filling and cost minimization. In addition, cluster analysis is used to identify the complementary relationship between power generation and consumption between users, to achieve dynamic scheduling and optimal configuration of resources, thereby improving the overall efficiency and economy of the system.

[0022] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow charts describe the operations (or steps) as sequential processes, many of the operations (or steps) therein can be implemented in parallel, concurrently, or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but can also have additional steps not included in the drawings; the process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0023] The technical solution of the present invention is further specifically described below through embodiments and in conjunction with the accompanying drawings.

[0024] Embodiment: This embodiment provides a method for dynamically aggregating virtual power plants based on the complementary characteristics of power generation and consumption by users, such as Figure 1 As shown, the following steps are included: S1. Make short-term predictions of photovoltaic power generation based on weather and historical power generation data. Short-term predictions include obtaining historical power generation data, weather forecast data, and time characteristics of the photovoltaic system, performing preprocessing steps including standardization and normalization on the input data, and constructing effective time series features; using machine learning or deep learning models suitable for time series prediction to process nonlinear data and capture time dependencies. Use historical data to train the model and optimize parameters to improve the accuracy of the prediction. In actual operation, use the latest weather data and time characteristics to input the model, and output the predicted value of photovoltaic power generation in the future in real time. .

[0025] S2. Dynamically control the operation of air conditioners to reduce load during peak hours. Dynamic control of air conditioner operation specifically includes designing a reward function R to balance user comfort and energy saving goals, using a reinforcement learning model to model the air conditioner demand response problem as a Markov decision process, defined as (S, A, P, R, γ), where S is the state space, including temperature, humidity and load information; A is the action space, including the air conditioner adjustment operation; P is the state transition probability, which is approximately learned through experience replay and training process in reinforcement learning; R is the reward function; γ is the discount factor.

[0026] S3. Dynamically adjust the charging and discharging strategy of the energy storage system to achieve peak shaving and valley filling and minimize costs. Use genetic algorithms to design the charging and discharging strategy of energy storage, balance supply and demand fluctuations, and reduce electricity costs. The objective function is to minimize the cost of the energy storage system. The constraints include energy storage capacity constraints and charging and discharging rate constraints.

[0027] S4. Dynamically identify the complementary relationship between power generation and consumption among users, perform cluster analysis on the time series data of power generation and consumption of users, establish power generation sequence and power consumption sequence respectively, and for each user, By setting the threshold Compare and judge the complementary characteristics of power generation and consumption between users i and j. Dynamic identification of the complementary relationship between power generation and consumption between users specifically includes, through cluster analysis, identifying the power generation and consumption patterns of different users, so as to determine which users have complementary power generation and power consumption demands, identify user groups with complementary characteristics, and achieve dynamic balance of power generation and power consumption between users. Perform cluster analysis on the user's power generation and power consumption time series data, establish power generation sequence and power consumption sequence respectively, and assume that the power generation of user i at time t is The electricity demand is , define the complementary characteristic index of user power generation and consumption It is used to measure the balance between power generation and consumption among different users. Identification of user groups with complementary characteristics. For each user, according to the complementary index Set a threshold ,like , it is considered that users i and j have complementary characteristics of power generation and consumption. Users in the period of surplus power generation supply power to users with power shortage during the peak period of power demand, thus optimizing the balance of power resources among users.

[0028] S5. Adjust the aggregation relationship between users in real time to achieve dynamic scheduling and optimal configuration of power generation and consumption resources. Dynamically aggregate resources between users based on the user's power generation forecast, the charging and discharging status of the energy storage system, and load demand, and adjust the aggregation relationship between users in real time to achieve load scheduling and optimal allocation of power resources. The multi-objective optimization functions implemented include minimizing peak load, maximizing user comfort, and maximizing energy storage benefits.

[0029] It also includes real-time monitoring, dynamic monitoring of users' electricity load, photovoltaic power generation output and energy storage status, and adjustment of the optimization parameters of power generation and consumption complementarity analysis and user dynamic aggregation based on the monitoring data to adapt to environmental changes and fluctuations in user demand, realize adaptive optimization of the system, use the monitoring results as feedback information input, and recalculate the optimization plan in the next time period.

[0030] Example 2 like Figure 2 As shown, for the dynamic aggregation of virtual power plants with complementary characteristics of power generation and consumption by users, a flow chart of a dynamic aggregation method of virtual power plants based on the complementary characteristics of power generation and consumption by users is provided, including: S1. Data Collection Obtain historical data of photovoltaic power generation systems, weather information, energy storage system status and user power load data; S2. Photovoltaic power generation forecast Use the LSTM model to make short-term predictions of photovoltaic power generation and provide basic data for virtual power plant scheduling; S3. Demand response control Dynamically adjust air conditioning loads through reinforcement learning algorithms to reduce unnecessary loads and peak power consumption during peak power periods; S4. Energy storage system dispatch Use genetic algorithms to optimize the charging and discharging strategies of energy storage systems to achieve peak load shifting, balance power supply and demand, and reduce costs; S5. Analysis on the complementarity of electricity generation and consumption among users Identify the power generation and consumption patterns among users through cluster analysis, find complementary user groups, and improve the efficiency of power resource utilization; S6. User dynamic aggregation Based on the NSGA-II optimization algorithm, the resource aggregation relationship between users is adjusted in real time according to the power generation and consumption situation and energy storage status to optimize load distribution; S7, real-time monitoring and adaptive adjustment Continuously monitor photovoltaic power generation, user load and energy storage status, dynamically adjust parameters to adapt to changes in the environment and demand, and achieve adaptive optimization of the system.

[0031] The following are the details of each step: S1: A photovoltaic power generation prediction method based on a long short-term memory network (LSTM) generates short-term power generation forecasts by utilizing historical power generation data, weather, time and other information to support the scheduling decisions of the virtual power plant system.

[0032] S2. The photovoltaic power generation prediction problem can be defined as time series prediction. Assuming that the input sequence ,in Including historical power generation, weather and time characteristics, the prediction target is the power generation in the next H time steps .

[0033] The state update equation of the LSTM model is as follows:

[0034]

[0035]

[0036]

[0037]

[0038] in, , , are the activation values ​​of the input gate, forget gate, and output gate; , , , and , , , is a trainable parameter. Based on the hidden state , the future power generation can be predicted as follows:

[0039] S3, intelligent demand response method for air conditioning load based on reinforcement learning, dynamically controls air conditioning load, reduces load during peak power demand, ensures user comfort and reduces peak power consumption. The following technical solutions are designed for this:

[0040] S3, intelligent demand response method for air conditioning load based on reinforcement learning, dynamically controls air conditioning load, reduces load during peak power demand, ensures user comfort and reduces peak power consumption. The following technical solutions are designed for this: Status definition: System status includes current indoor temperature, outdoor temperature, air conditioning power and real-time electricity price; Action definition: The system can control the power and running time of the air conditioner, including turning the air conditioner on and off, adjusting the set temperature, adjusting the fan speed, etc. Reward function design: Design a reward function to balance user comfort and energy saving goals. The reward function is defined as follows:

[0041] Where: is a penalty factor used to balance user comfort and energy saving requirements; To set the temperature; is the current actual temperature, For users' electricity consumption.

[0042] Using the reinforcement learning model, the air conditioning demand response problem is modeled as a Markov decision process (MDP), defined as (S, A, P, R, γ), and the value function is defined as follows:

[0043] Where: S is the state space, including temperature, humidity, and load information; A is the action space, including the adjustment operation of the air conditioner; P is the state transition probability, which is approximately learned through experience replay and training process in reinforcement learning, so it is implicit in the model training process and is not reflected here; R is the reward function; γ is the discount factor.

[0044] S4. Use artificial intelligence optimization algorithms (this patent design uses genetic algorithms) to design energy storage charging and discharging strategies to balance supply and demand fluctuations and reduce electricity costs. The specific design is as follows: Objective function: Minimize the cost of the energy storage system, expressed as follows:

[0045] Where: is the electricity price during discharge, is the electricity price during charging, and are the discharge and charge amounts of the energy storage system at time t, respectively.

[0046] Constraints: including energy storage capacity constraints and charge and discharge rate constraints. The corresponding expressions are as follows: Energy storage capacity constraints:

[0047] Where: E is the maximum capacity of the energy storage device.

[0048] Charge and discharge rate constraints:

[0049] Where: is the maximum charging rate of the energy storage device; is the maximum discharge rate of the energy storage device.

[0050] Genetic algorithms are used to search for the optimal solution, and the charging and discharging strategies are iteratively optimized through operations such as selection, crossover, and mutation.

[0051] S5. The user power generation and consumption complementary characteristics analysis module analyzes the power generation and consumption patterns of different users to identify user groups with complementary characteristics, so that the power generation and consumption between users can be dynamically balanced. Through the cluster analysis method, the power generation and consumption patterns of different users are identified to determine which users have complementary power generation and power consumption needs. By identifying user groups with complementary characteristics, the dynamic balance of power generation and consumption between users can be better achieved, and the overall efficiency of the system can be improved.

[0052] Cluster analysis: cluster analysis is performed on the user's power generation and power consumption time series data to establish power generation sequence and power consumption sequence respectively. The specific expression is as follows: Assume that the power generation of user i at time t is The electricity demand is . Define the complementary characteristics of power generation and consumption of users It is used to measure the balance between power generation and consumption among different users. The specific formula is as follows

[0053] in, The smaller it is, the more complementary the power generation of user i and the power consumption of user j are in time.

[0054] Identification of complementary user groups: For each user, Set a threshold ,like , it is considered that users i and j have complementary characteristics of power generation and consumption. Aggregating complementary user groups allows users in surplus power generation periods to supply power to users with power shortages during peak power demand periods, thereby optimizing the balance of power resources among users.

[0055] Dynamic balance realization: By matching the power generation and power consumption between complementary user groups, the supply and demand balance between users is dynamically adjusted in time sequence. The optimization goal is to achieve the optimal state of the overall power generation and power consumption balance of the system, thereby maximizing the efficiency of the system.

[0056] S6, the user dynamic aggregation module is based on a multi-objective optimization algorithm. It aggregates the power generation and consumption resources between users in real time, dynamically adjusts the aggregation relationship between users, and realizes load scheduling and optimal allocation of power resources. The user dynamic aggregation module aims to dynamically aggregate resources between users according to the user's power generation forecast, the charging and discharging status of the energy storage system, and the load demand through the NSGA-II optimization algorithm, and adjusts the aggregation relationship between users in real time to achieve load scheduling and optimal allocation of power resources. The multi-objective optimization functions of this module include minimizing peak load, maximizing user comfort, and maximizing energy storage benefits. The specific objective functions are as follows: Minimize the peak load f1:

[0057] Where: is the electricity demand of user i at time t; is the power generation of user j at time t; and are the discharge and charge amounts of the energy storage system at time t respectively.

[0058] Maximize user comfort f2: User comfort is related to air conditioning load regulation. Let the reward function R i,t represents the comfort benefit of user i at time t, which is defined as follows:

[0059]

[0060] Maximize energy storage benefits f3:

[0061] Where: is the electricity price during discharge, is the electricity price during charging, and are the discharge and charge amounts of the energy storage system at time t, is the power generation of the photovoltaic system at time t.

[0062] The NSGA-II algorithm is used to solve the Pareto optimal solution, generate a real-time user aggregation strategy, and achieve dynamic load scheduling and resource optimization.

[0063] S7. Dynamically monitor the user's power load, photovoltaic power generation output and energy storage status, and adjust the optimization parameters of the power generation and consumption complementarity analysis module and the user dynamic aggregation module based on the monitoring data to adapt to environmental changes and user demand fluctuations, and realize adaptive optimization of the system.

[0064] Real-time monitoring unit: This unit continuously collects the user's power load curve, the actual output of photovoltaic power generation and the current status of the energy storage system, and pre-processes the data, including filtering, normalization and data smoothing.

[0065] Parameter adjustment mechanism: According to the changes in real-time monitoring data, the clustering algorithm parameters in the power generation and consumption complementarity analysis module and the optimization parameters in the user dynamic aggregation module are dynamically adjusted to ensure that the optimization strategy is automatically updated as the environment and user needs change.

[0066] Feedback optimization mechanism: This unit inputs the monitoring results into the system as feedback information so that the optimization plan can be recalculated in the next time period based on factors such as changes in user demand, weather conditions and electricity market prices, thereby realizing the adaptive optimization capability of the virtual power plant system.

[0067] Example 3 In the actual application case designed, a virtual power plant model with two users is considered. Each user has a certain photovoltaic power generation system and air conditioning load, and the system is also equipped with energy storage equipment. As an intermittent energy source, photovoltaic power generation is affected by factors such as weather and season, so short-term power generation forecasts are required. The demand for air conditioning load is affected by factors such as temperature and user habits. Especially during peak power consumption periods, adjusting the air conditioning load can significantly reduce the peak load. The actual effect is as follows: Figure 3 shown.

[0068] The core principles of the technical solution: 1. Photovoltaic power generation forecast By collecting historical power generation data, meteorological data and time characteristics (such as season, time period, etc.) of the photovoltaic system, the LSTM model predicts the power generation in the next few hours and provides accurate data support for the virtual power plant dispatching system.

[0069] 2. Demand response and air conditioning load management The system adjusts the air conditioner's operating mode by real-time monitoring of the air conditioner's operating status (such as indoor and outdoor temperature, air conditioner power, real-time electricity price, etc.). The state, action, and reward function are set through the Markov decision process (MDP) model, and the reinforcement learning algorithm is continuously trained to enable the system to make the optimal load adjustment decision at every moment.

[0070] 3. Energy storage system optimization The genetic algorithm searches for the optimal solution in multiple iterations by considering factors such as electricity price fluctuations, energy storage system capacity limitations, and charge and discharge rate limitations. The system adjusts the energy storage charge and discharge plan in real time based on the current energy storage status, power demand, and power generation to improve energy utilization efficiency and economy.

[0071] 4. Analysis and dynamic aggregation of power generation and consumption complementarity among users The clustering algorithm analyzes the power generation and load demand of different users to identify user combinations with complementary power demands. The system dynamically adjusts the power distribution relationship between users according to the time period changes to achieve more efficient resource utilization. For example, when the photovoltaic power generation is low, the system can balance supply and demand by storing electricity in advance or dispatching complementary users to avoid energy waste.

[0072] 5. Dynamic aggregation and multi-objective optimization The system collects users’ photovoltaic power generation, power load and energy storage status information in real time, and calculates the optimal dispatch strategy for each user through the NSGA-II algorithm. The system dynamically adjusts the power resource aggregation relationship between users during peak load to ensure the optimization of the overall system’s energy efficiency.

[0073] The following is a sample result graph: Environmental preparation: 1. Python version: Python 3.9.

[0074] 2. Install necessary libraries: TensorFlow: For deep learning models (such as LSTM and Dense layers).

[0075] NumPy: For numerical computing and array manipulation.

[0076] scikit-learn: for data preprocessing Table 1 Dynamic aggregation data of photovoltaic power generation, air conditioning load and energy storage system

[0077] The specific embodiments described herein are merely examples of the spirit of the present invention. The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of protection of the present application. It should be noted that a technician in the technical field to which the present application belongs may make various modifications or supplements to the described specific embodiments or replace them in a similar manner, but will not deviate from the spirit of the present application or exceed the scope defined in the attached claims. For a person of ordinary skill in the art, multiple deformations and improvements may also be made without departing from the concept of the present application. Therefore, the scope of protection of the present application shall be subject to the attached claims.

Claims

1. A method for dynamic aggregation of virtual power plants based on the complementary characteristics of power generation and consumption by users, characterized in that: The following steps are involved: S1. Make short-term forecasts of photovoltaic power generation based on weather and historical power generation data; S2. Dynamically control air conditioning operation to reduce load during peak hours; S3. Dynamically adjust the charging and discharging strategy of the energy storage system to achieve peak load shifting and cost minimization; S4. Dynamically identify the complementary relationship between power generation and consumption among users, perform cluster analysis on the time series data of power generation and consumption of users, establish power generation sequence and power consumption sequence respectively, and for each user, By setting the threshold Compare and judge the complementary characteristics of power generation and consumption between users i and j; S5. Adjust the aggregation relationship between users in real time to achieve dynamic scheduling and optimal configuration of power generation and consumption resources.

2. According to claim 1, a method for dynamic aggregation of virtual power plants based on the complementary characteristics of power generation and consumption by users, characterized in that: Short-term forecasting includes obtaining historical power generation data, weather forecast data and time characteristics of the photovoltaic system, performing preprocessing steps including standardization and normalization on the input data, and constructing effective time series features; Use machine learning or deep learning models suitable for time series forecasting to handle non-linear data and capture time dependencies.

3. According to claim 2, a method for dynamic aggregation of virtual power plants based on the complementary characteristics of power generation and consumption by users, characterized in that: Use historical data to train the model and optimize parameters to improve the accuracy of the prediction. In actual operation, use the latest weather data and time characteristics to input the model and output the predicted value of photovoltaic power generation in the future in real time. .

4. According to claim 1, a method for dynamic aggregation of virtual power plants based on the complementary characteristics of power generation and consumption by users, characterized in that: Dynamic control of air-conditioning operation specifically includes designing a reward function R to balance user comfort and energy-saving goals. Using a reinforcement learning model, the air-conditioning demand response problem is modeled as a Markov decision process, defined as (S, A, P, R, γ), where S is the state space, including temperature, humidity, and load information; A is the action space, including the air-conditioning adjustment operation; P is the state transition probability, which is approximately learned through experience replay and training process in reinforcement learning; R is the reward function; and γ is the discount factor.

5. According to claim 1, a method for dynamic aggregation of virtual power plants based on the complementary characteristics of power generation and consumption by users, characterized in that: Genetic algorithms are used to design charging and discharging strategies for energy storage to balance supply and demand fluctuations and reduce electricity costs. The objective function is to minimize the cost of the energy storage system. Constraints include energy storage capacity constraints and charging and discharging rate constraints.

6. According to claim 1, a method for dynamic aggregation of virtual power plants based on the complementary characteristics of power generation and consumption by users, characterized in that: Dynamically identifying the complementary relationship between power generation and consumption among users specifically includes, through cluster analysis, identifying the power generation and consumption patterns of different users, thereby determining which users have complementary power generation and consumption demands, identifying user groups with complementary characteristics, and achieving a dynamic balance between power generation and consumption among users.

7. A method for dynamic aggregation of virtual power plants based on the complementary characteristics of power generation and consumption by users according to claim 6, characterized in that: Cluster analysis is performed on the user's power generation and power consumption time series data to establish power generation sequence and power consumption sequence respectively. Suppose the power generation of user i at time t is The electricity demand is , define the complementary characteristics of user power generation and consumption Used to measure the balance between power generation and consumption among different users.

8. A method for dynamic aggregation of virtual power plants based on the complementary characteristics of power generation and consumption by users according to claim 6 or 7, characterized in that: Identify user groups with complementary characteristics. For each user, we use the complementary index Set a threshold ,like , it is considered that users i and j have complementary characteristics of power generation and consumption. Users in the period of surplus power generation supply power to users with power shortage during the peak period of power demand, thus optimizing the balance of power resources among users.

9. A method for dynamic aggregation of virtual power plants based on the complementary characteristics of power generation and consumption by users according to claim 1, 6 or 7, characterized in that: Dynamically aggregate resources among users based on their power generation forecasts, the charging and discharging status of the energy storage system, and load demand, and adjust the aggregation relationship between users in real time to achieve load scheduling and optimal allocation of power resources. The multi-objective optimization functions implemented include minimizing peak load, maximizing user comfort, and maximizing energy storage benefits.

10. A method for dynamic aggregation of virtual power plants based on the complementary characteristics of power generation and consumption by users according to claim 1, 6 or 7, characterized in that: It also includes real-time monitoring, dynamic monitoring of users' electricity load, photovoltaic power generation output and energy storage status, and adjustment of the optimization parameters of power generation and consumption complementarity analysis and user dynamic aggregation based on the monitoring data to adapt to environmental changes and fluctuations in user demand, realize adaptive optimization of the system, use the monitoring results as feedback information input, and recalculate the optimization plan in the next time period.

Images