Cooperative scheduling method for multiple microgrids

Abstract

The invention relates to a multi-microgrid-oriented collaborative scheduling method, which belongs to the field of energy scheduling and smart grids, and comprises the following steps: constructing a community microgrid price-driven scheduling model, establishing a power and energy constraint model based on the operation characteristics of controllable resources in a community microgrid, and constructing a community-side electricity price-load interaction mechanism; modeling a day-ahead electricity price formulating problem into a dynamic optimization task in a continuous action space, and training an electricity price generator by adopting a TD3 algorithm to obtain an excitation electricity price curve meeting economical efficiency and a peak-valley adjustment target; receiving a power curve obtained by optimization of a family side in a real-time operation stage, calculating a power demand in combination with a real-time load, photovoltaic output and an energy storage state, constructing a real-time optimization model of energy storage charging and discharging and photovoltaic compensation, and correcting and smoothing the power curve to obtain final real-time scheduling power; the corrected community side actual scheduling result serves as feedback data to be input into the electricity price generation model, and the reinforcement learning strategy is updated to train the reference electricity price of the next period.

Description

Technical Field

[0001] This invention belongs to the field of energy dispatch and smart grid technology, and relates to a collaborative dispatch method for multiple microgrids. Background Technology

[0002] With the continuous advancement of new power system construction, energy supply methods are evolving from centralized to deeply integrated distributed systems. The high proportion of renewable energy integration has led to significant randomness and volatility in the power system, posing greater challenges to the grid in terms of power balance, flexible regulation, and security and stability. The rapid popularization of distributed photovoltaic, energy storage, and electric vehicles at the community level has made community microgrids increasingly key units for supporting diverse energy devices and achieving distributed consumption and demand-side regulation. Unlike traditional loads that are merely passive recipients, current end-side loads possess greater controllability and interactivity, enabling them to participate in system operation optimization through price signals, demand response mechanisms, and intelligent dispatch strategies. This provides new possibilities for building a flexible and efficient dispatch system.

[0003] With the coexistence of multiple resource types, the scheduling complexity of community microgrids has significantly increased. On the one hand, photovoltaic output exhibits significant volatility and prediction errors, requiring real-time balancing through flexible resources such as energy storage and electric vehicles. On the other hand, user-side flexible loads and electric vehicle charging behavior are highly sensitive to prices, and different users exhibit varying and uncertain response patterns, resulting in nonlinear characteristics in the overall load. Furthermore, the behavioral coupling between the community and household levels is complex, requiring scheduling strategies to simultaneously consider user economics, system peak-valley regulation capabilities, and security constraints, making effective coordination difficult to achieve with traditional centralized optimization methods.

[0004] Existing research largely focuses on independent dispatching at the household or community level, lacking in-depth exploration of cross-level price guidance mechanisms and load response feedback. At the community level, most studies employ prediction-based static optimization methods to generate electricity prices, making it difficult to adjust strategies based on actual response behavior. At the user level, demand response behavior has not formed a stable quantitative mechanism, resulting in significant discrepancies between incentive signals and actual load changes. Furthermore, the coupling relationship between the day-ahead and real-time operation phases is insufficiently addressed, and a sustainable iterative strategy update mechanism is lacking, leading to unstable electricity price regulation effects, high system operating costs, and difficulty in further improving the utilization level of renewable energy.

[0005] With the development of reinforcement learning, advanced optimization algorithms, and high-precision sensing devices, it has become possible to construct electricity pricing strategies that can learn the real-world response patterns of communities. Reinforcement learning has advantages such as modeling uncertain systems, learning strategies through interactions, and handling continuous action spaces, making it suitable for resolving complex causal relationships between electricity pricing and load response. Meanwhile, real-time optimization can utilize energy storage and controllable loads to quickly correct power deviations, achieving smoother and more stable operation on the community side. However, a feasible and systematic method remains to integrate reinforcement learning, pricing mechanisms, and real-time optimization into a closed-loop system, achieving full-process coordination from day-ahead pricing to real-time scheduling and policy feedback. Summary of the Invention

[0006] In view of this, the purpose of this invention is to provide a cooperative scheduling method for multiple microgrids.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A collaborative scheduling method for multiple microgrids includes the following steps: S1: Construct a price-driven dispatch model for community microgrids, establish a power and energy constraint model based on the operating characteristics of controllable resources within the community microgrid, and construct a community-side electricity price-load interaction mechanism to characterize the impact of electricity price changes on the total load of the community. S2: The day-ahead electricity pricing problem is modeled as a dynamic optimization task in a continuous action space. The electricity price generator is trained using the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm to obtain the incentive price curve that satisfies the objectives of economic efficiency and peak-valley regulation. S3: During the real-time operation phase, it receives the power curve optimized by the household side, calculates the power demand by combining the real-time load, photovoltaic output and energy storage status, and builds a real-time optimization model for energy storage charging and discharging and photovoltaic compensation to correct and smooth the power curve to obtain the final real-time dispatch power. S4: After completing the power correction for the entire real-time operation cycle, the corrected actual scheduling results on the community side are used as feedback data to input into the electricity price generation model, which is used to update the reinforcement learning strategy and train the benchmark electricity price for the next cycle.

[0008] Furthermore, the controllable resources within the community microgrid mentioned in step S1 include photovoltaic systems (PV), battery energy storage systems (BSS), and electric vehicles (EV), wherein: The Baseline Controller (BSS) is used to smooth out fluctuations in purchased power and regulate load, modeling and describing the physical constraints and long-term performance degradation during charging and discharging. The BSS must meet charging and discharging power constraints within each scheduling period. (1) Considering the battery life degradation effect, the aging cost model based on the full life cycle cost amortization is as follows: (2) In the formula, For total energy storage capacity, The aging cost per unit of electricity is composed of the total investment cost. Total number of charging cycles With total energy storage capacity constitute; The real-time changes in the state of charge are described by the battery dynamics update equation, and capacity constraints are added to ensure that the energy storage state remains within a safe range. (3) As a renewable energy supply unit, PV is modeled using a data-driven approach; During idle periods, EVs participate in scheduling as distributed energy storage units. Their charging and discharging process is similar to that of BSS, satisfying power constraints and aging cost models. Based on a travel dataset generated using a Monte Carlo algorithm, EVs during travel periods... Domestic demand exit scheduling: (4) At the moment of departure The EV indicator light meets users' minimum travel needs: (5).

[0009] Furthermore, the construction of the community-side electricity price-load interaction mechanism described in step S1, used to characterize the impact of electricity price changes on the total community load, specifically includes: The TD3 electricity price generator outputs a continuous electricity price curve based on the load characteristics of the previous cycle and the peak-valley adjustment target. It encourages load shifting with low electricity prices and inhibits load growth with high electricity prices, thereby forming energy consumption guidance. After receiving the electricity price, households implement price response strategies based on their own equipment constraints and energy preferences. When the electricity price is high, they reduce or postpone the operation of flexible loads, and when the electricity price is low, they increase the amount of charging or movable loads. The community aggregates the response behavior of each household to form the load adjustment pattern caused by electricity price changes. This pattern describes the elasticity of the community's total load under price-driven conditions and provides feedback for subsequent electricity price optimization.

[0010] Furthermore, step S2 specifically includes: At the community microgrid level, based on day-ahead load forecast data and the benchmark electricity price system, a dynamic time-of-use pricing incentive signal is designed to coordinate the distributed demand response of lower-level household microgrids. By smoothing the aggregated load curve through a decentralized response strategy, secondary peak load phenomena caused by users' synchronous demand adjustments are avoided. The incentive signal optimization problem is modeled as follows: : (6) in, For the load power after demand response of all home microgrids, To optimize the maximum load power within a cycle, Indicates the excitation signal; and This is used to ensure that the excitation signal is within a reasonable range and to avoid the problem of the excitation signal increasing indefinitely.

[0011] Furthermore, in step S2, the process is modeled as a sequential decision-making problem involving interactive experience learning, and the TD3 algorithm is used to generate and update the community-level incentive signals; specifically, it is modeled as an MDP containing the following elements: State space for the future Forecast load data for each time slot Meteorological data and base electricity price data , denoted as: (7) The action space refers to the excitation signals of the community microgrid in each time slot within the optimization cycle. : (8) The reward function is designed by comprehensively considering the reduction of the peak-to-valley difference in the load curve after response and the cost difference before and after incentive, and is formally expressed as: (9).

[0012] Furthermore, in step S3, by formulating the optimal charging and discharging strategy for the ESS unit and integrating photovoltaic power generation with idle EV energy storage capacity during non-driving periods, the power purchase curve is smoothed and the power supply pressure on the main grid is alleviated; optimization problem The expression is as follows: (10) in, To constrain the economic viability of energy storage systems, this is used to ensure that the aging cost of energy storage units does not exceed the revenue generated by electricity arbitrage. Physical constraints for energy storage and photovoltaic systems; By combining the adjustable capabilities of centralized energy storage and electric vehicles, constraints are established for charging and discharging power, SOC evolution, and linearization or piecewise linearization of power balance, ensuring that the original constraints remain convex. By introducing auxiliary variables and quadratic smoothing terms, the objective function is transformed into a convex optimization problem that can be directly solved by a quadratic programming or second-order cone programming solver. The energy storage and electric vehicle adjustment power sequence obtained after solving is used to correct the original household-side power curve in real time, thereby forming the final community dispatch power.

[0013] Furthermore, in step S4, the state-action-reward-next state quadruple is obtained by sampling from the experience pool. First, a target action with policy smoothing is generated based on community load, deviation data, and the electricity price signal of the previous cycle. The target value is formed using a dual Q-value network structure, which simultaneously includes peak-to-valley reduction and price penalty. The evaluation network and policy network ensure training stability through a soft update mechanism.

[0014] The beneficial effects of this invention are as follows: This invention proposes a collaborative scheduling method for multiple microgrids, constructing a two-level model of community and household and combining controllable resources such as flexible loads, electric vehicles, and energy storage to achieve collaborative optimization of multi-level loads. By introducing the "electricity price-load" interaction mechanism on the community side, it characterizes the impact of electricity price changes on user response, enabling electricity prices to have a predictable and controllable guiding effect on the demand side. Furthermore, this invention models electricity price generation as a Markov decision process in a continuous action space and uses an improved TD3 reinforcement learning to achieve adaptive updates of the electricity price strategy, forming a closed-loop mechanism of "electricity price incentive—load response—strategy iteration". In real-time scheduling, by rolling optimization and coordination of energy storage, electric vehicles, and photovoltaic resources, community power smoothing and cost reduction are achieved, and the scheduling results are used to update the electricity price strategy in reverse, ensuring that the electricity price continuously matches the actual load elasticity. This method can maintain good adaptability in energy scenarios with multiple stakeholders, strong coupling, and uncertain responses, significantly improving demand response performance and system economy, and has broad engineering application value. Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0015] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 This is a diagram of the community microgrid system architecture of the present invention; Figure 2This is a flowchart of the price-driven two-stage optimization process of the present invention. Detailed Implementation

[0016] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0017] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0018] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0019] Example 1: Please see Figures 1-2 The collaborative scheduling method for multiple microgrids described in this invention specifically includes the following steps: Step 1: Construct a price-driven dispatch model for the community microgrid. Based on the operating characteristics of controllable resources such as photovoltaic systems (PV), battery energy storage systems (BSS), and electric vehicles (EV) within the community microgrid, establish corresponding power and energy constraint models, and construct a community-side electricity price-load interaction mechanism to characterize the impact of electricity price changes on the total load of the community. Step 1.1: The BSS is mainly used to smooth out fluctuations in purchased power and regulate load. Its modeling focuses on describing the physical constraints and long-term performance degradation during the charging and discharging process. The BSS must meet the charging and discharging power constraints in each scheduling period: (1) Considering the battery life degradation effect, the aging cost model based on the full life cycle cost amortization is as follows: (2) In the formula, For total energy storage capacity, The aging cost per unit of electricity is composed of the total investment cost. Total number of charging cycles With total energy storage capacity constitute.

[0020] To prevent overcharging and over-discharging of energy storage, the model describes the real-time changes in SOC through battery dynamics update equations and incorporates capacity constraints to ensure that the energy storage state remains within a safe range. (3) As a renewable energy supply unit, PV is modeled using a data-driven approach to ensure that the fluctuations in photovoltaic power generation can be accurately reflected over different time periods.

[0021] During idle periods, EVs participate in scheduling as distributed energy storage units, and their charging and discharging process is similar to that of BSS, satisfying power constraints and aging cost models. However, the scheduling capability of EVs is strictly limited by user travel behavior. Using a travel dataset generated based on the Monte Carlo algorithm, EVs during travel periods... Domestic demand exit scheduling: (4) At the same time, at the departure time The EV's on status must meet the user's minimum travel needs: (5) Step 1.2: The Twin Delayed Deep Deterministic Policy Gradient (TD3) electricity price generator outputs a continuous electricity price curve based on the load characteristics of the previous cycle and the peak-valley adjustment target. Low prices encourage load shifting, while high prices suppress load growth, thus forming a clear energy consumption guideline. Upon receiving the electricity price, households implement price response strategies based on their own equipment constraints and energy preferences. During high prices, they reduce or postpone flexible load operation; during low prices, they increase the amount of charging or movable loads. The community aggregates the response behaviors of each household to form a load adjustment pattern caused by electricity price changes. This pattern describes the elasticity of the community's total load under price-driven conditions and provides feedback for subsequent electricity price optimization.

[0022] Step 2: Model the day-ahead electricity pricing problem as a dynamic optimization task in a continuous action space, and use the TD3 reinforcement learning algorithm to train the electricity price generator to obtain the incentive price curve that satisfies the economic and peak-valley regulation objectives; At the community microgrid level, based on day-ahead load forecasting data and the benchmark electricity price system, a dynamic time-of-use pricing incentive signal is designed to coordinate the distributed demand response of lower-level household microgrids. This mechanism aims to smooth out aggregated load curves through a decentralized response strategy, thereby avoiding secondary peak load phenomena caused by users' synchronous demand adjustments. The optimization problem of this incentive signal is modeled as follows: : (6) in, For the load power after demand response of all home microgrids, To optimize the maximum load power within a cycle, This indicates an excitation signal. and This is used to ensure that the excitation signal is within a reasonable range, thereby avoiding the problem of the excitation signal increasing indefinitely.

[0023] The process is modeled as a sequential decision-making problem involving interactive experience learning, and the TD3 algorithm is used to generate and update community-level incentive signals. Specifically, it is modeled as an MDP containing the following elements: State space for the future Forecast load data for each time slot Meteorological data and base electricity price data , denoted as: (7) The action space refers to the excitation signals of the community microgrid in each time slot within the optimization cycle. : (8) The reward function design needs to consider two objectives: firstly, to reduce the peak-to-trough difference in the load curve after response; and secondly, to mitigate the cost difference before and after excitation, thereby improving the acceptability and stability of the signal. Formalized as: (9) By using the TD3 algorithm to stably learn strategies in the continuous action space, the load curve after demand response reduces the difference between load peaks and valleys, and avoids the phenomenon of secondary peaks caused by concentrated response.

[0024] Step 3: During the real-time operation phase, receive the power curve obtained from the home side optimization, calculate the power demand by combining the real-time load, photovoltaic output and energy storage status, and build a real-time optimization model for energy storage charging and discharging and photovoltaic compensation. Correct and smooth the power curve to obtain the final real-time dispatch power. By developing optimal charging and discharging strategies for ESS units and effectively integrating photovoltaic power generation with idle EV energy storage capacity during non-driving periods, this phase aims to smooth the power purchase curve and alleviate the supply pressure on the main grid. This optimization problem... This can be expressed as: (10) in, To constrain the economic viability of energy storage systems, this is used to ensure that the aging cost of energy storage units does not exceed the revenue generated by electricity arbitrage. Physical constraints for energy storage and photovoltaic systems.

[0025] By combining the adjustable capabilities of centralized energy storage and electric vehicles, linearized or piecewise linearized constraints are established for charging and discharging power, SOC evolution, and power balance, ensuring the original constraints remain convex. By introducing auxiliary variables and quadratic smoothing terms, the objective function is transformed into a convex optimization problem that can be directly solved by quadratic programming or second-order cone programming solvers. The resulting energy storage and electric vehicle regulation power sequences are used to correct the original household-side power curve in real time, thus forming the final community-dispatch power.

[0026] Step 4: After completing the power correction for the entire real-time operation cycle, the corrected actual scheduling results on the community side are used as feedback data to input into the electricity price generation model to update the reinforcement learning strategy and train the benchmark electricity price for the next cycle.

[0027] The state-action-reward-next state quadruple is obtained by sampling from the experience pool. First, based on community load, deviation data, and the previous cycle's electricity price signal, a target action with strategy smoothing is generated: (11) And the target value is formed using a dual-Q value network structure: (12) The rewards include both "peak-valley difference reduction" and "price penalty," ensuring that the electricity price generator can proactively guide household responses and avoid secondary peaks. The evaluation network and policy network utilize a soft update mechanism to ensure training stability.

[0028] Example 2: An electronic device, comprising a memory and a processor; The memory is used to store computer programs; The processor is configured to implement the method described in Embodiment 1 when executing the computer program.

[0029] Example 3: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in Embodiment 1.

[0030] Example 4: A computer program product includes a computer program that, when executed by a processor, implements the method described in Example 1.

[0031] In the above embodiments, the reference to "this embodiment" in the specification indicates that a specific feature, structure, or characteristic described in connection with the embodiment is included in at least some embodiments, but not necessarily all embodiments. Multiple appearances of "this embodiment" do not necessarily all refer to the same embodiment.

[0032] In the above embodiments, although the invention has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory structures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed. The embodiments of the invention are intended to cover all such substitutions, modifications, and variations falling within the broad scope of the appended claims.

[0033] As will be understood by those skilled in the art, the computer-readable storage medium described in this embodiment allows for the implementation of all or part of the steps in the above method embodiments by computer program-related hardware. The aforementioned computer program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0034] The electronic terminal provided in this embodiment includes a processor, a memory, a transceiver, and a communication interface. The memory and the communication interface are connected to the processor and the transceiver and complete communication between them. The memory is used to store computer programs, the communication interface is used to perform communication, and the processor and the transceiver are used to run the computer programs, so that the electronic terminal performs the steps of the above method.

[0035] In this embodiment, the memory may include random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device.

[0036] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0037] This invention can be used in a wide range of general-purpose or special-purpose computing system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices, etc.

[0038] This invention can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This invention can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0039] 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 it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A cooperative scheduling method for multiple microgrids, characterized in that: Includes the following steps: S1: Construct a price-driven dispatch model for community microgrids, establish a power and energy constraint model based on the operating characteristics of controllable resources within the community microgrid, and construct a community-side electricity price-load interaction mechanism; S2: The day-ahead electricity pricing problem is modeled as a dynamic optimization task in a continuous action space. The electricity price generator is trained using the dual-delay deep deterministic policy gradient algorithm TD3 to obtain an incentive price curve that satisfies the objectives of economic efficiency and peak-valley regulation. S3: During the real-time operation phase, it receives the power curve optimized by the household side, calculates the power demand by combining the real-time load, photovoltaic output and energy storage status, and builds a real-time optimization model for energy storage charging and discharging and photovoltaic compensation to correct and smooth the power curve to obtain the final real-time dispatch power. S4: After completing the power correction for the entire real-time operation cycle, the corrected actual scheduling results on the community side are used as feedback data to input into the electricity price generation model, which is used to update the reinforcement learning strategy and train the benchmark electricity price for the next cycle.

2. The cooperative scheduling method for multiple microgrids according to claim 1, characterized in that: The controllable resources within the community microgrid mentioned in step S1 include photovoltaic (PV) systems, battery storage systems (BSS), and electric vehicles (EVs), wherein: The Baseline Controller (BSS) is used to smooth out fluctuations in purchased power and regulate load, modeling and describing the physical constraints and long-term performance degradation during charging and discharging. The BSS must meet charging and discharging power constraints within each scheduling period. (1) Considering the battery life degradation effect, the aging cost model based on the full life cycle cost amortization is as follows: (2) In the formula, For total energy storage capacity, The aging cost per unit of electricity is composed of the total investment cost. Total number of charging cycles With total energy storage capacity constitute; The real-time changes in the state of charge are described by the battery dynamics update equation, and capacity constraints are added to ensure that the energy storage state remains within a safe range. (3) As a renewable energy supply unit, PV is modeled using a data-driven approach; During idle periods, EVs participate in scheduling as distributed energy storage units. Their charging and discharging process is similar to that of BSS, satisfying power constraints and aging cost models. Based on a travel dataset generated using a Monte Carlo algorithm, EVs during travel periods... Domestic demand exit scheduling: (4) At the moment of departure The EV indicator light meets users' minimum travel needs: (5)。 3. The cooperative scheduling method for multiple microgrids according to claim 2, characterized in that: Step S1 describes the construction of a community-side electricity price-load interaction mechanism to characterize the impact of electricity price changes on the total community load. Specifically, this includes: The TD3 electricity price generator outputs a continuous electricity price curve based on the load characteristics of the previous cycle and the peak-valley adjustment target. It encourages load shifting with low electricity prices and inhibits load growth with high electricity prices, thereby forming energy consumption guidance. After receiving the electricity price, households implement price response strategies based on their own equipment constraints and energy preferences. When the electricity price is high, they reduce or postpone the operation of flexible loads, and when the electricity price is low, they increase the amount of charging or movable loads. The community aggregates the response behavior of each household to form the load adjustment pattern caused by electricity price changes. This pattern describes the elasticity of the community's total load under price-driven conditions and provides feedback for subsequent electricity price optimization.

4. The cooperative scheduling method for multiple microgrids according to claim 1, characterized in that: Step S2 specifically includes: At the community microgrid level, based on day-ahead load forecast data and the benchmark electricity price system, a dynamic time-of-use pricing incentive signal is designed to coordinate the distributed demand response of lower-level household microgrids. By smoothing the aggregated load curve through a decentralized response strategy, secondary peak load phenomena caused by users' synchronous demand adjustments are avoided. The incentive signal optimization problem is modeled as follows: : (6) in, For the load power after demand response of all home microgrids, To optimize the maximum load power within a cycle, Indicates the excitation signal; and This is used to ensure that the excitation signal is within a reasonable range and to avoid the problem of the excitation signal increasing indefinitely.

5. The cooperative scheduling method for multiple microgrids according to claim 4, characterized in that: In step S2, the process is modeled as a sequential decision-making problem involving interactive experience learning, and the TD3 algorithm is used to generate and update the community-level incentive signals; specifically, it is modeled as an MDP containing the following elements: State space for the future Forecast load data for each time slot Meteorological data and base electricity price data , denoted as: (7) The action space refers to the excitation signals of the community microgrid in each time slot within the optimization cycle. : (8) The reward function is designed by comprehensively considering the reduction of the peak-to-valley difference in the load curve after response and the cost difference before and after incentive, and is formally expressed as: (9)。 6. The cooperative scheduling method for multiple microgrids according to claim 1, characterized in that: In step S3, by formulating the optimal charging and discharging strategy for the ESS unit and integrating photovoltaic power generation with idle EV energy storage capacity during non-driving periods, the power purchase curve is smoothed and the power supply pressure on the main grid is alleviated; optimization problem The expression is as follows: (10) in, To constrain the economic viability of energy storage systems, this is used to ensure that the aging cost of energy storage units does not exceed the revenue generated by electricity arbitrage. Physical constraints for energy storage and photovoltaic systems; By combining the adjustable capabilities of centralized energy storage and electric vehicles, constraints are established for charging and discharging power, SOC evolution, and linearization or piecewise linearization of power balance, ensuring that the original constraints remain convex. By introducing auxiliary variables and quadratic smoothing terms, the objective function is transformed into a convex optimization problem that can be directly solved by a quadratic programming or second-order cone programming solver. The energy storage and electric vehicle adjustment power sequence obtained after solving is used to correct the original household-side power curve in real time, thereby forming the final community dispatch power.

7. The cooperative scheduling method for multiple microgrids according to claim 1, characterized in that: In step S4, the state-action-reward-next state quadruple is obtained by sampling from the experience pool. First, a target action with strategy smoothing is generated based on community load, deviation data, and the electricity price signal of the previous cycle. The target value is formed using a dual-Q value network structure, which simultaneously includes peak-to-valley reduction and price penalty; the evaluation network and policy network ensure training stability through a soft update mechanism.

Images