A wind storage scheduling method based on battery state of health charge and discharge depth limitation

Abstract

The application discloses a wind storage scheduling method based on battery health degree charge and discharge depth limitation, comprising the following steps: obtaining real-time load data of a transformer area power grid, real-time wind power data of a fan power generation system and operation state data of an energy storage battery, obtaining a transformer area power grid load prediction value, a fan prediction output value and a real-time battery health degree SOH; constructing a nonlinear mapping relationship between the real-time battery health degree SOH and a battery charge and discharge depth DOD limitation range, obtaining a dynamic DOD limitation range; using a reinforcement learning scheduling model to make decisions, generating a charge and discharge power instruction of the energy storage battery and a fan actual output instruction; according to the charge and discharge power instruction and the fan actual output instruction, controlling the energy storage system to charge and discharge within the dynamic DOD range, and adjusting the fan output, so as to realize wind storage collaborative optimization scheduling; during scheduling, the safety and stability of the transformer area power grid, the efficient consumption of new energy and the long-term service of the energy storage battery can be considered at the same time, so as to adapt to complex and changeable transformer area operation scenes.

Description

Technical Field

[0001] This invention relates to the field of energy storage dispatching technology, and in particular to a wind-storage dispatching method based on the charge-discharge depth limitation of battery health. Background Technology

[0002] With the rapid development of new energy power generation technologies, wind power, as a clean and low-carbon energy form, is increasingly accounting for a larger share of distribution network areas. However, wind power output is highly intermittent, volatile, and unpredictable, posing a significant challenge to voltage stability, frequency regulation, and power supply reliability in the distribution network area if directly connected to the grid. Energy storage systems, as key equipment for mitigating new energy fluctuations and enhancing grid flexibility, combined with wind turbines to form a wind-storage integrated system, have become a core technological approach to resolving the contradiction between new energy consumption and grid safety in distribution areas. However, current wind-storage dispatching methods primarily focus on two main objectives: grid stability and new energy consumption. These methods aim to achieve source-load power balance by optimizing energy storage charging and discharging strategies and wind turbine output regulation. However, existing technologies still have many shortcomings in practical applications, specifically in the following aspects: First, in the application of intelligent scheduling algorithms in the field of power grid dispatching, the state space of existing models is incomplete, ignoring key features such as the cumulative number of battery cycles and dynamic DOD constraints, resulting in insufficient ability of the algorithm to perceive the health status of equipment. Second, the action space is inaccurate, failing to fully consider the minimum safe operating output constraints of wind turbines, leading to equipment feasibility risks in dispatching instructions. Third, the reward function is incomplete, focusing on a single objective, such as only focusing on power grid stability or new energy consumption, making it difficult to adapt to complex and ever-changing distribution area operation scenarios. Furthermore, the dynamic management of battery health status in wind power and energy storage systems is disconnected from charge and discharge constraints. Existing methods generally employ a static SOH-DOD mapping strategy based on periodic monitoring, relying on passive adjustments based on human experience. This approach cannot respond in real time to the dynamic decay of SOH under complex operating conditions such as cycle count, temperature, and current. Secondly, the lag and subjectivity of the adjustment behavior lead to the energy storage battery operating at unreasonable charge and discharge depths for extended periods, which not only accelerates its lifespan degradation but also harbors potential safety risks of overcharging / over-discharging.

[0003] In summary, existing technologies have shortcomings in dynamic management of battery health status, spatial design of scheduling models, and multi-objective adaptive balancing, making it difficult for wind-storage integrated systems to simultaneously ensure the safety and stability of the power grid in the distribution area, the efficient absorption of new energy sources, and the long-term service life of energy storage batteries. Therefore, there is an urgent need for a collaborative scheduling scheme that integrates real-time battery health perception, refined dynamic constraints on DOD (Data Occurrence), and multi-objective adaptive optimization to adapt to complex and ever-changing distribution area operation scenarios. Summary of the Invention

[0004] In view of the above-mentioned prior art, the present invention provides a wind power and energy storage scheduling method based on the charging and discharging depth limitation of battery health, which mainly solves the technical problems existing in the background art.

[0005] To achieve the above objectives, the technical solution of the present invention is implemented as follows: A wind and energy storage scheduling method based on battery health-based charge / discharge depth limitations includes the following steps: Step S1: Obtain real-time load data of the power grid in the distribution area, real-time wind power data of the wind turbine power generation system, and operating status data of the energy storage battery; Step S2: Based on the real-time load data, the real-time wind power data, and the operating status data, obtain the predicted load value of the power grid in the distribution area, the predicted output value of the wind turbine, and the real-time battery health status (SOH). Step S3: Construct a nonlinear mapping relationship between the real-time battery health status (SOH) and the battery charge / discharge depth (DOD) limit range to obtain the dynamic DOD limit range; Step S4: Based on the real-time load data, the real-time wind power data, the operating status data, the predicted load value of the power grid in the distribution area, the predicted output value of the wind turbine, the real-time battery health status (SOH), and the dynamic DOD limit range, a reinforcement learning scheduling model is used to make decisions and generate charging and discharging power commands for the energy storage battery and actual output commands for the wind turbine. Step S5: Based on the charging and discharging power command and the actual output command of the wind turbine, control the energy storage system to charge and discharge within the dynamic DOD range, and adjust the wind turbine output to achieve wind-storage coordinated and optimized scheduling.

[0006] Optionally, the real-time load data includes: real-time active power data, real-time reactive power data, real-time voltage data, and real-time frequency data; The real-time wind power data of the wind turbine power generation system includes: real-time wind power output data and wind speed data; The operating status data includes the remaining SOC data of the energy storage battery, charge / discharge current data, charge / discharge voltage data, cell temperature data, module temperature data, cumulative cycle count data, internal resistance data, initial rated capacity data, and current rated capacity data.

[0007] Optionally, step S2 includes: Based on the Long Short-Term Memory (LSTM) network model, we constructed a power grid load prediction model, a wind turbine output prediction model, and a battery health prediction model, and trained them respectively to obtain the trained power grid load prediction model, wind turbine output prediction model, and battery health prediction model. The predicted load value of the distribution area power grid is obtained based on the trained distribution area power grid load prediction model. The predicted output value of the wind turbine is obtained based on the trained wind turbine output prediction model. The real-time battery health status (SOH) is obtained based on the trained battery health prediction model.

[0008] Optionally, the expression for step S3 is:

[0009]

[0010]

[0011] 50%≤ ≤100% <

[0012] in, These are the fitting coefficients. This is the minimum limit for dynamic DOD. For the health of energy storage batteries, This is the maximum limit value for dynamic DODs.

[0013] Optionally, the reinforcement learning scheduling model is constructed based on the Deep Deterministic Policy Gradient (DDPG) algorithm, and includes: Define a state space, in which the state vectors include: predicted grid load of the distribution area, real-time wind power output, predicted wind turbine output, battery SOC, battery health SOH, cell temperature, cumulative battery cycle count, minimum dynamic DOD limit, and maximum dynamic DOD limit. Define an action space, wherein the action vectors of the action space include: energy storage charging and discharging power, and wind turbine output adjustment coefficient; A reward function is set, and the weight coefficient of the reward function is dynamically adjusted according to the peak, flat and valley periods of the transformer area load.

[0014] Optionally, the expression for the reward function is:

[0015]

[0016]

[0017]

[0018]

[0019] in, Rewards for ensuring the stability of the power grid in the distribution area; Incentives for the consumption of new energy sources; The reward is for the health of energy storage batteries; α is the weighting coefficient for the power grid stability reward in the distribution area; β is the weighting coefficient for the new energy consumption reward; and γ is the weighting coefficient for the health reward of energy storage batteries.

[0020] Optionally, the weighting coefficients of the reward function are dynamically adjusted according to the peak, flat, and valley periods of the transformer area load, including: During peak hours, the focus is on the grid stability reward for the distribution area, and the weighting coefficient of the grid stability reward for the distribution area is the maximum value; During normal periods, the weighting coefficients of the power grid stability reward, the renewable energy consumption reward, and the energy storage battery health reward are balanced. During off-peak hours, the focus is on health rewards for energy storage batteries, with the weighting coefficient for these rewards set to the maximum value.

[0021] Optionally, the step of using a reinforcement learning scheduling model to make decisions and generate charging and discharging power commands for the energy storage battery and actual output commands for the wind turbine includes: The reinforcement learning scheduling model is used to output energy storage charging and discharging power commands and wind turbine output adjustment coefficients. The actual output command of the fan is calculated based on the fan output adjustment coefficient. The calculation formula is as follows:

[0022] in, This is the actual output command of the wind turbine. This is the fan output adjustment coefficient. This refers to the maximum power that the wind turbine can generate in real time.

[0023] The beneficial effects of this invention are as follows: A dynamic coupling mechanism between SOH and operating boundary was established: the battery health (SOH) is converted into a dynamic limit window of charge-discharge depth (DOD) in real time through nonlinear mapping, which avoids the operating conditions that accelerate battery aging from the source of scheduling, avoids the lag and subjectivity of manual adjustment in traditional strategies, and significantly improves the economic efficiency of the energy storage system throughout its entire life cycle.

[0024] A closed-loop decision-making mechanism for multi-objective collaborative optimization was achieved: Based on the reinforcement learning framework, a weighted reward function was constructed with the objectives of grid stability, renewable energy consumption, and battery health. This enabled the reinforcement learning scheduling model to autonomously learn the optimal scheduling strategy that takes into account multiple constraints, solving the optimization problem of conflicting objectives. Its continuous action space output ensured the smoothness and real-time performance of the control, significantly improving the robustness of the system in the face of uncertainty.

[0025] A time-adaptive strategy preference adjustment method is proposed: by dynamically adjusting the weight coefficient of the reward function during peak, flat and valley periods, the scheduling strategy can prioritize grid stability during peak electricity demand and focus on battery maintenance during valley periods, thereby enhancing the flexibility and practicality of actual wind and energy storage scheduling. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating a wind power and energy storage scheduling method based on battery health charge / discharge depth limitations provided in an embodiment of the present invention. Detailed Implementation

[0027] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. In the following description, the expression "some embodiments" refers to a subset of all possible embodiments; however, it should be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with each other without conflict.

[0028] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention can be practiced without one or more of these details. In other instances, certain technical features well-known in the art have not been described in order to avoid obscuring the invention.

[0029] It should be understood that the present invention can be embodied in various forms and should not be construed as being limited to the embodiments set forth herein. Rather, providing these embodiments will make the disclosure thorough and complete, and will fully convey the scope of the invention to those skilled in the art. Furthermore, the terminology used herein is intended only to describe particular embodiments and is not intended to limit the invention. When used herein, the singular forms “a,” “an,” and “the” are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the terms “compose” and / or “comprising,” when used in this specification, identify the presence of the stated features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups. When used herein, the term “and / or” includes any and all combinations of the associated listed items.

[0030] It should also be noted that when an element is referred to as being "fixed to" another element, it can be directly attached to the other element or there may be an intervening element. When an element is referred to as being "connected to" another element, it can be directly connected to the other element or there may be an intervening element. The terms "vertical," "horizontal," "inner," "outer," "left," "right," and similar expressions used herein are for illustrative purposes only and do not represent the only possible implementation.

[0031] To fully understand this invention, a detailed structure will be presented in the following description to illustrate the technical solution proposed by this invention. Optional embodiments of the invention are described in detail below; however, in addition to these detailed descriptions, the invention may have other embodiments.

[0032] Example 1 Please refer to the attached document. Figure 1 This application provides a wind power and energy storage scheduling method based on the charge / discharge depth limitation of battery health, including the following steps: Step S1: Obtain real-time load data of the power grid in the distribution area, real-time wind power data of the wind turbine power generation system, and operating status data of the energy storage battery; Step S2: Based on the real-time load data, the real-time wind power data, and the operating status data, obtain the predicted load value of the power grid in the distribution area, the predicted output value of the wind turbine, and the real-time battery health status (SOH). Step S3: Construct a nonlinear mapping relationship between the real-time battery health status (SOH) and the battery charge / discharge depth (DOD) limit range to obtain the dynamic DOD limit range; Step S4: Based on the real-time load data, the real-time wind power data, the operating status data, the predicted load value of the power grid in the distribution area, the predicted output value of the wind turbine, the real-time battery health status (SOH), and the dynamic DOD limit range, a reinforcement learning scheduling model is used to make decisions and generate charging and discharging power commands for the energy storage battery and actual output commands for the wind turbine. Step S5: Based on the charging and discharging power command and the actual output command of the wind turbine, control the energy storage system to charge and discharge within the dynamic DOD range, and adjust the wind turbine output to achieve wind-storage coordinated and optimized scheduling.

[0033] Specifically, real-time load data of the power grid in the distribution area can reflect the real-time power demand distribution, load fluctuation characteristics, and equipment load status in the area, directly determining the charging and discharging requirements of energy storage batteries. Load analysis can provide a basis for precise allocation of power resources in the future. During peak load periods, the batteries need to discharge to meet power supply needs, and during off-peak periods, they need to charge and store excess energy to avoid wind curtailment or grid overload. The uncertainty of the wind turbine power generation system needs to be regulated by energy storage batteries. When the wind speed suddenly drops and the output decreases, the energy storage batteries need to discharge to compensate. When the output is excessive, the batteries need to be charged and stored to avoid wind curtailment. Real-time wind power data facilitates the adjustment of the operation mode in advance. The operating status data of the energy storage batteries can reflect the operating status of the energy storage batteries and provide a basis for adjusting the charging and discharging strategies in the future. SOH is a key indicator of battery aging, affecting its charging and discharging capacity and lifespan. Real-time calculation of SOH using battery operating status data provides a basis for dynamically adjusting DOD limits. The relationship between SOH and DOD limits is non-linear (the lower the SOH, the smaller the allowed DOD). Real-time SOH data can be used to dynamically adjust the upper limit of DOD to avoid overcharging / over-discharging of the batteries.

[0034] Next, based on real-time load data, real-time wind power data, operating status data, predicted grid load in the distribution area, predicted wind turbine output, real-time battery health status (SOH), and dynamic DOD (Device of Demand) limits, a reinforcement learning scheduling model is used for decision-making. This model can simultaneously optimize multiple conflicting objectives. For example, while maximizing renewable energy absorption and extending battery life, it flexibly adjusts priorities through a reward function to achieve optimal wind-storage coordinated scheduling. Furthermore, in this technical solution, by considering dynamic DOD limits, the charging and discharging of batteries are ensured to be within a safe range. Energy storage is used to regulate the fluctuations in wind turbine output. Dynamic DOD limits ensure that batteries are fully charged when wind turbine output is excessive and reliably discharged when output is insufficient, ultimately achieving wind-storage coordinated optimization scheduling that considers battery life.

[0035] As an optional implementation, the real-time load data includes: real-time active power data, real-time reactive power data, real-time voltage data, and real-time frequency data; The real-time wind power data of the wind turbine power generation system includes: real-time wind power output data and wind speed data; The operating status data includes the remaining SOC data of the energy storage battery, charge / discharge current data, charge / discharge voltage data, cell temperature data, module temperature data, cumulative cycle count data, internal resistance data, initial rated capacity data, and current rated capacity data.

[0036] Specifically, real-time load data of the power grid in the distribution area is collected through the distribution automation system, real-time wind power data of the wind turbine power generation system is collected through the wind turbine controller and the weather station, and operating status data of the energy storage battery is collected through the battery management system (BMS). After obtaining real-time load data of the power grid, real-time wind power data of the wind turbine generation system, and operating status data of the energy storage battery, the data needs to be preprocessed. The specific steps of preprocessing are as follows: outliers are identified using the physical constraint method and removed. The physical constraints include limiting the SOC to between 0-100%, the charging and discharging current not exceeding 1.5 times the rated current, and the voltage not exceeding ±10% of the rated voltage. For short-term missing data, linear interpolation or forward imputation is used to complete the data. The Min-Max normalization method is used to standardize the completed data.

[0037] As an optional implementation, step S2 includes: Based on the Long Short-Term Memory (LSTM) network model, we constructed a power grid load prediction model, a wind turbine output prediction model, and a battery health prediction model, and trained them respectively to obtain the trained power grid load prediction model, wind turbine output prediction model, and battery health prediction model. The predicted load value of the distribution area power grid is obtained based on the trained distribution area power grid load prediction model. The predicted output value of the wind turbine is obtained based on the trained wind turbine output prediction model. The real-time battery health status (SOH) is obtained based on the trained battery health prediction model.

[0038] Specifically, LSTM (Long Short-Term Memory) is a special type of recurrent neural network suitable for processing long-term dependencies in time-series data. Distribution area loads are affected by user behavior, seasons, and other factors, exhibiting daily, weekly, or even seasonal patterns. Wind power is influenced by meteorological factors such as wind speed and direction, with wind speed changes potentially lasting for hours or even days. Battery health (SOH) degradation is a cumulative process, affected by factors such as the number of charge-discharge cycles and temperature. Using LSTM network models to construct distribution area grid load prediction models, wind turbine output prediction models, and battery health prediction models can effectively capture the distribution area's power grid load. The long-term historical information in the real-time load data of the power grid, the long-term trend in the real-time wind power data of the wind turbine power generation system, and the impact of historical operations on the current state in the operating status data of the energy storage battery are all considered. At the same time, since the influencing factors of the power grid load, wind power output, and energy storage battery health status (SOH) are nonlinear, the LSTM network model is used to construct the power grid load prediction model, wind turbine output prediction model, and battery health status prediction model. This model can fit the complex patterns well, resulting in more accurate predicted values ​​of power grid load, predicted wind turbine output, and real-time battery health status (SOH).

[0039] For example, based on the preprocessed real-time load data of the distribution network, the load data for the next 24 hours is predicted using the distribution network load prediction model to obtain the distribution network load prediction value; based on the real-time wind power data of the wind turbine power generation system, the wind turbine output data for the next 24 hours is predicted using the wind turbine output prediction model to obtain the wind turbine predicted output value; based on the operating status data (voltage, current, temperature, internal resistance) of the energy storage battery, the battery health is calculated in real time using the battery health prediction model to obtain the real-time battery health status (SOH).

[0040] As an optional implementation, the expression for step S3 is:

[0041]

[0042]

[0043] 50%≤ ≤100% <

[0044] in, These are the fitting coefficients. This is the minimum limit for dynamic DOD. For the health of energy storage batteries, This is the maximum limit value for dynamic DODs.

[0045] For example, at a specific moment in this embodiment, the battery health status (SOH) output by the model is 85%. Using a decay model with fitting coefficients calibrated by the least squares method, the calculated battery health status of 85% is substituted into the nonlinear mapping function characterizing the relationship between SOH and DOD to calculate the charge / discharge depth limit range under the current health state. In this embodiment, the calculation result is: It is 20%. The limit is 95%, meaning that the State of Charge (SOC) must not be lower than 80% during discharge and can reach up to 95% during charging.

[0046] It should be noted that existing methods generally adopt a static SOH-DOD mapping strategy based on periodic monitoring. That is, based on a pre-established fixed relationship between SOH and DOD (such as a lookup table method), the corresponding estimated SOH value is directly output by inputting the current DOD value. Under this method, after determining the mapping relationship, it does not dynamically adjust with real-time operating conditions (such as temperature and current fluctuations) during battery use, resulting in poor adaptability to dynamic operating conditions. In contrast, in this technical solution, the obtained real-time battery health SOH parameter sequence is substituted into the formula to calculate the battery charge / discharge depth DOD limit range. That is, the upper and lower limits of DOD are obtained according to the actual situation of the energy storage battery, which better adapts to dynamic characteristics.

[0047] As an optional implementation, the reinforcement learning scheduling model is constructed based on the Deep Deterministic Policy Gradient (DDPG) algorithm, and includes: Define a state space, in which the state vectors include: predicted grid load of the distribution area, real-time wind power output, predicted wind turbine output, battery SOC, battery health SOH, cell temperature, cumulative battery cycle count, minimum dynamic DOD limit, and maximum dynamic DOD limit. Define an action space, wherein the action vectors of the action space include: energy storage charging and discharging power, and wind turbine output adjustment coefficient; A reward function is set, and the weight coefficient of the reward function is dynamically adjusted according to the peak, flat and valley periods of the transformer area load.

[0048] As an optional implementation, the expression for the reward function is:

[0049]

[0050]

[0051]

[0052]

[0053] in, Rewards for ensuring the stability of the power grid in the distribution area; Incentives for the consumption of new energy sources; The reward is for the health of energy storage batteries; α is the weighting coefficient for the power grid stability reward in the distribution area; β is the weighting coefficient for the new energy consumption reward; and γ is the weighting coefficient for the health reward of energy storage batteries.

[0054] Specifically, grid stability in a distribution area may require rapid response from energy storage batteries, leading to frequent charging and discharging, which can damage battery health. New energy consumption may require prolonged charging and discharging of energy storage batteries, increasing DOD (Distance Over Time) and accelerating aging. Battery health protection may limit battery charging and discharging capabilities, affecting grid stability and consumption efficiency. Therefore, during wind-storage scheduling, by determining the weighting coefficients of the reward function, the rewards for grid stability, new energy consumption, and battery health can be dynamically balanced. This guides the wind-storage system to maximize new energy utilization efficiency while extending energy storage lifespan, thus maximizing its full lifecycle benefits, all while ensuring grid security.

[0055] As an optional implementation, the weighting coefficients of the reward function are dynamically adjusted according to the peak, flat, and valley periods of the transformer area load, including: During peak hours, the focus is on the grid stability reward for the distribution area, and the weighting coefficient of the grid stability reward for the distribution area is the maximum value; During normal periods, the weighting coefficients of the power grid stability reward, the renewable energy consumption reward, and the energy storage battery health reward are balanced. During off-peak hours, the focus is on health rewards for energy storage batteries, with the weighting coefficient for these rewards set to the maximum value.

[0056] For example, the weighting coefficients during peak hours are: α=0.6, β=0.2, γ=0.2; the weighting coefficients during normal hours are: α=0.33, β=0.34, γ=0.33; and the weighting coefficients during trough hours are: α=0.2, β=0.2, γ=0.6.

[0057] As an optional implementation, the step of using a reinforcement learning scheduling model to make decisions and generate charging and discharging power commands for the energy storage battery and actual output commands for the wind turbine includes: The reinforcement learning scheduling model is used to output energy storage charging and discharging power commands and wind turbine output adjustment coefficients. The actual output command of the fan is calculated based on the fan output adjustment coefficient. The calculation formula is as follows:

[0058] in, This is the actual output command of the wind turbine. This is the fan output adjustment coefficient. This refers to the maximum power that the wind turbine can generate in real time.

[0059] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A wind power and energy storage scheduling method based on battery health-related charge / discharge depth limitations, characterized in that, Includes the following steps: Step S1: Obtain real-time load data of the power grid in the distribution area, real-time wind power data of the wind turbine power generation system, and operating status data of the energy storage battery; Step S2: Based on the real-time load data, the real-time wind power data, and the operating status data, obtain the predicted load value of the power grid in the distribution area, the predicted output value of the wind turbine, and the real-time battery health status (SOH). Step S3: Construct a nonlinear mapping relationship between the real-time battery health status (SOH) and the battery charge / discharge depth (DOD) limit range to obtain the dynamic DOD limit range; Step S4: Based on the real-time load data, the real-time wind power data, the operating status data, the predicted load value of the power grid in the distribution area, the predicted output value of the wind turbine, the real-time battery health status (SOH), and the dynamic DOD limit range, a reinforcement learning scheduling model is used to make decisions and generate charging and discharging power commands for the energy storage battery and actual output commands for the wind turbine. Step S5: Based on the charging and discharging power command and the actual output command of the wind turbine, control the energy storage system to charge and discharge within the dynamic DOD range, and adjust the wind turbine output to achieve wind-storage coordinated and optimized scheduling.

2. The wind and energy storage scheduling method based on battery health charge / discharge depth limitations according to claim 1, characterized in that, The real-time load data includes: real-time active power data, real-time reactive power data, real-time voltage data, and real-time frequency data; The real-time wind power data of the wind turbine power generation system includes: real-time wind power output data and wind speed data; The operating status data includes the remaining SOC data of the energy storage battery, charge / discharge current data, charge / discharge voltage data, cell temperature data, module temperature data, cumulative cycle count data, internal resistance data, initial rated capacity data, and current rated capacity data.

3. The wind and energy storage scheduling method based on battery health charge / discharge depth limitations according to claim 1, characterized in that, Step S2 includes: Based on the Long Short-Term Memory (LSTM) network model, we constructed a power grid load prediction model, a wind turbine output prediction model, and a battery health prediction model, and trained them respectively to obtain the trained power grid load prediction model, wind turbine output prediction model, and battery health prediction model. The predicted load value of the distribution area power grid is obtained based on the trained distribution area power grid load prediction model. The predicted output value of the wind turbine is obtained based on the trained wind turbine output prediction model. The real-time battery health status (SOH) is obtained based on the trained battery health prediction model.

4. The wind and energy storage scheduling method based on battery health charge / discharge depth limitations according to claim 1, characterized in that, The expression for step S3 is: 50%≤ ≤100% < in, These are the fitting coefficients. This is the minimum limit for dynamic DOD. For the health of energy storage batteries, This is the maximum limit value for dynamic DODs.

5. The wind and energy storage scheduling method based on battery health charge / discharge depth limitations according to claim 1, characterized in that, The reinforcement learning scheduling model is constructed based on the Deep Deterministic Policy Gradient (DDPG) algorithm, and includes: Define a state space, in which the state vectors include: predicted grid load of the distribution area, real-time wind power output, predicted wind turbine output, battery SOC, battery health SOH, cell temperature, cumulative battery cycle count, minimum dynamic DOD limit, and maximum dynamic DOD limit. Define an action space, wherein the action vectors of the action space include: energy storage charging and discharging power, and wind turbine output adjustment coefficient; A reward function is set, and the weight coefficient of the reward function is dynamically adjusted according to the peak, flat and valley periods of the transformer area load.

6. A wind power and energy storage scheduling method based on battery health charge / discharge depth limitations according to claim 5, characterized in that, The expression for the reward function is: in, Rewards for ensuring the stability of the power grid in the distribution area; Incentives for the consumption of new energy sources; The reward is for the health of energy storage batteries; α is the weighting coefficient for the power grid stability reward in the distribution area; β is the weighting coefficient for the new energy consumption reward; and γ is the weighting coefficient for the health reward of energy storage batteries.

7. A wind power and energy storage scheduling method based on battery health charge / discharge depth limitations according to claim 6, characterized in that, The weighting coefficients of the reward function are dynamically adjusted according to the peak, flat, and valley periods of the transformer area load, including: During peak hours, the focus is on the grid stability reward for the distribution area, and the weighting coefficient of the grid stability reward for the distribution area is the maximum value; During normal periods, the weighting coefficients of the power grid stability reward, the renewable energy consumption reward, and the energy storage battery health reward are balanced. During off-peak hours, the focus is on health rewards for energy storage batteries, with the weighting coefficient for these rewards set to the maximum value.

8. The wind and energy storage scheduling method based on battery health charge / discharge depth limitations according to claim 1, characterized in that, The process of using a reinforcement learning scheduling model to make decisions and generate charging and discharging power commands for the energy storage battery and actual output commands for the wind turbine includes: The reinforcement learning scheduling model is used to output energy storage charging and discharging power commands and wind turbine output adjustment coefficients. The actual output command of the fan is calculated based on the fan output adjustment coefficient. The calculation formula is as follows: in, This is the actual output command of the wind turbine. This is the fan output adjustment coefficient. This refers to the maximum power that the wind turbine can generate in real time.

Images