Wind storage system and grid-connected power adaptive regulation method thereof

Abstract

The application discloses a wind storage system and a grid-connected power adaptive regulation method thereof, and relates to the technical field of new energy sources. In the method, the grid-connected regulation process is divided into a first control stage and a second control stage according to the acquisition time of actual wind power output data and expected grid-connected power data; the grid-connected power plan quantity is determined in the first control stage; the wind power component, the energy storage power component and the actual grid-connected power quantity are determined in the second control stage; the regulation state quantity is adjusted bidirectionally according to power deviation, state of charge deviation and prediction deviation, and the adjusted regulation state quantity is used for the next round of regulation. The application can solve the problem that the grid-connected power plan quantity does not match the actual bearing capacity of the wind storage system in the related art, and make the actual grid-connected power quantity close to the grid-connected power plan quantity.

Description

Technical Field

[0001] This application relates to the field of new energy technology, and in particular to a wind-storage system and its grid-connected power adaptive regulation method. Background Technology

[0002] A wind-storage system refers to a power generation system formed by an electrical connection between a wind farm and an energy storage power station. The two systems converge at a common grid connection point and are then connected to the grid as a whole. Wind power output is random and intermittent, and there is a discrepancy between predicted and actual wind power output. To address this, the wind-storage system employs an adaptive grid-connected power control method that determines the planned grid-connected power output based on the current state of the system, replacing the traditional grid-connected power control method based on a fixed power target.

[0003] When determining the planned grid-connected power of a wind-storage system, the relevant technologies use a fixed risk preference parameter. This risk preference parameter is not adjusted according to changes in the state of charge of the energy storage equipment, the prediction deviation of wind power output forecast data relative to actual wind power output data, and historical grid connection deviations. This results in a large deviation between the actual grid-connected power and the planned grid-connected power.

[0004] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] This application provides a wind-storage system and its grid-connected power adaptive control method, which makes the actual grid-connected power approach the planned grid-connected power.

[0006] To solve the above-mentioned technical problems, this application provides the following technical solution: This application provides a method for adaptive regulation of grid-connected power in a wind-storage system, including: Based on the acquisition time of actual wind power output data and expected grid-connected power data, the grid-connected control process of the wind-storage system is divided into the first control stage and the second control stage. In the first control phase, the planned grid-connected power is determined based on the predicted wind power output data, the predicted values ​​of the expected grid-connected power data, the current state of charge of the energy storage device, and the control state variables; the values ​​of the control state variables are within a preset range. In the second control phase, based on the actual output data of wind power, the actual value of the expected grid-connected power data, and the planned grid-connected power, the wind power component and the energy storage power component are determined respectively, and based on the wind power component and the energy storage power component, the actual grid-connected power of the wind-storage system is determined. The control state quantity is adjusted bidirectionally based on the power deviation between the actual grid-connected power and the planned grid-connected power, the charge deviation of the energy storage device from the preset safe range, and the prediction deviation of the wind power output prediction data relative to the actual wind power output data. The adjusted control state quantity will be used in the first control stage of the next round of grid connection control process.

[0007] Another aspect of this application provides a wind power storage system, including a wind farm, an energy storage power station, and a server; the wind farm includes a wind farm power controller, and the energy storage power station includes an energy storage converter; The server includes at least a processor, memory, and a communication interface; The processor acquires grid-connected power expectation data through the communication interface and implements the above-mentioned wind-storage system grid-connected power adaptive control method by executing the computer program stored in the memory; it generates wind power dispatch power instructions based on the wind power components and generates energy storage charging and discharging power instructions based on the energy storage power components. The processor sends the wind power dispatch power command to the wind farm power controller and the energy storage charging and discharging power command to the energy storage converter through the communication interface.

[0008] The advantage of the technical solution provided in this application lies in dividing the grid connection control process of the wind-storage system into a first control stage and a second control stage according to the acquisition time of actual wind power output data and expected grid-connected power data. This ensures that the first control stage determines the planned output before data acquisition, and the second control stage allocates the actual power after data acquisition. This avoids the problem of timing misalignment between planned and actual execution caused by not dividing the stages according to the information availability point, and ensures that the deviation reference relationship between the planned and actual grid-connected power matches the data availability point. Simultaneously, the introduction of a control state variable with a limited value range in the first control stage participates in the determination of the planned output, allowing the planned grid-connected power to continuously change within a fixed range with the control state variable, and then, after the second control stage, allocates the power based on the actual grid connection. The power deviation from the planned output, the charge deviation from the safe range, and the prediction deviation are used to bidirectionally adjust the control state quantity. This adjusted control state quantity is then returned to the next round of the first control stage. This allows the control state quantity to accumulate and decrease with each round of control, causing the planned output to increase or decrease accordingly. This enables the grid-connected power plan to dynamically adjust based on the current charge pressure of the wind-storage system, wind power output prediction deviation, and historical grid connection deviation. This avoids the mismatch between the grid-connected power plan and the actual risk tolerance of the wind-storage system caused by fixed risk preference parameters. Therefore, when actual wind power output fluctuates, the tracking deviation of the actual grid-connected power from the planned output can be corrected round by round, making the actual grid-connected power approach the planned grid-connected power. Furthermore, this application also provides a corresponding wind-storage system for the adaptive control method of grid-connected power in wind-storage systems, further enhancing the practicality of the method and providing the wind-storage system with corresponding advantages.

[0009] The technical features mentioned above, those to be mentioned below, and those shown individually in the accompanying drawings can be arbitrarily combined, as long as the combined technical features are not contradictory. All feasible combinations of features are the technical content explicitly described in this application. Any one of the multiple sub-features contained in the same statement can be applied independently, without necessarily being applied together with other sub-features.

[0010] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit this application. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of this application or related technologies, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A flowchart illustrating an adaptive control method for grid-connected power of a wind-storage system provided in this application; Figure 2 A structural framework diagram of an exemplary embodiment of the grid-connected power adaptive control device for the wind-storage system provided in this application; Figure 3 A structural diagram of an exemplary embodiment of the electronic device provided in this application; Figure 4 This is a structural diagram of an exemplary embodiment of the wind storage system provided in this application. Detailed Implementation

[0013] To enable those skilled in the art to better understand the technical solutions of this application, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. The terms "first," "second," "third," "fourth," etc., used in the specification and the aforementioned drawings are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. The term "exemplary" means "serving as an example, embodiment, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as superior to or better than other embodiments.

[0014] The wind-storage system consists of a wind farm and an energy storage power station. The wind farm converts wind energy into electrical energy and connects it to the grid via collector lines. The energy storage power station is connected to the same grid connection point or a nearby busbar, utilizing the rapid charging and discharging capabilities of energy storage to absorb excess wind energy or make up for output gaps. The wind-storage system periodically acquires wind power output data and expected grid-connected power data, and issues wind power dispatching power commands and energy storage charging and discharging power commands. As the proportion of wind power in the power system increases, wind power output exhibits randomness and intermittency. There is a deviation between the predicted wind power output data and the actual wind power output data. When the deviation between the actual grid-connected power and the expected grid-connected power exceeds the allowable range, deviation assessment fees will be incurred.

[0015] The grid connection control process of wind and energy storage systems is asynchronous in time. Before obtaining the actual values ​​of wind power output data and expected grid connection power data, the planned grid connection power for this round needs to be determined as a reference for subsequent deviations and the basis for dispatch pressure allocation. Only after obtaining the actual values ​​of wind power output data and expected grid connection power data can the wind power component and energy storage power component be allocated. The planned grid connection power in the previous stage determines the grid connection power deviation reference and control pressure in the next stage. The allocation results of wind power component and energy storage power component in the next stage affect the state of charge of energy storage equipment, deviation assessment costs, and the deviation risk that can be tolerated in the next round of grid connection control. This forms a characteristic of the two control stages being interconnected and the risk tolerance capacity dynamically changing. In the application scenario of the real-time electricity market, around each round of grid connection control, the wind and energy storage system sequentially executes capacity declaration, market clearing, real-time energy management, market settlement, and system status update. Among them, capacity declaration occurs before market clearing, real-time energy management occurs after market clearing, and market settlement and system status update are carried out after the first two stages are completed.

[0016] Regarding the aforementioned grid connection regulation process, some related technologies directly treat the allocation of wind power and energy storage power components during grid connection regulation as the output action of a single reinforcement learning model, without considering the impact of the grid-connected power plan on subsequent allocation. The fixed output action parameters are difficult to adjust according to changes in state-of-charge pressure, prediction deviation, and historical grid connection deviation. Other related technologies, although employing hierarchical reinforcement learning to handle tasks at different time scales, typically base their hierarchical division on time scale or task splitting, without considering the actual differences in the available data points during the wind-storage system grid connection regulation process. They also fail to reflect the impact of the grid-connected power plan on subsequent deviation calculations and power allocation, as well as the feedback effect of the power allocation results on the determination conditions of the next round of grid-connected power plan.

[0017] To address the risk of deviations with low probability but high losses, related technologies also employ methods such as conditional value of risk, risk thresholds, scenario risk costs, or robust uncertainty sets. However, such risk preferences are typically determined by fixed confidence levels, fixed risk weights, or fixed thresholds. They cannot be adjusted based on changes in risk tolerance caused by historical deviation assessment costs, the state-of-charge pressure of energy storage equipment, and deviations in wind power output forecasts. This results in the same planned grid-connected power output potentially having similar expected returns but significantly different tail losses under different operating trajectories, making it difficult to achieve a dynamic balance between return targets and deviation risk control as the wind-storage system status changes.

[0018] To address the issues in the aforementioned related technologies where the planned grid-connected power adopts a fixed risk preference and is mismatched with the actual capacity of the wind-storage system, this application provides an adaptive grid-connected power control method for wind-storage systems. Based on the acquisition time of actual wind power output data and expected grid-connected power data, the grid-connected control process is divided into a first control stage and a second control stage. This ensures that the planned grid-connected power is determined before the actual data is revealed, and power component allocation is performed after the actual data is revealed. In the first control stage, the planned grid-connected power is determined based on the predicted wind power output data, the predicted value of the expected grid-connected power data, the current state of charge of the energy storage equipment, and the control status variables. This allows the planned grid-connected power to adapt to the current deviation capacity of the wind-storage system within a preset range. In the second control stage, the wind power component and energy storage power component are determined based on the actual wind power output data, the actual value of the expected grid-connected power data, and the planned grid-connected power, respectively, thereby determining the actual grid-connected power of the wind-storage system. Then, based on the power deviation between the actual and planned grid-connected power, the charge deviation of the energy storage device from the preset safe range, and the prediction deviation of the wind power output forecast data relative to the actual wind power output data, the control state quantity is adjusted bidirectionally. This causes the control state quantity to contract in the decreasing direction when the deviation exceeds the allowable range and recover in the increasing direction during continuous operation. The adjusted control state quantity is used in the first control stage of the next round of grid-connected control process, thereby making the actual grid-connected power approach the planned grid-connected power. After introducing the technical solution of this application, various non-limiting embodiments of this application are described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] Please see first. Figure 1 According to the adaptive regulation method for grid-connected power of a wind-storage system provided in this application, it can be implemented as a computer program product, installed and run in the management platform or control platform of the wind-storage system, for adaptive regulation of the grid-connected power of the wind-storage system. In some embodiments of the method, the method includes the following steps: S101: Based on the acquisition time of actual wind power output data and expected grid-connected power data, the grid-connected control process of the wind-storage system is divided into the first control stage and the second control stage.

[0020] The wind-storage system is an integrated power generation unit consisting of a wind farm and energy storage equipment connected to the grid. Actual wind power output data refers to the active power actually output by the wind farm in the current time period. Expected grid-connected power data is the current or predicted value of the active power that the wind-storage system is expected to deliver to the grid, issued by the power dispatching agency, representing the grid's demand for grid-connected power from the wind-storage system. Acquisition time refers to the moment when the above two types of data are provided by the relevant measurement or prediction modules and can be read by the wind-storage system. The first control phase refers to the period before the actual values ​​of the actual wind power output data and expected grid-connected power data are revealed; the second control phase refers to the period after the actual values ​​of the actual wind power output data and expected grid-connected power data are revealed. The first control phase is executed before the acquisition of the actual wind power output data and expected grid-connected power data, and the second control phase is executed after the acquisition of the actual wind power output data and expected grid-connected power data. One round of grid-connected regulation and control consists of a first control phase and a subsequent second control phase.

[0021] In this step, for ease of description, the duration of a single round of grid connection control is defined as the control period length, and the sequence number of each round of grid connection control is defined as the control period index. The wind-storage system divides each operating day into multiple control periods, and each control period corresponds to one round of grid connection control. For example, each operating day is divided into multiple periods of length... During the period of regulation, among which For the index of the control period, This refers to the duration of a single control period. For example, if the single dispatch update cycle is 15 minutes, each operating day is divided into 96 control periods of equal length; if the single dispatch update cycle is 5 minutes, each operating day is divided into 288 control periods of equal length; and if the single dispatch update cycle is 1 minute, each operating day is divided into 1440 control periods of equal length. Of course, in addition to dividing by equal length, control periods can also be divided by the update time of the grid-connected power expectation data, as long as the actual wind power output data and the actual grid-connected power expectation data can be read within the same time window within a single control period. This application does not impose any restrictions on this. In the application scenario of the real-time electricity market, around each round of grid-connected control process, the wind-storage system sequentially executes capacity declaration, market clearing, real-time energy management, market settlement, and system status update; the first control stage corresponds to the capacity declaration based on forecast information before market clearing, the second control stage corresponds to the real-time energy management based on actual data after market clearing, and market settlement and system status update are based on the execution results of the first two control stages.

[0022] S102: In the first control phase, the planned grid-connected power is determined based on the predicted wind power output data, the predicted values ​​of the expected grid-connected power data, the current state of charge of the energy storage device, and the control state variables.

[0023] Among them, wind power output forecast data is the prediction result of the actual wind power output data for the current period. Grid-connected power expectation data is the prediction result of the grid-connected power expectation data for the current period issued by the dispatching agency. Energy storage equipment includes electrochemical energy storage units and their converters; the current state of charge of the energy storage equipment refers to the ratio of the remaining electricity to the rated electricity at the start of the first control phase, with a value range of 0 to 1. The control state quantity is a bounded variable reflecting the degree of deviation risk borne by the wind-storage system, and its value range is within a preset range interval. The preset range interval is a closed interval formed by the upper and lower bounds of the control state quantity, such as [0, 1]. The planned grid-connected power is the target grid-connected power value that the wind-storage system attempts to achieve in the second control phase, determined in the first control phase of this round.

[0024] In this step, the predicted values ​​of wind power output forecast data, grid-connected power expectation data, the current state of charge of energy storage devices, and control status variables are used as inputs, and the planned grid-connected power is output. For example, the input data for the first control stage can be combined as follows: ,in, This is the set of input data for the first control stage. This indicates the first control phase. For wind power output forecast data, This is a predicted value for the settlement price of grid-connected electricity, and it belongs to the predicted value of the expected grid-connected power data. This represents the current state of charge of the energy storage device. It is at least one of the following: time characteristics, historical wind power dispatch results, and historical grid connection deviation information. To regulate the state variable, its value range is: , Represents the set consisting of the listed state variables; the output data satisfies ,Right now ,in, This is the output of the first control stage. This is the planned grid-connected power output. For the maximum grid-connected power of the wind-storage system, the interval This indicates the permissible range of the planned grid-connected power capacity. This refers to a wind-storage system. In the first control stage, the control state variable is used as one of the input data to determine the grid-connected power plan, causing the grid-connected power plan to change with the value of the control state variable. For example, various methods can be used to map the input data to the grid-connected power plan, such as neural networks, preset mapping tables, or rule-based calculations based on the control state variable. The specific form of the mapping relationship is not a limitation of this step, as long as changes in the value of the control state variable can cause corresponding changes in the grid-connected power plan; this application does not impose any limitations on this. In addition to the predicted values ​​of wind power output forecast data, expected grid-connected power data, the current state of charge of the energy storage equipment, and the control state variable, the predicted value of the grid-connected power settlement price, as well as time characteristics, historical wind power dispatch results, and historical grid-connected deviation information are optional supplementary input data. Including these supplementary inputs allows the grid-connected power plan to more fully reflect the deviation tolerance capacity of the wind-storage system.

[0025] S103: In the second control stage, based on the actual output data of wind power, the actual value of the expected grid-connected power data, and the planned grid-connected power, the wind power component and the energy storage power component are determined respectively, and based on the wind power component and the energy storage power component, the actual grid-connected power of the wind-storage system is determined.

[0026] Among them, the wind power component is the wind power component that is dispatched and utilized by the wind farm in the second control phase. Energy storage power component. It is the charging and discharging power component of the energy storage device in the second control phase. Indicates discharge, This indicates charging. The actual grid-connected power is the sum of the wind power component and the energy storage power component, determined by the physical constraints of the wind farm and the energy storage equipment.

[0027] In this step, after reading the actual values ​​of the wind power output data and the expected grid-connected power data, the wind power component and the energy storage power component are determined based on the actual values ​​of the wind power output data, the expected grid-connected power data, and the planned grid-connected power. Then, the wind power component and the energy storage power component are combined to synthesize the actual grid-connected power, satisfying the following requirements. ,in, This represents the actual grid-connected power. For wind power components, This refers to the energy storage power component. For example, the input data for the second control stage can be combined as follows: ,in, This is the set of input data for the second control stage. This indicates the second control phase. This refers to the actual power output corresponding to the wind power generation data. The actual value of the settlement price for grid-connected electricity. The actual value of the grid-connected power expectation data is typically used in practical applications as a market clearing instruction; the output data can be... ,in, This is the output of the second control phase. The actual value of the grid-connected power expectation data is formed by the planned grid-connected power and external conditions, such as load, network status, and other grid-connected entity declarations. In an implementation where the wind-storage system acts as a price taker and network congestion is not considered, the following can be adopted: For example, one feasible allocation method is to make the wind power components satisfy... ,in, The actual power output data of wind power corresponds to the available power, so that the energy storage power component meets the requirements. ,in, This is the absolute value of the preset maximum charging power. To preset the maximum discharge power, the upper horizontal line indicates the upper limit of the corresponding physical quantity, allowing the state of charge of the energy storage device to be determined according to... Transfer, and satisfy ,in, and These are charging efficiency and discharging efficiency, respectively. For the rated energy of energy storage, , This is the preset lower limit of charge for energy storage devices. This is the preset upper limit of the charge capacity for the energy storage device. When the actual wind power output exceeds the planned grid-connected power, the excess power is absorbed by the charging power in the energy storage power component. When the actual wind power output is less than the planned grid-connected power, the shortfall is made up by the discharging power in the energy storage power component. The unutilized wind power can be expressed as... Alternatively, the probability distribution parameters of the wind power component and the energy storage power component can be obtained by sampling the output of the neural network, or by performing rule calculations based on the smaller value between the actual value of the expected grid-connected power data and the planned grid-connected power. As long as the wind power component and the energy storage power component can be combined to synthesize the actual grid-connected power, this application does not impose any restrictions on this.

[0028] S104: Adjust the control state quantity bidirectionally based on the power deviation between the actual grid-connected power and the planned grid-connected power, the charge deviation of the energy storage device from the preset safe range, and the prediction deviation of the wind power output prediction data relative to the actual wind power output data.

[0029] Among them, power deviation is the dimensionless result of the deviation between the actual grid-connected power and the planned grid-connected power. Charge deviation is the distance by which the state of charge of energy storage devices deviates from the preset safe range. The preset safe range is a closed interval formed by pre-set safe upper and lower bounds of the state of charge. Prediction deviation is the dimensionless result of the prediction deviation between the predicted wind power output data and the actual wind power output data. Bidirectional adjustment means that the control state quantity can be adjusted both in the increasing direction and in the decreasing direction.

[0030] In this step, after the allocation of wind power components and energy storage power components is completed in the second control stage, and the actual grid-connected power is determined, the power deviation, charge deviation, and prediction deviation are calculated separately. Then, these three deviations are used as inputs to adjust the control state variables bidirectionally. For example, the power deviation can be the dimensionless value after normalization of the absolute value of the difference between the actual grid-connected power and the planned grid-connected power; the charge deviation can be the dimensionless value after normalization of the distance between the charge state of the energy storage device and the preset safe range; and the prediction deviation can be the dimensionless value after normalization of the absolute value of the deviation between the predicted wind power output data and the actual wind power output data. Other measurement methods that can reflect the above three deviations can also be used. After obtaining the three deviations, the adjustment amount of the control state variables in the decreasing direction can be obtained through various methods such as weighted summation, taking the maximum value, or mapping and summing separately. The control state variables are then adjusted in the increasing direction when the conditions are continuously met. This application does not impose any limitations on the value of the control state variables as long as they remain within the preset range.

[0031] S105: The adjusted control state quantity will be used in the first control stage of the next round of grid connection control process.

[0032] In this step, after completing the output actions of the second control phase of this round and calculating the actual grid-connected power, power deviation, charge deviation, and predicted deviation, the adjusted control state quantity is transmitted to the first control phase of the next round of grid-connected control process as input, forming a transmission chain of deviation risk between adjacent grid-connected control processes. This bidirectional adjustment process only changes the values ​​of the control state quantity, without changing the value ranges of the wind power component and the energy storage power component, the charge state transition equation of the energy storage device, or the synthesis method of the actual grid-connected power, ensuring that the power allocation result is always executable.

[0033] In the technical solution provided in this application embodiment, the grid connection control process of the wind-storage system is divided into a first control stage and a second control stage according to the acquisition time of the actual wind power output data and the expected grid connection power data. This ensures that the first control stage determines the planned quantity before data acquisition, and the second control stage allocates the actual power after data acquisition. This avoids the problem of timing misalignment between the planned value and the actual execution caused by not dividing the stages according to the information availability point, and ensures that the deviation reference relationship between the planned grid connection power and the actual grid connection power matches the data availability point. Simultaneously, a control state quantity with a limited value range is introduced into the first control stage to participate in the determination of the planned quantity, allowing the planned grid connection power to continuously change within a fixed range with the control state quantity, and the actual grid connection power is allocated after the second control stage. The power deviation from the planned power, the charge deviation from the safe range, and the prediction deviation are used to adjust the control state quantity bidirectionally. The adjusted control state quantity is then returned to the first control stage of the next round. This allows the control state quantity to increase or decrease with the cumulative deviation generated in each round of control, and to increase or decrease the planned power in the next round accordingly. This enables the grid-connected power planned power to be dynamically adjusted according to the current charge pressure of the wind and energy storage system, the wind power output prediction deviation, and the historical grid connection deviation. This avoids the problem of mismatch between the grid-connected power planned power and the actual risk-bearing capacity of the wind and energy storage system caused by fixed risk preference parameters. Thus, when there are fluctuations in the actual wind power output, the tracking deviation of the actual grid-connected power to the planned power can be corrected round by round, so that the actual grid-connected power approaches the grid-connected power planned power.

[0034] In the above embodiments, there is no limitation on how to adjust the control state variables bidirectionally. To avoid the problems of inconsistent dimensions of the three deviations and the inability of the control state variables to recover their values ​​after continuous operation, this embodiment also provides an exemplary bidirectional adjustment implementation method, which may include the following steps: Power deviation, charge deviation, and prediction deviation are normalized separately. The normalized power deviation, charge deviation, and prediction deviation are then weighted and summed according to their respective weights to obtain the multi-source risk cost. Based on the multi-source risk cost and a preset risk consumption coefficient, the risk consumption amount is determined. If the power deviation is lower than a preset deviation threshold, the prediction deviation is lower than a preset prediction error threshold, and the state of charge is within a preset safe range, then the recovery amount is determined based on the upper limit difference between the current value of the control state quantity and its upper limit value. The value of the control state quantity is adjusted in the direction of increasing recovery amount, while simultaneously, the value of the control state quantity is adjusted in the direction of decreasing risk consumption. The recovery amount increases with the increase of the upper limit difference, and the decrease in the control state quantity increases with the increase of risk consumption. If the power deviation is not lower than a preset deviation threshold, or the prediction deviation is not lower than a preset prediction error threshold, or the state of charge is not within a preset safe range, then the value of the control state quantity is adjusted in the direction of decreasing risk consumption, and the decrease in the control state quantity increases with the increase of risk consumption.

[0035] Among them, multi-source risk cost is a comprehensive risk index obtained by normalizing and weighting power deviation, charge deviation, and prediction deviation. The preset risk consumption coefficient is a manually set proportional coefficient that converts multi-source risk cost into risk consumption. Risk consumption is the product of multi-source risk cost and the preset risk consumption coefficient. The preset deviation threshold is a threshold for judging whether the power deviation is too large. The preset prediction error threshold is a threshold for judging whether the prediction deviation is too large. Recovery amount is the amount by which the control state quantity is adjusted in the increasing direction.

[0036] In this embodiment, the power deviation, charge deviation, and prediction deviation are normalized by dividing each by its respective preset reference value, and then weighted summation is performed according to their weights to obtain the multi-source risk cost. For example, the multi-source risk cost can be obtained through... The calculation yielded, where For multi-source risk costs, Indicates risk, , and These are the normalized power deviation, charge deviation, and prediction deviation, respectively. , and The non-negative weights of the above three items are assigned respectively; then the risk consumption is calculated according to the preset risk consumption coefficient. ,in Risk consumption amount A preset risk consumption coefficient is used. The condition for the safe operation indication value is: when , and hour, ;otherwise ;in The preset deviation threshold is the threshold for assessing cost deviation. To preset the prediction error threshold, For safe operation indication, Indicates safety conditions. Recovery amount can be... The calculation yielded that, For recovery amount, Indicates recovery. For the safety recovery factor, This is the upper limit difference value. The bidirectional adjustment of the control state quantity satisfies... ,in This is the adjusted control state quantity. This is the amount adjusted according to the decreasing trend of risk consumption. This is the amount adjusted in the direction of increasing recovery. Indicates the variable The portion less than 0 is set to 0, and the portion greater than 1 is set to 1. For example, the weights of power deviation, charge deviation, and prediction deviation can be set to 0.4, 0.3, and 0.1 respectively, with a preset risk consumption coefficient. It can be set to 0.05, the safety recovery coefficient. It can be set to 0.01. Of course, in addition to the three deviations mentioned above, for example, a normalized term of the current operating loss can be further introduced as a fourth term of the multi-source risk cost, with its weight... It can be set to 0.2 to enable the control state variable to make additional responses to the current operating losses; this application does not impose any limitations on this. When determining whether the power deviation is lower than the preset deviation threshold, the deviation assessment cost before normalization can be directly compared with the preset deviation threshold, or the dimensionless value after normalization can be compared; the same applies when determining whether the prediction deviation is lower than the preset prediction error threshold; as long as the above judgment results are consistent with the conditions in the corresponding steps, this application does not impose any limitations on this.

[0037] As can be seen from the above, this embodiment solves the problem of inconsistent dimensions of the three deviations by normalizing the power deviation, charge deviation, and prediction deviation respectively and then performing a weighted summation operation according to their weights to obtain the multi-source risk cost. By determining the risk consumption amount based on the multi-source risk cost and the preset risk consumption coefficient, the reduction of the control state quantity changes synchronously with the multi-source risk cost. By determining the recovery amount according to the upper limit difference value when all three deviations are below their respective thresholds and the charge state is within the preset safe range, and adjusting the control state quantity in the increasing direction, the control state quantity recovers its value after continuous operation. By adjusting the control state quantity in the decreasing direction according to the risk consumption amount when any deviation exceeds the allowable range, the control state quantity contracts when the deviation risk increases. Thus, the control state quantity can achieve a dynamic balance between deviation risk and recovery.

[0038] In the above embodiments, there is no limitation on how to determine the preset range of the control state quantity's value range. To avoid the control state quantity from crossing the boundary during bidirectional adjustment or continuing to accumulate at the boundary, resulting in abnormal values, this embodiment also provides an exemplary method for determining the preset range, which may include the following steps: Obtain the preset range of the control state quantity; set the initial value of the control state quantity within the preset range, and determine the maximum and minimum values ​​of the control state quantity; when performing bidirectional adjustment, stop the value reduction operation when the control state quantity decreases to the minimum value; stop the value increase operation when the control state quantity increases to the maximum value, so that the value range of the control state quantity is limited to between the minimum and maximum values.

[0039] The preset range interval is a closed interval formed by the upper and lower bounds of the pre-defined control state quantity. The maximum value of the control state quantity is the upper bound of the preset range interval. The minimum value of the control state quantity is the lower bound of the preset range interval. The initial value is the value of the control state quantity at the beginning of this round of grid connection control process.

[0040] In this embodiment, the preset range interval of the control state quantity is set to That is, the control state quantity satisfies The minimum value is The maximum value is 1; and the initial value at the start of this round of grid connection regulation is obtained by sampling according to a uniform distribution, and the sampling method satisfies ,in, In the formula, As the initial value, Representing an interval Uniform distribution on and These are the lower and upper limits of the initial value, respectively, with symbols... Indicates sampling according to distribution. Define the relationship between the upper and lower limits; when performing bidirectional adjustment, when the control state quantity decreases to the minimum value... The operation of lowering the value is then stopped, and the operation of raising the value is stopped when the control state quantity increases to the maximum value of 1. The initial value can also be set by a fixed value or by random sampling within a preset range, etc., and this application does not limit it in any way.

[0041] As can be seen from the above, this embodiment sets the initial value of the control state quantity within a preset range and determines the maximum and minimum values ​​of the control state quantity, so that the control state quantity always changes within the boundary; by truncating the values ​​that cross the boundary during the bidirectional adjustment process, the value range of the control state quantity is limited to between the minimum and maximum values, thus avoiding abnormal values ​​caused by the continued accumulation of the control state quantity at the boundary.

[0042] In the above embodiments, there is no limitation on how to determine the grid-connected power plan. To avoid the problem that the relationship between the grid-connected power plan and the control status quantity is unclear, thus preventing the adjustment of the control status quantity from being transmitted to the grid-connected power plan, this embodiment also provides an exemplary method for determining the grid-connected power plan, which may include the following steps: Based on the positive correlation between the planned grid-connected power and the values ​​of the control status variables, the planned grid-connected power is determined according to the predicted values ​​of wind power output, expected grid-connected power, current state of charge, and the values ​​of the control status variables.

[0043] The positive correlation refers to the monotonically increasing value of the grid-connected power plan as the value of the control state variable increases. In this embodiment, the predicted values ​​of wind power output forecast data, expected grid-connected power data, current state of charge, and control state variable are used as inputs. The grid-connected power plan is obtained based on the positive correlation between the grid-connected power plan and the control state variable. For example, when the control state variable is near its maximum value, the grid-connected power plan is the larger proportion of the higher of the predicted values ​​of wind power output forecast data and expected grid-connected power data; when the control state variable is near its minimum value, the grid-connected power plan is the smaller proportion of the lower of the predicted values ​​of wind power output forecast data and expected grid-connected power data; when the control state variable is at an intermediate value, the grid-connected power plan is interpolated using a monotonic function. Exemplarily, the positive correlation can be achieved using various specific forms such as neural networks, preset mapping tables, or multiplication by a scaling factor that monotonically increases with the control state variable. This application does not impose any limitations on these methods.

[0044] As can be seen from the above, this embodiment establishes a positive correlation between the value of the grid-connected power plan and the value of the control state quantity, so that the adjustment of the control state quantity can be directly transmitted to the grid-connected power plan. The grid-connected power plan is determined based on the predicted values ​​of wind power output forecast data, grid-connected power expectation data, current state of charge, and the value of the control state quantity, so that the grid-connected power plan is updated synchronously when the control state quantity changes.

[0045] In the above embodiments, there are no limitations on how to calculate the power deviation between the actual grid-connected power and the planned grid-connected power, the charge deviation of the energy storage device from the preset safe range, and the prediction deviation of the wind power output prediction data relative to the actual wind power output data. To avoid the problems of inconsistent dimensions of the three deviations and double calculation of prediction deviation and power deviation, this embodiment also provides an exemplary method for calculating the three deviations, which may include the following steps: Calculate the absolute value of the power difference between the actual grid-connected power and the planned grid-connected power, and obtain the normalized value of the absolute power as the power deviation; obtain the lower deviation distance when the state of charge is below the lower boundary of the preset safety range, and obtain the upper deviation distance when the state of charge is above the upper boundary of the preset safety range, and determine the charge deviation based on the lower and upper deviation distances; calculate the predicted absolute value of the difference between the actual wind power output data and the predicted wind power output data, and obtain the normalized value of the predicted absolute value as the prediction deviation.

[0046] Among them, the absolute power quantity is the absolute value of the difference between the actual grid-connected power and the planned grid-connected power. The preset reference quantity is a reference quantity used to convert the absolute power quantity into a dimensionless value. The preset safety range is a closed interval formed by artificially set upper and lower boundaries of the safe state of charge; the lower boundary of the preset safety range is the minimum value, and the upper boundary is the maximum value. The lower deviation distance is the difference between the state of charge and the lower boundary of the preset safety range. The upper deviation distance is the difference between the state of charge and the upper boundary of the preset safety range. The predicted absolute quantity is the absolute value of the difference between the actual wind power output data and the predicted wind power output data.

[0047] In this embodiment, the difference between the actual grid-connected power and the planned grid-connected power is first calculated, which can be expressed by a relational formula. To calculate, where, For this difference, This represents the actual grid-connected power. This serves as a reference for calculating power deviation, and in practical applications, it can be represented by a power dispatch command; the absolute value of this difference is then taken. As an absolute quantity of power, it is mapped to deviation assessment fees according to the tiered pricing rules for positive and negative deviations in the target market, and then based on the relationship... The power deviation is obtained by normalization, where The power deviation after normalization. For deviation assessment fees, This serves as a reference value for deviation assessment costs used for normalization. The lower deviation distance when the state of charge is below the lower boundary of the preset safety range and the upper deviation distance when the state of charge is above the upper boundary of the preset safety range are calculated. The upper and lower deviation distances are then summed to obtain the charge deviation. For example, this can be achieved through a relational formula. The charge deviation is calculated, where For charge deviation, The state of charge after the action is performed. and These are the lower and upper boundaries of the preset safe zone, respectively. Since the state of charge cannot be simultaneously below the lower boundary and above the upper boundary, at most one of the two distances is non-zero. Therefore, taking the larger of the two distances is equivalent to adding them together. When the state of charge is within the preset safe zone... Calculate the absolute value of the difference between the actual wind power output data and the predicted wind power output data. As the absolute quantity for prediction, and then based on the relationship... After normalization, the prediction bias is obtained, where The prediction bias after normalization. This refers to the rated power of the wind farm. For example, the reference value for deviation assessment costs. It can be set to 500, and the preset safety range can be set to... Rated power of wind farm The capacity can be set to 50 MW; the tiered pricing ratio can be set to 0.25, 0.60, and 1.00, with the negative deviation ratio being higher than the positive deviation ratio. Furthermore, the reference quantity used to calculate the power deviation can be either the planned grid-connected power or the actual value of the expected grid-connected power data, each corresponding to a different method of obtaining the settlement reference power; as long as the absolute value of the deviation can be converted into deviation assessment fees, this application does not impose any restrictions on this.

[0048] As can be seen from the above, this embodiment obtains the power deviation by normalizing the absolute value of the difference between the actual grid-connected power and the planned grid-connected power, making the power deviation a dimensionless deviation. By taking the larger of the lower deviation distance when the state of charge is below the lower boundary of the preset safety range and the upper deviation distance when the state of charge is above the upper boundary of the preset safety range as the load deviation, the problem of repeated calculation when the state of charge crosses both boundaries is avoided. By normalizing the absolute value of the difference between the actual wind power output data and the predicted wind power output data, the prediction deviation is obtained, making the dimensions of the prediction deviation consistent with those of the power deviation.

[0049] In the above embodiments, there is no limitation on how to calculate the power deviation between the actual grid-connected power and the planned grid-connected power. To avoid the problem of the power deviation being overly sensitively included in the control state quantity when the wind-storage system is operating smoothly, this embodiment also provides an exemplary method for refining the power deviation, which may include the following steps: Obtain the preset blind zone and compare the absolute power value with the preset blind zone; when the absolute power value is not greater than the preset blind zone, set the power deviation to zero; when the absolute power value is greater than the preset blind zone, take the difference between the absolute power value and the preset blind zone as the absolute power value and perform normalization processing to obtain the power deviation.

[0050] The preset blind zone is a manually set allowable deviation range; deviations within the preset blind zone are not included in the power deviation calculation. The width of the preset blind zone is its specific value. In this embodiment, the preset blind zone is read, and its width satisfies... ,in The width of the preset blind spot, To be according to The ratio determines the proportional coefficient of the preset blind zone. To take the absolute value, Minimum blind zone width, This indicates taking the maximum value within the parentheses; when the absolute power quantity... When the power deviation is not greater than the preset blind zone, set it to zero; when the absolute power exceeds the preset blind zone, adjust the power deviation accordingly. Calculate the difference between the absolute power and the preset blind zone, and use this difference as the absolute power. The deviation amount exceeding the preset blind zone. The deviation is represented, and then normalized according to a preset reference value to obtain the power deviation. When When the preset blind zone is not exceeded, this formula takes the value of 0. For example, the proportionality coefficient... It can be set to 5%, minimum blind zone width. It can be set to 0.50 MW, and the larger of the two values ​​is taken as the width of the preset blind zone; as long as the width of the preset blind zone is greater than zero, the power deviation during stable operation can be avoided from being overly sensitively included in the control state quantity. This application does not impose any restrictions on this.

[0051] As can be seen from the above, this embodiment sets a preset blind zone so that the portion of the absolute power within the preset blind zone is not included in the power deviation; by normalizing the portion of the absolute power greater than the preset blind zone to obtain the power deviation, the power deviation can distinguish whether the deviation is too large, thereby avoiding the power deviation being overly sensitively included in the control state quantity when the wind-storage system is operating smoothly.

[0052] In the above embodiments, there are no limitations on how to determine the wind power component and the energy storage power component based on the actual wind power output data and the grid-connected power plan, respectively. To avoid the problem that the wind power component or the energy storage power component exceeds the physical boundary, or that the difference between the actual wind power output data and the grid-connected power plan is not absorbed by wind power and energy storage, this embodiment also provides an exemplary power component allocation method, which may include the following steps: The wind power component is limited to a value not less than zero and not greater than the generating power corresponding to the actual wind power output data; the charging power of the energy storage power component is limited to an absolute value not exceeding the preset maximum charging power and a discharging power not exceeding the preset maximum discharging power; the state of charge of the energy storage device is limited to a value not lower than the preset lower charge limit and not higher than the preset upper charge limit; the actual wind power output data is compared with the grid-connected power plan; when the actual wind power output data is greater than the grid-connected power plan, the excess power is absorbed through the charging power in the energy storage power component; when the actual wind power output data is less than the grid-connected power plan, the shortfall is made up through the discharging power in the energy storage power component.

[0053] Among them, the generating power is the maximum active power that the wind farm can generate in the current time period; the preset maximum charging power is the absolute value of the maximum power allowed in the charging direction of the energy storage device; the preset maximum discharging power is the maximum power allowed in the discharging direction of the energy storage device; the preset lower charge limit is the upper limit of the state of charge that the energy storage device is allowed to operate in; and the preset upper charge limit is the upper limit of the state of charge that the energy storage device is allowed to operate in.

[0054] In this embodiment, it can be constrained , Limit the wind power component, energy storage power component, and the state of charge of energy storage devices; This is the preset lower limit of charge for energy storage devices. This sets a preset upper limit for the charge capacity of the energy storage device. Then, it compares the actual wind power output with the planned grid-connected power: when the actual wind power output exceeds the planned grid-connected power, the excess power is absorbed by the charging power in the energy storage power component, and the charging power does not exceed the difference between the actual wind power output and the planned grid-connected power; when the actual wind power output is less than the planned grid-connected power, the shortfall is made up by the discharging power in the energy storage power component, and the discharging power does not exceed the difference between the planned grid-connected power and the actual wind power output. For example, if the energy storage device is only allowed to be charged by wind power for external transmission wind storage systems, it also requires... The charging power is limited to the amount of wind power available for charging based on the power flow structure. For example, the wind power component and the energy storage power component can be obtained directly by rule-based calculation, or the wind power component and the energy storage power component can be output by a neural network; as long as the wind power component and the energy storage power component meet the above constraints, this application does not impose any limitations on them.

[0055] As can be seen from the above, this embodiment ensures that the wind power component and the energy storage power component always meet physical constraints by limiting the wind power component to between zero and the generating power, limiting the charging and discharging power of the energy storage component to between the preset maximum charging power and the preset maximum discharging power, and limiting the state of charge of the energy storage device to between the preset lower charge limit and the preset upper charge limit. By comparing the actual wind power output data with the grid-connected power plan and absorbing or supplementing the excess or deficiency through the energy storage charging and discharging power, the wind-storage system can balance the actual wind power output data and the grid-connected power plan in real time.

[0056] In the above embodiments, there is no limitation on how to determine the grid-connected power plan. To avoid the problem that the grid-connected power plan cannot reflect the tail information of the reward value, this embodiment also provides an exemplary method for determining the grid-connected power plan based on the reward value distribution, which may include the following steps: A critic network is constructed to output multiple quantile values, where each quantile value represents the reward distribution corresponding to different grid-connected power plans executed under a given state. The grid-connected regulation process is executed multiple times, and the current state and the current grid-connected power plan input to the critic network are recorded after each execution. The parameters of the critic network are adjusted based on the error between the quantile values ​​input to the critic network and the actual reward values ​​obtained after execution. The average of the multiple quantile values ​​output by the trained critic network is taken as the expected feature, and the average of the portion of the multiple quantile values ​​before a specified proportion after ascending order is taken as the lower tail feature. The regulation state quantity is used as the weight of the expected feature, and the complement of the regulation state quantity is used as the weight of the lower tail feature. A weighted summation operation is performed on the expected feature and the lower tail feature to obtain the adjustment target. The grid-connected power plan is selected according to the adjustment target.

[0057] The commentator network is an artificial neural network used to estimate the reward value distribution. Quantile values ​​are the values ​​at certain quantile points in the reward value distribution, used to approximate the complete shape of the reward value distribution. The given state is the input to the commentator network, consisting of predicted wind power output data, predicted values ​​of grid-connected power expectation data, the current state of charge of energy storage devices, and regulation state variables. The expected feature is the mean of the reward value distribution. The lower tail feature is the conditional expectation of the lower tail portion of the reward value distribution. The specified proportion is a pre-set parameter that determines the proportion of the lower tail feature. The complement of the regulation state variable is the difference between the maximum value and the current value of the regulation state variable. The adjustment target is a comprehensive index obtained by weighted summation of the expected feature and the lower tail feature. The larger the value of the regulation state variable, the greater the proportion of the expected feature in the adjustment target, and the closer the selected grid-connected power plan is to the plan corresponding to maximizing the expected feature; the smaller the value of the regulation state variable, the greater the proportion of the lower tail feature in the adjustment target, and the closer the selected grid-connected power plan is to the plan corresponding to maximizing the lower tail feature.

[0058] In this embodiment, the constructed critic network is based on The quantile values ​​approximate the reward value distribution. The reward value distribution satisfies... ,in This is the random discount bonus value obtained after executing the grid-connected power plan under a given condition. Given a state, This is the planned grid-connected power output. The method for determining the grid-connected power plan. This represents the offset round number from the current round. As a discount factor, For the first Reward value for a given time period This represents the summation of reward values ​​over subsequent rounds. This indicates that the random variables on both sides are identically distributed. Represents an infinite time domain. Multiple quantile values ​​approximate the distribution of reward values. .in, For parameters The distribution of controlled reward values ​​is approximately... For the first There are n predicted quantile values, where i is from 1 to N. This represents the total number of quantile values. This represents the set consisting of the predicted quantile values. This is an approximation. The quantile level corresponding to each quantile value is... ,in For the first A quantile level of 0.5 is used to ensure that the quantile is located in the middle of each equal division interval. This is achieved using the target commentator network parameters. Construct target quantile samples ,in For the first Target quantile samples, with tilde indicating the target quantity. This is the reward value for the current round. The output of the target critic network quantile value. Given the state for the next round, This is the planned grid-connected power output for the next round. The commentator network minimizes the pairwise quantile Huber loss. Training was conducted, among which For the loss function of the commentator network, the subscript is... This refers to the network of critics. For online critic parameters, This represents the total number of pairs of online quantile values ​​with the target quantile value. quantile level and Huber threshold The corresponding quantile Huber loss, The error between the target quantile value and the online predicted quantile value. The quantile Huber loss can be defined as... ,in For error The corresponding quantile Huber loss, For quantile levels, For Huber threshold, This is an indicator function that takes a value of 1 when the error is less than 0 and a value of 0 otherwise. Represents absolute value. For Huber's loss; when hour, ,otherwise The expected feature estimate is: ,in For the desired features, Represents the mathematical expectation. The reward distribution estimated for the commentator network. The lower tail feature is estimated as follows: ,in This is a characteristic of the lower tail. Represents conditional risk value. For a specified ratio, , This refers to the number of predicted quantile values ​​whose quantile level is no greater than a specified proportion. The control state variable is used as the weight of the desired feature and as a complement to the control state variable. The weights of the lower tail features are used for weighted summation to obtain the adjustment target. ,in To adjust the target, the planned grid-connected power output can be determined by minimizing the target. ,in The loss function for this determination method, subscript This indicates the first control phase. For parameters to be updated, For experience replay pool, This indicates that a given state is sampled from the experience replay pool. Indicates by parameter strategy The sampling action, which is the grid-connected power plan, The temperature parameter is the entropy normal. Given a state The logarithmic probability of selecting this grid-connected power plan quantity. This represents the expectation of the sampled distribution of a given state and the planned grid-connected power. Then, the planned grid-connected power is selected based on the adjustment target: when the control state variable takes a larger value within a preset range, the adjustment target is mainly determined by the expected characteristics, and the selected planned grid-connected power approaches the larger proportion of the higher of the predicted values ​​of wind power output and the expected grid-connected power; when the control state variable takes a smaller value within a preset range, the adjustment target is mainly determined by the lower tail characteristics, and the selected planned grid-connected power approaches the smaller proportion of the lower tail. For example, the total number of quantile values... It can be set to 32 to specify the ratio. It can be set to 0.1; as long as the total number of quantile values ​​is greater than zero and the specified proportion is met. This application does not impose any restrictions on this matter.

[0059] As can be seen from the above, this embodiment constructs a commentator network for outputting multiple quantile values, enabling the commentator network to estimate the complete distribution of reward values ​​rather than being limited to the expected value; by adjusting the parameters of the commentator network according to the error between the quantile values ​​corresponding to the input of the commentator network and the actual reward value obtained after execution, the output of the commentator network is made consistent with the distribution of reward values; by taking the average of multiple quantile values ​​as the expected feature and taking the average of the first specified proportion as the lower tail feature, the expected information and lower tail risk information of the reward value are explicitly extracted; by using the control state quantity as the weight of the expected feature and the complement of the control state quantity as the weight of the lower tail feature, a weighted summation operation is performed on the expected feature and the lower tail feature to obtain the adjustment target, so that the adjustment target can dynamically balance the pursuit of benefits and risk control according to the deviation tolerance of the wind and energy storage system; by selecting the grid-connected power plan according to the adjustment target, the grid-connected power plan can reflect the complete distribution information of reward values.

[0060] In the above embodiments, there is no limitation on how to determine the wind power component and the energy storage power component separately. To avoid the problem of performing suboptimal actions such as energy storage charging or wind curtailment when the wind power component crosses the physical boundary or when there is a power deficit, this embodiment also provides an exemplary power component allocation method based on reinforcement learning, which may include the following steps: A critic network and an actor network are constructed. The actor network outputs the probability distribution parameters of wind power and energy storage power components, while the critic network evaluates the value of the actions output by the actor network. The grid-connected control process is executed multiple times, and the reward value obtained after each execution is recorded. The parameters of both the critic network and the actor network are updated based on the reward value to ensure that the actor network's output approaches the action that maximizes the reward value. During training, if the wind power component exceeds the available power corresponding to the actual wind power output data, the corresponding reward value is reduced. If the energy storage device is charged when the actual grid-connected power is less than the planned grid-connected power, the corresponding reward value is reduced. If the wind power component is lower than the available power corresponding to the actual wind power output data when the actual grid-connected power is less than the planned grid-connected power, the corresponding reward value is reduced.

[0061] The actor network is an artificial neural network used to output the probability distribution parameters of the wind power component and the energy storage power component. The critic network is an artificial neural network used to evaluate the value of state-action pairs. Value is the expected long-term reward of the state-action pair.

[0062] In this embodiment, the Soft Actor-Critic (SAC) algorithm is used to train the allocation of wind power components and energy storage power components. An actor network is constructed to output the probability distribution parameters of the wind power components and energy storage power components, which are then sampled and mapped to the corresponding physical range through action scaling. A critic network is constructed to evaluate the value corresponding to a given state and action. The target network remains stable through soft updates, and experience data is stored in an experience replay pool. The actor network balances exploration and reward under the action of entropy regularization, while the target of the critic network consists of the current reward value and the value of the next state. The economic reward value in the reward value satisfies... ,in As economic reward value, This indicates an economic reward. Settlement of electricity consumption discrepancies between real-time market and day-ahead clearing. For deviation assessment fees, For energy storage operation or degradation costs, To cover the cost of wind curtailment, the above-mentioned , , and These represent electrical energy deviation, energy storage, and wind curtailment, respectively; among them, real-time market electricity sales revenue satisfies , The actual value of the settlement price for grid-connected electricity. This represents the actual grid-connected power. For the clearing power of the day before, To adjust the length of the time period, superscript This indicates that the reward value for the second control phase meets the requirements of [previous date]. ,in This is the reward value for the second control phase. This indicates the second control phase. As economic reward value, For physical constraint violation quantity, The penalty for suboptimal behavior may include continuing to perform energy storage charging when the actual grid-connected power is lower than the settlement reference power, or continuing to perform wind curtailment when the available wind power can reduce the deficit. and These are physical constraint shaping weights and suboptimal behavior shaping weights. Physical constraint violations include wind power component violations, energy storage power component violations, state-of-charge boundary violations, and grid connection constraint violations prohibiting grid-to-energy storage charging. Suboptimal behavior penalties include controlling energy storage charging even when actual grid-connected power is lower than planned grid-connected power, or implementing wind curtailment even when actual wind power output data can reduce the deficit. These shaping terms are used to accelerate training, and the physical constraint projection ensures the final executableness of actions when put into actual use. For out-of-bounds actions generated in the early stages of training, the degree of out-of-bounds violation is calculated and added to the physical constraint violation, or the action is projected to the feasible region before being put into actual use. During training, if the wind power component exceeds the available power, or if energy storage charging is controlled when the actual grid-connected power is lower than the planned grid-connected power, or if the wind power component is lower than the available power when the actual grid-connected power is lower than the planned grid-connected power, the corresponding reward value is reduced. For example, the allocation of wind power components and energy storage power components can be trained first under a fixed rule for determining the grid-connected power plan, and then the allocation method after the training is completed can be frozen to train the method for determining the grid-connected power plan, and joint fine-tuning can be performed as needed; other training sequences that can realize the state transition and reward feedback relationship between the first control stage and the second control stage can also be adopted; the training sequence is only used to obtain executable parameters and does not constitute a necessary limitation of this application.

[0063] As can be seen from the above, this embodiment constructs a critic network and an actor network and updates the parameters of the critic network and the actor network simultaneously with the reward value, so that the output of the actor network approaches the action that maximizes the reward value; by reducing and correcting the reward value in three cases, namely when the wind power component exceeds the available power, when the actual grid-connected power is less than the planned grid-connected power, and when the actual grid-connected power is less than the planned grid-connected power, the actor network avoids the above suboptimal actions during training.

[0064] It should be noted that there is no strict order of execution for the steps in this application. As long as they conform to a logical order, these steps can be executed simultaneously or in a certain preset order. Figure 1 This is just an illustrative example and does not mean that this is the only possible execution order.

[0065] This application also provides a corresponding device for the adaptive regulation of grid-connected power in a wind-storage system, further enhancing the practicality of the method. The device can be described from both a functional module perspective and a hardware perspective. The following describes the adaptive regulation device for grid-connected power in a wind-storage system provided in this application. This device is used to implement the adaptive regulation method for grid-connected power in a wind-storage system provided in this application. In this embodiment, the adaptive regulation device for grid-connected power in a wind-storage system may include or be divided into one or more program modules. These program modules are stored in a storage medium and executed by one or more processors to complete the adaptive regulation method for grid-connected power in a wind-storage system disclosed in the embodiment. The program module referred to in this embodiment is a series of computer program instruction segments capable of performing specific functions, which is more suitable than the program itself for describing the execution process of the adaptive regulation device for grid-connected power in a wind-storage system in the storage medium. The following description will specifically introduce the functions of each program module in this embodiment. The adaptive regulation device for grid-connected power in a wind-storage system described below can be referred to in correspondence with the adaptive regulation method for grid-connected power in a wind-storage system described above.

[0066] From the perspective of functional modules, see Figure 2 , Figure 2 This is a structural diagram of the grid-connected power adaptive control device for the wind-storage system provided in this embodiment, in one specific implementation. The device may include: The phase division module 201 is used to divide the grid connection regulation process of the wind-storage system into a first control phase and a second control phase based on the acquisition time of the actual wind power output data and the expected grid-connected power data. The planned quantity determination module 202 is used to determine the planned grid-connected power quantity in the first control stage based on the predicted values ​​of wind power output forecast data, grid-connected power expectation data, the current state of charge of energy storage equipment, and the control state quantity; the value range of the control state quantity is within a preset range. The power component determination module 203 is used to determine the wind power component and the energy storage power component respectively in the second control stage based on the actual value of the wind power output data, the actual value of the grid-connected power expectation data, and the grid-connected power plan. The actual quantity determination module 204 is used to determine the actual grid-connected power of the wind power and energy storage system based on the wind power component and the energy storage power component. The state quantity adjustment module 205 is used to adjust and regulate the state quantity in both directions based on the power deviation between the actual grid-connected power and the planned grid-connected power, the charge deviation of the energy storage device from the preset safe range, and the prediction deviation of the wind power output prediction data relative to the actual wind power output data. The state quantity transmission module 206 is used to apply the adjusted control state quantity to the first control stage of the next round of grid connection control process.

[0067] The grid-connected power adaptive control device for the wind-storage system mentioned above is described from the perspective of functional modules. Furthermore, this application also provides an electronic device, which is described from the perspective of hardware. Figure 3 This is a schematic diagram of the electronic device provided in one embodiment of the present application. The electronic device includes a memory 301 and a processor 302. The memory 301 stores a computer program, and the processor 302 is configured to run the computer program to execute the steps in any of the above-described embodiments of the grid-connected power adaptive control method for wind and energy storage systems.

[0068] It is understood that if the grid-connected power adaptive control method of the wind-storage system in the above embodiments is implemented as a software functional unit and sold or used as an independent product, it can be stored in a non-volatile storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the related technology, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and executes all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes, but is not limited to, various media capable of storing program code, such as: USB flash drive, mobile hard disk, read-only memory (ROM), random access memory (RAM), electrically erasable programmable ROM, register, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, removable disk, CD-ROM, magnetic disk, or optical disk. Based on this, this application also provides a non-volatile storage medium on which a computer program is stored. This computer program is configured to execute the steps in any of the above embodiments of the grid-connected power adaptive control method of the wind-storage system when it runs.

[0069] It is understood that if the adaptive regulation method for grid-connected power of the wind-storage system in the above embodiments is implemented as a software functional unit and sold or used as an independent product, the computer software product may not need to be stored in a physical storage medium. For example, it can be directly transmitted to a computer or other device with information processing capabilities via a wired or wireless network to execute all or part of the steps of the methods in the various embodiments of this application. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the related technology, or all or part of the technical solution, can be embodied in the form of a software product. Based on this, this application also provides a computer program product, which stores a computer program that is configured to execute the steps in any of the embodiments of the adaptive regulation method for grid-connected power of the wind-storage system described above when it is run.

[0070] Finally, this application also provides a wind storage system, see [link to relevant documentation]. Figure 4 The system may include a wind farm 401, an energy storage power station 402, and a server 403. The wind farm 401 includes a wind farm power controller for controlling the start-up, shutdown, and power output of wind turbine generators. The energy storage power station 402 includes an energy storage converter for controlling the charging and discharging power of energy storage devices. The server 403 includes at least a processor, a memory, and a communication interface. The processor acquires grid-connected power expectation data through the communication interface and implements the steps of the adaptive grid-connected power control method for the wind-storage system described in any of the above embodiments by executing a computer program stored in the memory. It generates wind power dispatch power commands based on wind power components and energy storage charging and discharging power commands based on energy storage power components. The processor sends the wind power dispatch power commands to the wind farm power controller and the energy storage charging and discharging power commands to the energy storage converter through the communication interface. The wind farm power controller controls the power output of the wind turbine generators according to the wind power dispatch power commands. The energy storage converter controls the charging and discharging power of the energy storage devices according to the energy storage charging and discharging power commands.

[0071] The processor is a central processing unit, application-specific integrated circuit, or field-programmable gate array (FPGA) with information processing capabilities. The memory is a non-volatile storage medium used to store computer programs and data. The communication interface is a wired or wireless communication module that supports data exchange with wind farms and energy storage power stations.

[0072] In this embodiment, when the processor executes the computer program stored in the memory, it completes the time allocation for acquiring actual wind power output data and expected grid-connected power data according to the method corresponding to the above embodiments; determines the grid-connected power plan in the first control stage; allocates wind power components and energy storage power components in the second control stage; performs bidirectional adjustment of power deviation, charge deviation, and prediction deviation; and transmits the control state variables to the next round of grid-connected control process. It also generates wind power dispatching power commands based on the wind power components and energy storage charging and discharging power commands based on the energy storage power components, and sends them to the wind farm power controller and energy storage converter respectively through the communication interface to realize the actual control of the grid-connected power of the wind-storage system. Exemplarily, an industrial control computer, a programmable logic controller, or an embedded controller can be used as the processor; as long as the processor has information processing capabilities and can exchange data with the wind farm power controller and energy storage converter, this application does not impose any limitations on this.

[0073] During the rolling operation, server 403 executes the following process in a round-by-round cycle: Before the start of the first control phase of each round of grid connection regulation, server 403 collects wind power output prediction data, predicted values ​​of grid-connected power expectation data, the current state of charge of energy storage devices, historical wind power dispatch results, and historical grid connection deviation information, updates the input data of the first control phase, and generates the grid-connected power plan, which is then sent to the dispatching agency via the grid-connected power application interface; after the actual value of the grid-connected power expectation data and the actual wind power output data are available, the server constructs the input data for the second control phase, generates wind power components and energy storage power components, and then the safety filter, based on the upper limit of the generateable power, the preset maximum charging power, the preset maximum discharging power, the preset lower limit of charge, and the preset upper limit of charge, performs the following process: In addition to grid connection constraints prohibiting grid-to-energy storage charging, the two power components mentioned above are clipped or projected, and wind power dispatch power commands and energy storage charging / discharging power commands are generated according to the clipped or projected results, and sent to the wind farm power controller and energy storage converter. After the end of this round of grid connection regulation, the server obtains the actual grid-connected power, grid-connected power settlement results, the state of charge of the energy storage device, and the actual wind power output data, calculates the economic reward value and power deviation, load deviation, and prediction deviation, and adjusts the regulation state variables bidirectionally accordingly. The adjusted regulation state variables, together with the new wind power output prediction data, the predicted value of the grid-connected power expectation data, and the current state of charge of the energy storage device, constitute the input data for the first control stage of the next round of grid connection regulation, thus continuously rolling out the process. In addition, the actor network used to determine the grid-connected power plan, the actor network used to determine the wind power component and the energy storage power component, the normalization parameters of the input data, and the adjustment parameters of the regulation state variables are all deployed on the server.

[0074] As can be seen from the above, the wind-storage system of this embodiment implements the method steps of the above embodiment by executing the computer program stored in the memory through the processor, so that the grid-connected power plan can be dynamically adjusted according to the current deviation tolerance of the wind-storage system and the wind power component and the energy storage power component meet the physical constraints; the wind farm power controller and the energy storage converter perform actual regulation according to the wind power dispatch power command and the energy storage charging and discharging power command, so that the wind-storage system can balance the actual wind power output data and the grid-connected power plan in real time.

[0075] Any of the components, modules, units, parts, methods, and operations described herein can be implemented using software, firmware, hardware (e.g., fixed logic circuitry), manual processing, or any combination thereof. Alternatively or additionally, any functionality described herein can be performed at least in part by one or more hardware logic components, such as, but not limited to, a central processing unit (CPU), a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), an application-specific standard product (ASSP), a system-on-a-chip (SoC), a complex programmable logic device (CPLD), a microprocessor (MCU), etc. The systems, computing devices, or apparatuses described herein encompass a wide range of means, devices, and machines for processing data, including, for example, one or more programmable processors, computers, SoCs, or combinations thereof. The apparatus may also include code that creates an execution environment for the computer program in question, such as code constituting processor firmware, a protocol stack, a database management system, an operating system, a cross-platform runtime environment, a virtual machine, or one or more combinations thereof. The aforementioned computer program (also known as a program, software, software application, app, script, or code) can be written in any form of programming language, including compiled or interpreted languages, declarative or procedural languages, and can be deployed in any form, including as a standalone program or as a module, component, subroutine, object, or other unit suitable for a computing environment.

[0076] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this specification can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0077] The foregoing has provided a detailed description of a wind-storage system and its grid-connected power adaptive regulation method provided in this application. The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. Several improvements and modifications can be made to this application without departing from the principles thereof, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for adaptive regulation of grid-connected power in a wind-storage system, characterized in that, include: Based on the acquisition time of actual wind power output data and expected grid-connected power data, the grid-connected control process of the wind-storage system is divided into the first control stage and the second control stage. In the first control phase, the planned grid-connected power is determined based on the predicted wind power output data, the predicted values ​​of the expected grid-connected power data, the current state of charge of the energy storage device, and the control state variables; the values ​​of the control state variables are within a preset range. In the second control phase, based on the actual output data of wind power, the actual value of the expected grid-connected power data, and the planned grid-connected power, the wind power component and the energy storage power component are determined respectively, and based on the wind power component and the energy storage power component, the actual grid-connected power of the wind-storage system is determined. The control state quantity is adjusted bidirectionally based on the power deviation between the actual grid-connected power and the planned grid-connected power, the charge deviation of the energy storage device from the preset safe range, and the prediction deviation of the wind power output prediction data relative to the actual wind power output data. The adjusted control state quantity will be used in the first control stage of the next round of grid connection control process.

2. The adaptive regulation method for grid-connected power of a wind-storage system according to claim 1, characterized in that, Bidirectional adjustment of the control state quantity includes: The power deviation, the charge deviation, and the prediction deviation are normalized respectively, and the normalized power deviation, charge deviation, and prediction deviation are weighted and summed according to their respective weights to obtain the multi-source risk cost. The amount of risk consumption is determined based on the multi-source risk cost and the preset risk consumption coefficient; If the power deviation is lower than a preset deviation threshold, the prediction deviation is lower than a preset prediction error threshold, and the state of charge is within the preset safe range, then a recovery amount is determined based on the upper limit difference between the current value of the control state quantity and its upper limit value. The value of the control state quantity is adjusted in the increasing direction according to the recovery amount, and at the same time, the value of the control state quantity is adjusted in the decreasing direction according to the risk consumption amount. The recovery amount increases as the upper limit difference value increases, and the decrease in the control state quantity increases as the risk consumption amount increases. If the power deviation is not lower than the preset deviation threshold, or the prediction deviation is not lower than the preset prediction error threshold, or the state of charge is not within the preset safe range, then the value of the control state quantity is adjusted in the decreasing direction according to the risk consumption amount, and the decrease in the control state quantity increases with the increase of the risk consumption amount.

3. The adaptive regulation method for grid-connected power of a wind-storage system according to claim 2, characterized in that, The process of determining the range of values ​​for the control state quantity includes: Obtain the preset range interval of the control state quantity; Set the initial value of the control state quantity within the preset range, and determine the maximum and minimum values ​​of the control state quantity; When performing the bidirectional adjustment, the value reduction operation is stopped when the control state quantity decreases to the minimum value; the value increase operation is stopped when the control state quantity increases to the maximum value, so that the value range of the control state quantity is limited to between the minimum value and the maximum value.

4. The adaptive regulation method for grid-connected power of a wind-storage system according to claim 3, characterized in that, Determine the planned grid-connected power output, including: Based on the positive correlation between the planned grid-connected power and the value of the control state quantity, the value of the planned grid-connected power is determined according to the predicted values ​​of wind power output data, expected grid-connected power data, current state of charge, and the value of the control state quantity.

5. The adaptive regulation method for grid-connected power of a wind-storage system according to claim 1, characterized in that, Before bidirectionally adjusting the control state quantity, the following is also included: Calculate the absolute value of the power difference between the actual grid-connected power and the planned grid-connected power, and obtain the normalized value of the absolute power as the power deviation; Obtain the lower deviation distance when the state of charge is below the lower boundary of the preset safety range, obtain the upper deviation distance when the state of charge is above the upper boundary of the preset safety range, and determine the charge deviation based on the lower deviation distance and the upper deviation distance; Calculate the absolute value of the difference between the actual wind power output data and the predicted wind power output data, and obtain the normalized value of the absolute value of the prediction as the prediction deviation.

6. The adaptive regulation method for grid-connected power of a wind-storage system according to claim 5, characterized in that, The process of obtaining the power deviation includes: Obtain a preset blind zone and compare the absolute power quantity with the preset blind zone; When the absolute power is not greater than the preset blind zone, the power deviation is set to zero; When the absolute power is greater than the preset blind zone, the difference between the absolute power and the preset blind zone is taken as the absolute power and normalized to obtain the power deviation.

7. The adaptive regulation method for grid-connected power of a wind-storage system according to any one of claims 1 to 6, characterized in that, The wind power component and the energy storage power component are determined separately, including: The wind power component is limited to a value that is not less than zero and not greater than the generating power corresponding to the actual wind power output data. The charging power of the energy storage power component is limited to an absolute value not exceeding a preset maximum charging power and a discharging power not exceeding a preset maximum discharging power. The state of charge of the energy storage device is restricted to be no lower than a preset lower charge limit and no higher than a preset upper charge limit; Compare the actual wind power output data with the planned grid-connected power; When the actual wind power output data is greater than the grid-connected power plan, the excess power will be absorbed by the charging power in the energy storage power component. When the actual wind power output is less than the grid-connected power plan, the shortfall will be made up by the discharge power in the energy storage power component.

8. The adaptive regulation method for grid-connected power of a wind-storage system according to any one of claims 1 to 6, characterized in that, Determine the planned grid-connected power output, including: Construct a commentator network to output multiple quantile values, where each quantile value represents the distribution of reward values ​​corresponding to different grid-connected power plans executed under a given state; The grid connection control process is executed multiple times, and the current state and the current grid connection power plan input to the commentator network are recorded after each execution. The parameters of the commentator network are adjusted according to the error between the quantile value input by the commentator network and the actual reward value obtained after execution. The average of multiple quantile values ​​output by the critic network after training is taken as the expected feature, and the average of the portion of the multiple quantile values ​​that is before a specified proportion after being arranged from smallest to largest is taken as the lower tail feature. The control state quantity is used as the weight of the desired feature, and the complement of the control state quantity is used as the weight of the lower tail feature. A weighted summation operation is performed on the desired feature and the lower tail feature to obtain the adjustment target. Select the grid-connected power plan according to the aforementioned adjustment targets.

9. The adaptive regulation method for grid-connected power of a wind-storage system according to any one of claims 1 to 6, characterized in that, The wind power component and the energy storage power component are determined separately, including: Construct a critic network and an actor network, wherein the actor network outputs probability distribution parameters of wind power components and energy storage power components, and the critic network evaluates the value corresponding to the actions output by the actor network; The grid-connected control process is executed multiple times. After each execution, the reward value obtained after the action is recorded. The parameters of the critic network and the actor network are updated simultaneously based on the reward value, so that the output of the actor network approaches the action that maximizes the reward value. During training, if the wind power component exceeds the generating power corresponding to the actual wind power output data, the corresponding reward value is reduced. If the energy storage device is charged when the actual grid-connected power is less than the planned grid-connected power, the corresponding reward value is reduced. If the wind power component is lower than the generating power corresponding to the actual wind power output data when the actual grid-connected power is less than the planned grid-connected power, the corresponding reward value is reduced.

10. A wind-storage system, characterized in that, It includes a wind farm, an energy storage power station, and a server; the wind farm includes a wind farm power controller, and the energy storage power station includes an energy storage converter. The server includes at least a processor, memory, and a communication interface; The processor acquires grid-connected power expectation data through the communication interface and implements the steps of the grid-connected power adaptive control method of the wind-storage system as described in any one of claims 1 to 9 by executing the computer program stored in the memory; generating wind power dispatch power instructions based on wind power components and generating energy storage charging and discharging power instructions based on energy storage power components. The processor sends the wind power dispatch power command to the wind farm power controller and the energy storage charging and discharging power command to the energy storage converter through the communication interface.

Images