Multi-time-scale hybrid energy storage capacity configuration method and system and storage medium

Abstract

The invention relates to the technical field of data mining application, and discloses a multi-time-scale hybrid energy storage capacity configuration method and system and a storage medium. The method comprises the following steps: acquiring data according to a wind power plant construction state and preprocessing to obtain standard wind power output power data; acquiring a local power grid load curve or scheduling demand data, and calculating a difference value to generate an energy storage reference power curve; splitting the power signal by adopting an SGMD decomposition algorithm, and identifying multi-time scale power fluctuation to generate feature data; setting constraints based on the feature data, and optimizing energy storage capacity configuration by using an immune hybrid particle swarm algorithm; and performing full-life-cycle simulation on an optimization result, verifying an effect and compliance, and evaluating economy to determine an optimal scheme. Through multi-time scale decomposition and closed-loop optimization, the problems that existing energy storage configuration is poor in adaptability and insufficient in economical efficiency are solved, and wind power integration stability and configuration economical efficiency are improved.

Description

Technical Field

[0001] This application relates to the field of data mining application technology, and in particular to a method, system and storage medium for configuring multi-timescale hybrid energy storage capacity. Background Technology

[0002] As wind power accounts for an increasing proportion of the energy structure, its output power is affected by natural factors such as wind speed and direction, exhibiting random fluctuations at multiple time scales, ranging from seconds, minutes, hours to weeks and above. If these fluctuations are directly connected to the grid, they can easily lead to an imbalance between grid load and demand, disrupt grid frequency stability, reduce power quality, and even cause equipment failures and grid connection risks. Therefore, it is necessary to configure energy storage systems to smooth out fluctuations and adapt to grid demand. A reasonable energy storage capacity configuration is the core prerequisite for ensuring the efficient operation of energy storage systems.

[0003] Existing energy storage capacity configuration technologies suffer from several shortcomings, making it difficult to meet the high precision and economic requirements of wind power grid connection. On the one hand, traditional methods often analyze power fluctuations based on a single time scale, failing to develop adaptive solutions for the different characteristics of fluctuations at different time scales. For example, they may focus only on minute-level fluctuation smoothing while ignoring second-level high-frequency impacts or week-level low-frequency peak-shaving needs, resulting in configurations that cannot comprehensively address fluctuations across multiple scenarios and impacting grid stability. On the other hand, existing technologies often prioritize meeting grid connection standards as the sole objective, failing to fully incorporate life-cycle economic considerations. This can easily lead to problems such as excessively high initial investment, uncontrolled operation and maintenance costs, or insufficient revenue coverage. Furthermore, the algorithms used in the optimization process often suffer from local optimum traps, failing to balance functionality and economy, and making it difficult to output a globally optimal configuration solution. In addition, some methods lack a precise mapping between fluctuation characteristics and energy storage response, and the consideration of constraints such as energy storage charging and discharging power and state of charge during configuration is insufficient, further reducing the practicality and reliability of the configuration scheme.

[0004] To address the aforementioned shortcomings, this application proposes a technical approach of "feature identification - optimized configuration - simulation verification" by identifying fluctuation characteristics through multi-timescale power decomposition, combining it with an immune hybrid particle swarm optimization algorithm for multi-objective optimization, and verifying it through full lifecycle simulation. This approach aims to achieve energy storage capacity configuration that balances multi-scale fluctuation mitigation, grid connection compliance, and economic efficiency. Summary of the Invention

[0005] This application relates to the field of data mining application technology, and discloses a method, system, and storage medium for configuring hybrid energy storage capacity across multiple time scales. This application addresses the problems of poor adaptability and insufficient economic efficiency in existing energy storage configurations through multi-time scale decomposition and closed-loop optimization, thereby improving the grid connection stability and configuration economy of wind power.

[0006] Firstly, this application provides a method for configuring multi-timescale hybrid energy storage capacity, the method comprising: Step S101: Based on the data collected from the construction status of the wind farm, preprocess the collected data to obtain standard wind power output data; Step S102: Obtain the load curve or grid dispatch demand data of the local power grid, calculate the difference between the standard wind power output power data and the load curve or grid dispatch demand data, and generate the reference power curve of the energy storage system. Step S103: The standard wind power output power data is decomposed into multiple frequency components using the SGMD decomposition algorithm. Based on the frequency components, power fluctuations at different time scales, such as second, minute, hour, day, week and above, are identified to generate multi-time scale fluctuation feature data. Step S104: Based on the multi-timescale fluctuation characteristic data, set energy storage response time constraints, and use the immune hybrid particle swarm algorithm to optimize the capacity configuration of each energy storage device, so as to prioritize meeting the national standards for wind power grid connection and the fluctuation requirements of the reference power curve, and secondly maximize economic benefits, while following the constraints of energy storage charging and discharging power, energy storage capacity, state of charge and response time. Step S105: Perform annual full life cycle simulation on the optimized energy storage capacity configuration to verify its effectiveness in smoothing load fluctuations and meeting peak shaving requirements at different time scales, as well as its compliance with national standards for wind power grid connection. At the same time, evaluate the economics to determine the optimal energy storage capacity configuration scheme.

[0007] Optionally, step S101 includes: Determine the construction status of the wind farm, which includes both completed and planned status; If the wind farm is already built, the actual output power data generated during its operation is directly collected. The actual output power data is then cleaned, smoothed, and outlier-handled to obtain the standard wind power output power data. If the project is in the planning stage, local wind resource data and wind turbine model data from the tender are collected. Based on the local wind resource data and wind turbine model data from the tender, the wind power output power is fitted using the LSTM algorithm to obtain the standard wind power output power data of the proposed wind farm.

[0008] Optionally, step S102 includes: Calculate the difference between the standard wind power output data and the load curve data or the grid dispatch demand data at each time step; The differences are integrated according to the time series to form the reference power curve, wherein the reference power curve is used to reflect the power gap between wind power output and grid load or dispatch demand; the load curve data includes daily periodic fluctuation data and seasonal fluctuation data, wherein the daily periodic fluctuation data reflects load changes at different times of the day, and the seasonal fluctuation data reflects the overall load level differences in different seasons.

[0009] Optionally, step S103 includes: Using the standard wind power output as the decomposition object, ensure that its time dimension matches the subsequent multi-time scale identification requirements, covering time spans of seconds, minutes, hours, days, weeks and above. The SGMD decomposition algorithm is used to decompose the standard wind power output signal, splitting the single power signal into multiple sub-signals with different frequency characteristics. Each sub-signal corresponds to a power component with a different fluctuation frequency. Based on the frequency characteristics of each sub-signal, the corresponding time scale fluctuation characteristics are determined. Among them, high-frequency sub-signals correspond to second-level power fluctuations, mid-to-high-frequency sub-signals correspond to minute-level power fluctuations, mid-to-low-frequency sub-signals correspond to hour-level and daily-level power fluctuations, and low-frequency sub-signals correspond to week-level and above power fluctuations. Simultaneously, the amplitude, duration, and period parameters of each time scale fluctuation are extracted. The identified fluctuation features at various time scales are integrated to generate the multi-time scale fluctuation feature data.

[0010] Optionally, the synchronous extraction of the amplitude, duration, and period parameters of fluctuations at each time scale includes: Taking the sub-signals corresponding to each time scale as the extraction objects, first determine the power reference value corresponding to each sub-signal. The power reference value is the average power value of the corresponding sub-signal within the complete fluctuation period. The absolute value of the difference between the peak power of the sub-signal and the power reference value is calculated as the amplitude of the fluctuation on the corresponding time scale. The monitoring sub-signal power is monitored for the complete time interval from the first deviation from the power reference value to the return to the power reference value. The duration of the time interval is the duration of the fluctuation on the corresponding time scale. For each sub-signal, identify the time points of two consecutive power peaks or two consecutive power valleys, calculate the time difference between two adjacent peaks or two adjacent valleys, and take the average of multiple calculation results as the period parameter of the corresponding time scale fluctuation. After extracting the amplitude, duration, and period parameters, each parameter is compared with the historical fluctuation parameters at the corresponding time scale. If the amplitude deviation exceeds the first preset value, the duration deviation exceeds the second preset value, or the period deviation exceeds the third preset value, the sub-signal decomposition results are re-examined and the parameters are extracted again to ensure that the extracted parameters conform to the actual characteristics of the fluctuation at the corresponding time scale.

[0011] Optionally, step S104 includes: Based on the multi-timescale fluctuation characteristic data, M particles are generated by logical mapping. Each particle is a matrix containing the capacity and power of each energy storage device. At the same time, initial antibodies are generated, which correspond to high-frequency or low-frequency compensation power and reference power for energy storage operation, respectively. Based on the primary objective of meeting the national standards for wind power grid connection and the secondary objective of maximizing economic benefits, the fitness of each particle is calculated by combining the maximum allowable fluctuation after smoothing, the time vector and the whole life cycle cost formula, and the discounted present value of operation and maintenance. The concentration selection mechanism selects N particles with suitable fitness from M particles, and combines them with the N particles newly generated by the logical mapping to form a population of M+N particles, thus avoiding the algorithm from getting trapped in local optima. The position and velocity of particles are updated based on particle fitness, the capacity and power parameters of each energy storage device are adjusted, and the optimization direction is updated synchronously with local and global optimal solutions. If the maximum number of iterations is reached or the global optimal position is obtained, the iteration stops and the optimal capacity configuration results of each energy storage device are output. If not, the iteration returns to the fitness calculation step to continue. The entire configuration process must comply with constraints on energy storage charging and discharging power, energy storage capacity, state of charge, and response time.

[0012] Optionally, step S105 includes: Based on the optimal capacity configuration results, a simulation model covering the entire life cycle of the energy storage system is constructed. The model input includes the multi-timescale fluctuation characteristic data, the reference power curve, and the charging and discharging efficiency and lifespan degradation parameters of the energy storage device. The functionality and compliance are verified through annual simulations. The functionality is verified by testing the effect of the optimal capacity configuration on smoothing load fluctuations at time scales of seconds, minutes, hours, days, weeks and above, to determine whether it meets the grid peak-shaving requirements. The compliance is verified by comparing the smoothed power curve of energy storage obtained from the simulation with the national standard for wind power grid connection to confirm whether it meets the grid connection requirements. The costs of the optimal capacity configuration result over its entire life cycle are calculated, including initial investment costs, operation and maintenance costs, energy storage replacement costs, and energy storage disposal costs. At the same time, the revenues such as peak shaving revenue and policy operation subsidies are also calculated to determine the economic benefits of the optimal capacity configuration result over its entire life cycle. By comparing the simulation verification results and economic evaluation data of various solutions based on the optimal capacity configuration, the solution that meets peak shaving requirements, complies with national grid connection standards, and has the best economic benefits is selected as the optimal energy storage capacity configuration solution.

[0013] Secondly, this application provides a multi-timescale hybrid energy storage capacity configuration system, the multi-timescale hybrid energy storage capacity configuration system comprising: The data acquisition and preprocessing module is used to collect data based on the construction status of wind farms, clean, smooth and handle outliers of actual output power data of completed wind farms, and use the LSTM algorithm to fit the power of proposed wind farms based on wind resources and wind turbine model data to obtain standard wind power output power data. The difference calculation module is used to obtain the local power grid load curve or dispatch demand data, calculate the difference between the standard wind power output power data and the data, and integrate the difference according to the time series to generate the reference power curve of the energy storage system. The power decomposition module is used to decompose standard wind power output power data into multi-frequency components using the SGMD decomposition algorithm, identify power fluctuations at multiple time scales based on frequency components, and simultaneously extract fluctuation parameters and integrate them to generate multi-time scale fluctuation feature data. The capacity configuration optimization module is used to set energy storage response time constraints based on fluctuation characteristic data, and to optimize energy storage capacity using an immune hybrid particle swarm optimization algorithm. It prioritizes meeting grid connection standards and reference power curve requirements, and secondly maximizes economic benefits while adhering to relevant constraints. The simulation verification module is used to perform annual full life cycle simulations on the optimized energy storage capacity configuration, verify the effect of smoothing fluctuations and grid connection compliance at multiple time scales, and evaluate the economics to determine the optimal energy storage capacity configuration scheme.

[0014] Thirdly, this application provides a multi-timescale hybrid energy storage capacity configuration device, including: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the multi-timescale hybrid energy storage capacity configuration device to execute the above-described multi-timescale hybrid energy storage capacity configuration method.

[0015] Fourthly, this application provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the aforementioned multi-timescale hybrid energy storage capacity configuration method.

[0016] This application proposes a multi-timescale hybrid energy storage capacity configuration method, system, and storage medium, applicable to the capacity configuration design of grid-connected energy storage systems for wind farms. It addresses the problems of poor adaptability of existing energy storage configurations to multi-timescale fluctuations, insufficient accuracy in wind power output prediction, difficulty in balancing economic efficiency and functionality, and lack of effective verification mechanisms. Compared with existing technologies, the beneficial effects of this application's technical solution are at least as follows: First, it achieves precise adaptation across multiple time scales. By using the SGMD decomposition algorithm, the wind power output is broken down into different frequency components, identifying power fluctuations at multiple time scales from second-level to minute-level to week-level and above. Energy storage capacity is then configured accordingly, which can comprehensively smooth load fluctuations at various scales, ensure the stability of the power grid under different operating conditions, and avoid the shortcomings of single-time-scale configuration in dealing with fluctuations in multiple scenarios.

[0017] Secondly, improve the reliability of wind power output data. For existing wind farms, preprocess the actual power data and use deep learning algorithms (such as LSTM) to fit the power of planned wind farms. This ensures that the standard wind power output data accurately reflects the actual situation, providing a reliable data foundation for subsequent energy storage configuration and solving the problem of insufficient data accuracy affecting the configuration effect in existing technologies.

[0018] Third, the system balances functionality and economic efficiency, prioritizing compliance with national standards for wind power grid connection while maximizing economic benefits. It employs an immune hybrid particle swarm optimization algorithm to optimize energy storage capacity, while also considering constraints such as charging and discharging power and state of charge to avoid issues such as excessive initial investment and uncontrolled operation and maintenance costs, thus achieving a balance between functionality and economic efficiency in the energy storage system.

[0019] Fourth, ensure the long-term effectiveness of the configuration scheme. Verify the energy storage configuration scheme through annual full life cycle simulation. This will not only verify the effectiveness of smoothing fluctuations and meeting peak-shaving requirements at multiple time scales and the compliance with grid connection, but also evaluate the economic benefits throughout the entire life cycle. This will ensure that the scheme can stably meet the standards during long-term operation and solve the problem of insufficient practicality of the scheme due to the lack of effective verification of existing technologies. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A flowchart of a multi-timescale hybrid energy storage capacity configuration method according to this application; Figure 2This is a structural diagram of a multi-timescale hybrid energy storage capacity configuration system according to this application; Figure 3 This is a structural diagram of a multi-timescale hybrid energy storage capacity configuration device according to this application. Detailed Implementation

[0022] This application provides a method, system, and storage medium for configuring multi-timescale hybrid energy storage capacity. The terms "first," "second," "third," "fourth," etc. (if present)," in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0023] This application relates to the field of data mining application technology, and discloses a method, system, and storage medium for configuring hybrid energy storage capacity across multiple time scales. The method includes: collecting and preprocessing data based on the construction status of wind farms to obtain standard wind power output data; acquiring local grid load curves or dispatch demand data, calculating the difference to generate a reference power curve for energy storage; using the SGMD decomposition algorithm to split the power signal, identifying power fluctuations across multiple time scales to generate feature data; setting constraints based on the feature data, and optimizing the energy storage capacity configuration using an immune hybrid particle swarm optimization algorithm; performing full lifecycle simulation on the optimization results to verify the effectiveness and compliance, and evaluating economic efficiency to determine the optimal solution. This application addresses the problems of poor adaptability and insufficient economic efficiency in existing energy storage configurations through multi-time scale decomposition and closed-loop optimization, improving the grid connection stability and configuration economy of wind power.

[0024] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of a multi-timescale hybrid energy storage capacity configuration method in this application includes: Step S101: Collect data based on the construction status of the wind farm, preprocess the collected data, and obtain standard wind power output data.

[0025] In one specific embodiment, the process of performing step S101 may specifically include the following steps: Determine the construction status of the wind farm, which includes both completed and planned status; If the wind farm is already built, the actual output power data generated during its operation is directly collected. The actual output power data is then cleaned, smoothed, and outlier-handled to obtain the standard wind power output power data. If the project is in the planning stage, local wind resource data and wind turbine model data from the tender are collected. Based on the local wind resource data and wind turbine model data from the tender, the wind power output power is fitted using the LSTM algorithm to obtain the standard wind power output power data of the proposed wind farm.

[0026] Specifically, for wind farms already in operation, it is necessary to collect actual output power data generated during their operation. The collection period should cover at least one year to include power variation characteristics under different seasons and weather conditions. The data collection frequency must meet the needs of subsequent multi-timescale analysis, reaching at least the second-level sampling rate. The collected actual output power data may contain abnormal jump values ​​due to sensor failures, zero or missing values ​​due to equipment maintenance shutdowns, and high-frequency noise caused by grid fluctuations. These data need to be processed. During data cleaning, abnormal fluctuation values ​​exceeding the rated power range of wind turbines are removed by setting a power threshold range. For zero values ​​generated during shutdowns, the shutdown period needs to be confirmed in conjunction with maintenance records before being marked or removed. For missing data values, linear interpolation or imputation using the mean of adjacent time periods is used to complete the data. Smoothing is performed using a moving average method with a 30-second sliding window to filter the cleaned data, reducing the impact of high-frequency noise on subsequent analysis. Outlier handling also requires identifying data points that deviate from the data mean by more than three times the standard deviation using the 3σ principle to further ensure data accuracy. After the above processing, standard wind power output power data is obtained, which can directly reflect the actual power output characteristics of the built wind farms.

[0027] For wind farms under construction, local wind resource data and data on the wind turbine models tendered need to be collected. Local wind resource data includes at least five consecutive years of data on wind speed, wind direction, and air density. Data collection points must be distributed across multiple meteorological towers within the planned wind farm area to ensure the data represents the wind resource distribution of the entire wind farm. Tendered wind turbine model data includes technical parameters such as rated power, cut-in wind speed, cut-out wind speed, rated wind speed, and power curve. These parameters are provided by the wind turbine manufacturer and must comply with relevant industry standards. The wind speed data from the collected local wind resource data is correlated with the power curve from the wind turbine model data. Using wind speed as the input variable and incorporating the air density correction formula for the power curve, the theoretical power output corresponding to different wind speeds is preliminarily calculated. Subsequently, the processed wind speed data and theoretical power output data are used as training samples and input into the LSTM algorithm model. The LSTM algorithm constructs a neural network structure with input, hidden, and output layers, and uses gating units (input gate, forget gate, and output gate) to capture the long-term dependency between wind speed and power output. During model training, wind speed-power data from similar actual wind farms are used as the validation set. By adjusting parameters such as the model's learning rate, the number of neurons in the hidden layer, and the number of iterations, the mean square error between the model's predicted power and the actual power is minimized. After the model converges, the predicted wind speed data for the proposed wind farm for the next year is input into the trained LSTM model to fit and obtain the standard wind power output data of the proposed wind farm. This data can predict the power output characteristics of the proposed wind farm in advance.

[0028] Step S102: Obtain the load curve or grid dispatch demand data of the local power grid, calculate the difference between the standard wind power output power data and the load curve or grid dispatch demand data, and generate the reference power curve of the energy storage system.

[0029] In one specific embodiment, the process of performing step S102 may specifically include the following steps: Calculate the difference between the standard wind power output data and the load curve data or the grid dispatch demand data at each time step; The differences are integrated according to the time series to form the reference power curve, wherein the reference power curve is used to reflect the power gap between wind power output and grid load or dispatch demand; the load curve data includes daily periodic fluctuation data and seasonal fluctuation data, wherein the daily periodic fluctuation data reflects load changes at different times of the day, and the seasonal fluctuation data reflects the overall load level differences in different seasons.

[0030] Specifically, obtaining local power grid load curves or power grid dispatch demand data requires connecting to the local power grid dispatch center database. Load curve data must cover at least a one-year period, including 24-hour daily load values, and the data sampling interval must be consistent with standard wind power output data to ensure time dimension matching. Power grid dispatch demand data must include parameters such as the power capacity limit for wind power grid connection at different times and peak-shaving demand thresholds. These parameters must comply with local power grid operation regulations and national standards for wind power grid connection. In the load curve data, daily periodic fluctuations reflect load changes at different times of the day. For example, on weekdays, load peaks occur between 8:00-10:00 AM and 6:00-10:00 PM due to increased industrial production and residential electricity consumption, while the load is at its lowest point between 2:00 AM and 6:00 AM. Seasonal fluctuations reflect the overall load level differences across seasons. For example, the overall load level is higher in summer due to air conditioning and in winter due to heating demand than in spring and autumn. These fluctuation characteristics need to be categorized, extracted, and labeled with time dimension information.

[0031] When calculating the difference between standard wind power output data and load curve data or grid dispatch demand data at each time step, the timestamp is used as the matching basis. The standard wind power output value at the same time step is subtracted from the load value in the load curve data (or the upper limit of power acceptance in the grid dispatch demand data). If the standard wind power output value is greater than the load value (or the upper limit of power acceptance), the difference is positive, indicating that wind power output is excessive and the energy storage system needs to charge and store the excess energy. If the standard wind power output value is less than the load value (or the upper limit of power acceptance), the difference is negative, indicating that wind power output is insufficient and the energy storage system needs to discharge to supplement the power gap. If the two are equal, the difference is zero, and the energy storage system does not need to take action. During the calculation process, each set of time-based data needs to be verified one by one to avoid calculation errors caused by timestamp misalignment. For time points with abnormal differences (such as the absolute value of the difference exceeding the maximum fluctuation range of the same period in history), it is necessary to backtrack and check whether there are any anomalies in the standard wind power output data and load / dispatch data at the corresponding time step. After ruling out data acquisition or transmission errors, the calculation is recalculated.

[0032] When integrating the differences by time series, the differences calculated at each moment are arranged sequentially in chronological order to form a continuous difference sequence. This sequence is then transformed into a reference power curve using data visualization tools. The horizontal axis of the reference power curve represents time (accurate to the same unit as the data sampling interval), and the vertical axis represents the power difference. Each point on the curve corresponds to the power deficit or surplus at a specific moment, which can intuitively reflect the dynamic balance between wind power output and grid load / dispatch demand.

[0033] Step S103: The standard wind power output data is decomposed into multiple frequency components using the SGMD decomposition algorithm. Based on the frequency components, power fluctuations at different time scales, such as second, minute, hour, day, week and above, are identified to generate multi-time scale fluctuation feature data.

[0034] In one specific embodiment, the process of executing step S103 may specifically include the following steps: Using the standard wind power output as the decomposition object, ensure that its time dimension matches the subsequent multi-time scale identification requirements, covering time spans of seconds, minutes, hours, days, weeks and above. The SGMD decomposition algorithm is used to decompose the standard wind power output signal, splitting the single power signal into multiple sub-signals with different frequency characteristics. Each sub-signal corresponds to a power component with a different fluctuation frequency. Based on the frequency characteristics of each sub-signal, the corresponding time scale fluctuation characteristics are determined. Among them, high-frequency sub-signals correspond to second-level power fluctuations, mid-to-high-frequency sub-signals correspond to minute-level power fluctuations, mid-to-low-frequency sub-signals correspond to hour-level and daily-level power fluctuations, and low-frequency sub-signals correspond to week-level and above power fluctuations. Simultaneously, the amplitude, duration, and period parameters of each time scale fluctuation are extracted. The identified fluctuation features at various time scales are integrated to generate the multi-time scale fluctuation feature data.

[0035] Specifically, the integrity of the time dimension of the standard wind power output is first verified to ensure that its time span covers seconds, minutes, hours, days, weeks and above, and that the data sampling frequency matches the identification requirements of each time scale. Specifically, the sampling interval for second-level data is no more than 1 second, and the sampling interval for minute-level data is no more than 1 minute. As for hourly to weekly and above data sampling, it can be obtained by aggregation calculation based on the preprocessed data of the previous time span. For example, hourly data is obtained by summing and averaging the minute-level data based on the preprocessed minute-level data, thereby ensuring the comprehensiveness and accuracy of subsequent multi-time scale fluctuation identification.

[0036] The SGMD decomposition algorithm is used to decompose the standard wind power output signal. Based on multi-scale geometric analysis theory, the core logic of SGMD is to construct adaptive basis functions to decompose a single power signal into multiple sub-signals in descending order of frequency. This process follows the energy conservation principle of signal decomposition, meaning the sum of the energy of all sub-signals after decomposition is equal to the energy of the original standard wind power output signal. Specifically, the standard wind power output signal is first initialized, and the number of decomposition layers and error threshold are set. The number of decomposition layers is determined based on the number of time scales to be identified (in this embodiment, it is set to 5 layers, corresponding to second-level, minute-level, hour-level, day-level, week-level, and above). The error threshold is set to 5% of the original signal energy. Then, through iterative calculation, each iteration extracts the high-frequency component of the current signal as a sub-signal, and the remaining signal is used as the input for the next iteration, until the number of iterations reaches the set number of decomposition layers or the remaining signal energy is less than the error threshold. Finally, five sub-signals with different frequency characteristics are obtained, each corresponding to a power component with a different fluctuation frequency.

[0037] Based on the frequency characteristics of each sub-signal, the corresponding time-scale fluctuation characteristics are determined. The center frequency of each sub-signal is calculated using a spectrum analysis tool. Sub-signals with a center frequency greater than 1Hz are considered high-frequency sub-signals, corresponding to second-level power fluctuations (such as power fluctuations caused by instantaneous airflow changes in wind turbine blades); sub-signals with a center frequency between 0.0167Hz and 1Hz are considered mid-to-high-frequency sub-signals, corresponding to minute-level power fluctuations (such as power fluctuations caused by short-term gusts); sub-signals with a center frequency between 2.778×10⁻⁻⁶ are considered high-frequency sub-signals. 4 Sub-signals between Hz and 0.0167Hz are low-to-medium frequency sub-signals, corresponding to hourly and daily power fluctuations (such as power fluctuations caused by intraday wind speed changes and diurnal temperature differences); the center frequency is less than 2.778×10⁻ 4 The Hz sub-signal is a low-frequency sub-signal, corresponding to power fluctuations at the week level and above (such as power fluctuations caused by weather changes or seasonal changes within the week).

[0038] The average power value within the complete fluctuation cycle of the sub-signal is taken as the power benchmark value, and the absolute value of the difference between the peak power value and the benchmark value is taken as the fluctuation amplitude. The time interval from the first deviation of the power from the benchmark value to its return to the benchmark value is monitored to obtain the duration. The time points of two consecutive peaks or valleys are identified, the time difference is calculated, and the average value is taken as the period parameter. After extraction, the parameters are compared with historical data of the corresponding time scale. If the amplitude deviation exceeds 10% to 30% (the threshold varies for each time scale), the duration deviation exceeds 20%, or the period deviation exceeds 15%, the decomposition results are re-examined and the parameters are extracted again to ensure that the parameters conform to the actual fluctuation characteristics. Finally, the fluctuation characteristics of each time scale (including sub-signal frequency range, fluctuation amplitude, duration, and period parameter) are integrated in ascending order of time scale to generate multi-time scale fluctuation characteristic data, which provides data support for the optimization of energy storage capacity configuration by the immune hybrid particle swarm algorithm in step S104.

[0039] In one specific embodiment, the synchronous extraction of the amplitude, duration, and period parameters of fluctuations at each time scale may specifically include the following steps: Taking the sub-signals corresponding to each time scale as the extraction objects, first determine the power reference value corresponding to each sub-signal. The power reference value is the average power value of the corresponding sub-signal within the complete fluctuation period. The absolute value of the difference between the peak power of the sub-signal and the power reference value is calculated as the amplitude of the fluctuation on the corresponding time scale. The monitoring sub-signal power is monitored for the complete time interval from the first deviation from the power reference value to the return to the power reference value. The duration of the time interval is the duration of the fluctuation on the corresponding time scale. For each sub-signal, identify the time points of two consecutive power peaks or two consecutive power valleys, calculate the time difference between two adjacent peaks or two adjacent valleys, and take the average of multiple calculation results as the period parameter of the corresponding time scale fluctuation. After extracting the amplitude, duration, and period parameters, each parameter is compared with the historical fluctuation parameters at the corresponding time scale. If the amplitude deviation exceeds the first preset value, the duration deviation exceeds the second preset value, or the period deviation exceeds the third preset value, the sub-signal decomposition results are re-examined and the parameters are extracted again to ensure that the extracted parameters conform to the actual characteristics of the fluctuation at the corresponding time scale.

[0040] Specifically, when simultaneously extracting the amplitude, duration, and period parameters of fluctuations at each time scale, the sub-signals corresponding to each time scale are used as the extraction objects. First, the power reference value corresponding to each sub-signal is determined. The power reference value is taken as the average power value of the corresponding sub-signal within the complete fluctuation period, and the calculation formula is as follows: ,in This is a power reference value (unit: kW). The complete fluctuation period of the sub-signal (unit: s). For the sub-signal at time t The power value (unit: kW) is obtained by integration to obtain the average power level within the period, ensuring that the reference value can reflect the overall power characteristics of the sub-signal.

[0041] The absolute value of the difference between the peak power of the sub-signal and the power reference value is calculated as the amplitude of the fluctuation on the corresponding time scale. The calculation formula is as follows: ,in Fluctuation amplitude (unit: kW) The peak power (in kW) of the sub-signal within its fluctuation period is derived based on the definition of fluctuation amplitude in physics. Taking a high-frequency sub-signal corresponding to a second-level power fluctuation as an example, if within a certain period... ,but That is, the fluctuation range is 80kW on a second-level scale.

[0042] The monitoring sub-signal power is used to track the complete time interval from the first deviation from the power reference value to the return to the power reference value. The duration of this time interval is the duration of the fluctuation on the corresponding time scale. The time of the first deviation is recorded using a timestamp. With the moment of return The formula for calculating the duration is: ,in Duration (in seconds). This is a timestamp (unit: seconds), directly reflecting the duration of power fluctuations. For example, for mid-to-high frequency sub-signals corresponding to minute-level power fluctuations, the first deviation from the reference value occurs at [time missing]. The regression baseline time is ,but That is, the duration of minute-level fluctuations is 120 seconds.

[0043] For each sub-signal, identify the time points of two consecutive power peaks or two consecutive power troughs, calculate the time difference between two adjacent peaks or two adjacent troughs, and take the average of multiple calculations as the period parameter of the corresponding time scale fluctuation. The calculation formula is as follows: ,in The average period (unit: s). The number of peaks or valleys identified. For the first i The timestamps of each peak or trough (in seconds) are averaged through multiple sampling to reduce random errors. Taking the low-to-medium frequency sub-signal corresponding to daily power fluctuations as an example, five consecutive peak timestamps were identified as follows: The time differences are respectively The daily fluctuation period is 86400s.

[0044] After extracting the amplitude, duration, and period parameters, each parameter is compared with historical fluctuation parameters at the corresponding time scale. If the amplitude deviation exceeds the first preset value (10%, 15%, 20%, 25%, and 30% for seconds, minutes, hours, days, and weeks and above, respectively), or the duration deviation exceeds the second preset value (20% for all time scales), or the period deviation exceeds the third preset value (15% for all time scales), the sub-signal decomposition results are re-examined and parameters are extracted again to ensure that the extracted parameters conform to the actual characteristics of fluctuations at the corresponding time scale. The identified fluctuation characteristics at each time scale (including the frequency range, fluctuation amplitude, duration, and period parameters of the corresponding sub-signal) are integrated and arranged in ascending order of time scale to generate multi-time scale fluctuation characteristic data. This data must include a unique identifier for each time scale, the corresponding sub-signal number, and the values ​​of each fluctuation parameter, providing data support for subsequent optimization of energy storage capacity configuration.

[0045] Step S104: Based on the multi-timescale fluctuation characteristic data, set energy storage response time constraints, and use the immune hybrid particle swarm algorithm to optimize the capacity configuration of each energy storage device, so as to prioritize meeting the national standards for wind power grid connection and the fluctuation requirements of the reference power curve, and secondly maximize economic benefits, while following the constraints of energy storage charging and discharging power, energy storage capacity, state of charge and response time.

[0046] In one specific embodiment, the process of executing step S104 may specifically include the following steps: Based on the multi-timescale fluctuation characteristic data, M particles are generated by logical mapping. Each particle is a matrix containing the capacity and power of each energy storage device. At the same time, initial antibodies are generated, which correspond to high-frequency or low-frequency compensation power and reference power for energy storage operation, respectively. Based on the primary objective of meeting the national standards for wind power grid connection and the secondary objective of maximizing economic benefits, the fitness of each particle is calculated by combining the maximum allowable fluctuation after smoothing, the time vector and the whole life cycle cost formula, and the discounted present value of operation and maintenance. The concentration selection mechanism selects N particles with suitable fitness from M particles, and combines them with the N particles newly generated by the logical mapping to form a population of M+N particles, thus avoiding the algorithm from getting trapped in local optima. The position and velocity of particles are updated based on particle fitness, the capacity and power parameters of each energy storage device are adjusted, and the optimization direction is updated synchronously with local and global optimal solutions. If the maximum number of iterations is reached or the global optimal position is obtained, the iteration stops and the optimal capacity configuration results of each energy storage device are output. If not, the iteration returns to the fitness calculation step to continue. The entire configuration process must comply with constraints on energy storage charging and discharging power, energy storage capacity, state of charge, and response time.

[0047] Specifically, the energy storage response time constraint is first set based on the multi-timescale fluctuation characteristic data generated in step S103. The energy storage response time constraint for second-level power fluctuations is no more than 0.5 seconds, for minute-level fluctuations no more than 30 seconds, for hour-level fluctuations no more than 1 hour, and for day-level fluctuations and above no more than 12 hours, to ensure that the operating speed of the energy storage device is adapted to the rate of change of fluctuations at different time scales.

[0048] When using the immune hybrid particle swarm optimization algorithm to optimize the capacity configuration of each energy storage device, M particles are first generated by logical mapping. The value of M is determined according to the type of energy storage device. If three types of devices are configured—flywheel energy storage, lithium battery energy storage, and compressed air energy storage—M is set to 30. The particles are a 6-row, 30-column matrix, with each row corresponding to the capacity (kWh) and power (kW) parameters of the three types of energy storage devices. That is, the elements in the matrix are, in order, the flywheel energy storage capacity, flywheel energy storage power, lithium battery energy storage capacity, lithium battery energy storage power, compressed air energy storage capacity, and compressed air energy storage power. At the same time, initial antibodies are generated, with the number of initial antibodies being the same as the number of particles. Each antibody corresponds to the high-frequency compensation power (adapting to second-level and minute-level fluctuations), the low-frequency compensation power (adapting to hour-level and above fluctuations), and the energy storage action reference power. The reference power value is based on the reference power curve generated in step S102, taking the average of the absolute values ​​of the power gap in each time period of the curve.

[0049] The fitness function should simultaneously reflect the primary objective of meeting the national standards for wind power grid connection and the secondary objective of maximizing economic benefits. The fitness function is constructed as shown in formula (1): (1) in, F This represents the particle fitness value. α The primary objective weight (value 0.8). β The secondary objective weight is set to 0.2, and α+β=1 to ensure a reasonable weight allocation. The value of the main objective function reflects the degree to which the energy storage configuration scheme meets the national standards for wind power grid connection; The value of the secondary objective function reflects the economic efficiency of the energy storage configuration scheme.

[0050] For the main objective function Calculations are made based on national standards such as wind power grid connection, ensuring that the fluctuation of the smoothed energy storage reference power curve meets the grid connection requirements. Combined with the specific clauses in "GB / T19963.1-2021 Technical Regulations for Wind Farm Access to Power Systems" that "the power fluctuation of wind power grid connection shall not exceed 10% of the rated power within 1 minute and 15% of the rated power within 1 hour", as shown in formula (2): (2) Where, Δ P To determine the extent to which the smoothed power fluctuation exceeds the standard, if the smoothed 1-minute fluctuation is ≤10% and the 1-hour fluctuation is ≤15%, then ΔP=0. =100; if the 1-minute fluctuation exceeds the standard by 1% or the 1-hour fluctuation exceeds the standard by 1%, then ΔP increases by 1, and γ is the penalty coefficient (value 5). This penalty coefficient indicates that for every 1% exceeding the standard, Deduct 5 points, minimum =0.

[0051] In specific calculations, the energy storage configuration scheme corresponding to the particle is substituted into the reference power curve generated in step S102, and the power curve is smoothed by simulating the energy storage charging and discharging process. For example, if the power fluctuation of a certain particle configuration scheme is 8% (≤10%) in 1 minute and 12% (≤15%) in 1 hour after smoothing, then Δ P =0, =100; If the power fluctuation of another particle configuration scheme is 12% after smoothing (exceeding the standard by 2%) and 16% after one hour (exceeding the standard by 1%), then Δ P =2+1=3, =100-5×3=85.

[0052] secondary objective function The calculation is based on a life-cycle cost and benefit model. It clarifies that life-cycle costs include initial investment costs, operation and maintenance costs, replacement costs, and disposal costs, while benefits include peak-shaving benefits and policy operating subsidies. The net life-cycle benefit (NPV) is calculated first, and then normalized. The value is shown in formula (3): (3) Wherein, NPV is the net lifecycle revenue of the particle configuration scheme, NPV = lifecycle revenue - lifecycle cost. The minimum net gain among all particles, To obtain the maximum net gain among all particles, the NPV is mapped to 0-100 using linear normalization. The normalization process ensures that net gains of different magnitudes can be converted into fitness scores of a uniform dimension, facilitating comparisons between particles.

[0053] The calculation of total life-cycle benefits is as follows: Peak shaving revenue = peak shaving power × peak shaving price × annual peak shaving duration × lifespan. For example, if the peak shaving power is 200kW, the peak shaving price is 0.5 yuan / kWh, the annual peak shaving duration is 3000h, and the lifespan is 20 years, the peak shaving revenue = 200 × 0.5 × 3000 × 20 = 6,000,000 yuan. Policy subsidy = annual subsidy amount × lifespan. For example, if the annual subsidy is 100,000 yuan, the total subsidy over 20 years = 100,000 × 20 = 2,000,000 yuan. Therefore, the total life-cycle revenue = 6,000,000 + 2,000,000 = 8,000,000 yuan.

[0054] The life cycle cost is calculated as follows: Initial investment cost: ,in For initial investment costs, The unit power investment cost The unit capacity investment cost For the configured energy storage rated power, This refers to the rated capacity of the configured energy storage. If using flywheel energy storage: =1000 yuan / kW =500 yuan / kWh, configuration =500kW =1000kWh, then the initial investment cost of the flywheel = 1000×500 + 500×1000 = 1,000,000 yuan; if lithium battery energy storage: The initial investment cost of lithium batteries Yuan; if compressed air is used for energy storage: Therefore, the initial investment cost of compressed air = 1000 × 400 + 300 × 1500 = 850,000 yuan; the total initial investment cost of the three types of equipment = 1,000,000 + 900,000 + 1,620,000 = 2,750,000 yuan.

[0055] The calculation of operation and maintenance costs is as follows: Discounted present value of maintenance (NC) = Annual maintenance cost × ,in r Let 8% be the discount rate and T be the maintenance lifespan, set to 20 years. Then the discounted present value of maintenance = ≈981,814 yuan; The replacement cost is calculated as follows: ,in To reduce replacement costs, The unit capacity investment cost For the configured energy storage capacity, if the lithium battery needs to be replaced once every 5 years and the physical energy storage does not need to be replaced for 20 years, then the lithium battery will need to be replaced 3 times in 20 years. The replacement cost of the lithium battery = 400 × 1500 × 3 = 1,800,000 yuan. The disposal cost is calculated as follows: Disposal cost = residual value of equipment minus cost; total disposal cost for the three types of equipment is 200,000 yuan. Total life cycle cost = 2,750,000 + 9,818,147 + 1,800,000 + 200,000 = 5,731,814 yuan.

[0056] In summary, NPV = 8,000,000 - 5,731,814 = 2,268,186 yuan. If all particles... , Yuan, then .

[0057] Will Substituting into the fitness function, we can obtain That is, the fitness value of the particle is .

[0058] Particles are selected using a concentration selection mechanism. The concentration of each particle is calculated (concentration = number of particles with a fitness difference < 5 / total number of particles). N particles (N=10) with a concentration < 10% and a top 10 fitness ranking are selected. Then, 10 new particles are generated using logical mapping (parameter values ​​range from 80% to 120% of existing particle parameters), forming a swarm of M+N=40 particles. Particle position and velocity are updated based on their fitness. Position updates follow the classic particle swarm optimization algorithm, while velocity updates introduce an antibody-based promotion and inhibition mechanism. For particles with fitness ≥ 80, the velocity adjustment coefficient is set to 1.2 (promoting movement towards the optimal solution); for particles with fitness < 60, the velocity adjustment coefficient is set to 0.8 (inhibiting ineffective searches). Local optima (parameters corresponding to the highest historical fitness of a particle) and global optima (parameters corresponding to the highest historical fitness of all particles) are updated simultaneously. If the number of iterations reaches 100 or the global optima remains unchanged for 10 consecutive iterations, iteration stops and the optimal capacity configuration result is output; otherwise, the iteration returns to the fitness calculation step to continue. The entire process must adhere to the following constraints: the energy storage charging and discharging power is constrained to be... Energy storage capacity constraint is The state of charge constraint is 20%≤SOC≤80%, and the response time constraint is executed as set above.

[0059] Step S105: Perform annual full life cycle simulation on the optimized energy storage capacity configuration to verify its effectiveness in smoothing load fluctuations and meeting peak shaving requirements at different time scales, as well as its compliance with national standards for wind power grid connection. At the same time, evaluate the economics to determine the optimal energy storage capacity configuration scheme.

[0060] In one specific embodiment, the process of executing step S105 may specifically include the following steps: Based on the optimal capacity configuration results, a simulation model covering the entire life cycle of the energy storage system is constructed. The model input includes the multi-timescale fluctuation characteristic data, the reference power curve, and the charging and discharging efficiency and lifespan degradation parameters of the energy storage device. The functionality and compliance are verified through annual simulations. The functionality is verified by testing the effect of the optimal capacity configuration on smoothing load fluctuations at time scales of seconds, minutes, hours, days, weeks and above, to determine whether it meets the grid peak-shaving requirements. The compliance is verified by comparing the smoothed power curve of energy storage obtained from the simulation with the national standard for wind power grid connection to confirm whether it meets the grid connection requirements. The costs of the optimal capacity configuration result over its entire life cycle are calculated, including initial investment costs, operation and maintenance costs, energy storage replacement costs, and energy storage disposal costs. At the same time, the revenues such as peak shaving revenue and policy operation subsidies are also calculated to determine the economic benefits of the optimal capacity configuration result over its entire life cycle. By comparing the simulation verification results and economic evaluation data of various solutions based on the optimal capacity configuration, the solution that meets peak shaving requirements, complies with national grid connection standards, and has the best economic benefits is selected as the optimal energy storage capacity configuration solution.

[0061] Specifically, a simulation model covering the entire lifecycle of the energy storage system is constructed based on the optimal capacity configuration results. The model input must be precisely matched with the multi-timescale analysis requirements. The multi-timescale fluctuation characteristic data uses a dataset generated in step S103, containing fluctuation amplitudes, durations, and period parameters at the second, minute, hour, day, week, and higher levels. This dataset directly relates to the fluctuation characteristics that the energy storage equipment must cope with at each time scale. The energy storage reference power curve uses time-series data generated in step S102, reflecting the power gap between wind power output and load / dispatch demand, providing a benchmark for the energy storage charging and discharging actions in the simulation. The charging and discharging efficiency and lifespan degradation parameters of the energy storage equipment are set in conjunction with the characteristics of common energy storage technologies, such as flywheel energy storage with a charging and discharging efficiency of 92% and an annual lifespan degradation rate of 0.5%, lithium battery energy storage with a charging and discharging efficiency of 88% and an annual lifespan degradation rate of 3%, and compressed air energy storage with a charging and discharging efficiency of 75% and an annual lifespan degradation rate of 1%. This ensures that the model input data is consistent with the actual operating characteristics of the energy storage equipment, laying the foundation for simulation accuracy.

[0062] When verifying functionality and compliance through annual simulations, the simulation cycle is set to the baseline lifespan of the energy storage system, 20 years. The simulation step size is determined based on the sampling interval of fluctuations at each time scale: a 1-second step size for second-level fluctuations, a 1-minute step size for minute-level fluctuations, a 1-hour step size for hour-level fluctuations, and a corresponding time unit step size for daily-level and above fluctuations, ensuring that the simulation process can accurately capture the dynamic changes of fluctuations at each scale. During the functionality verification process, the capacity, power, response time, and other parameters of each energy storage device in the optimal capacity configuration scheme are substituted into the simulation model to simulate the energy storage charging and discharging actions under fluctuations at different time scales. For example, second-level power fluctuations trigger flywheel energy storage to complete the charging and discharging response within ≤0.5 seconds, and hour-level power fluctuations trigger the coordinated charging and discharging of lithium battery energy storage and compressed air energy storage. By monitoring the power curves at each time scale output in real time, the amplitude and duration of the smoothed power fluctuations are calculated. If the amplitude of second-level fluctuations is ≤2%, minute-level fluctuations are ≤5%, and hour-level and above fluctuations are ≤10%, then the grid peak-shaving requirements are met. During the compliance verification process, the smoothed power curve of the energy storage obtained from the simulation is extracted and compared with the clauses in the national standard for wind power grid connection, which stipulate that "power fluctuation ≤ 10% in 1 minute and power fluctuation ≤ 15% in 1 hour". If the fluctuations in all time periods of the curve meet the standard clauses, the grid connection is deemed compliant.

[0063] When calculating the economic benefits over the entire life cycle, cost and benefit calculations must be based on clear formulas and parameters. For cost accounting, the initial investment cost is calculated as "Initial Investment Cost = Unit Power Investment Cost × Rated Power + Unit Capacity Investment Cost × Rated Capacity". For example, a flywheel energy storage system with a rated power of 500kW and a rated capacity of 1000kWh, a unit power investment cost of 2000 yuan / kW, and a unit capacity investment cost of 500 yuan / kWh, has an initial investment cost of 2000 × 500 + 500 × 1000 = 1,500,000 yuan. Lithium battery energy storage and compressed air energy storage are calculated using the same formula and then summed. Operation and maintenance costs are calculated as "Discounted Value of Operation and Maintenance = Annual Operation and Maintenance Cost ×..." Calculate the discount rate. r Take 8%, lifespan TTaking a 20-year period, the annual operation and maintenance cost is 100,000 yuan (total of the three types of equipment). Substituting into the formula, the present value of operation and maintenance is approximately 981,814 yuan. The replacement cost of energy storage is calculated as replacement cost = unit capacity investment cost × rated capacity × number of replacements. Lithium batteries are replaced once every 5 years, and physical energy storage does not need to be replaced for 20 years. Therefore, lithium batteries are replaced 3 times in 20 years. If the rated capacity of lithium batteries is 1500kWh and the unit capacity investment cost is 400 yuan / kWh, the replacement cost is 400 × 1500 × 3 = 1,800,000 yuan. The disposal cost is calculated according to the equipment residual value deduction rule, totaling 200,000 yuan for the three types of equipment. The total cost is the sum of the above costs. In terms of revenue calculation, peak shaving revenue is calculated as peak shaving revenue = peak shaving power × peak shaving price × annual peak shaving duration × lifespan. For example, if the annual peak shaving power is 200kW, the peak shaving price is 0.5 yuan / kWh, and the annual peak shaving duration is 3000h, then the peak shaving revenue over 20 years is 200 × 0.5 × 3000 × 20 = 6,000,000 yuan. Policy operation subsidies are calculated as policy subsidies = annual subsidy amount × lifespan. If the annual subsidy is 100,000 yuan, the total subsidy over 20 years is 2,000,000 yuan. The total revenue is the sum of peak shaving revenue and subsidies. The economic benefit over the entire life cycle = total revenue - total cost.

[0064] When comparing simulation verification results and economic evaluation data based on various optimal capacity configurations, at least three differentiated configuration schemes need to be generated. These may include adjusting the capacity ratio of flywheel energy storage and lithium battery energy storage, and the power parameters of compressed air energy storage. The simulation and calculation process described above is repeated for each scheme to obtain the functional compliance rate (peak-shaving demand satisfaction rate, grid connection compliance rate) and economic indicators (net life-cycle return, investment payback period) for each scheme. Based on the optimization logic of prioritizing compliance with national wind power grid connection standards and reference power curve fluctuation requirements, and then maximizing economic benefits, the scheme with a 100% functional compliance rate is selected first. Then, the scheme with the highest net life-cycle return among the compliant schemes is chosen as the optimal energy storage capacity configuration scheme.

[0065] The above describes a multi-timescale hybrid energy storage capacity configuration method in the embodiments of this application. The following describes a multi-timescale hybrid energy storage capacity configuration system 200 in the embodiments of this application. Please refer to [link to relevant documentation]. Figure 2 One embodiment of the multi-timescale hybrid energy storage capacity configuration system 200 in this application includes: The data acquisition and preprocessing module 201 is used to collect data based on the construction status of wind farms, clean, smooth and handle outlier values ​​of the actual output power data of completed wind farms, and use the LSTM algorithm to fit the power of the proposed wind farms based on wind resources and wind turbine model data to obtain standard wind power output power data. The difference calculation module 202 is used to obtain the local power grid load curve or dispatch demand data, calculate the difference between the standard wind power output power data and the data, and integrate the difference according to the time series to generate the reference power curve of the energy storage system. Power decomposition module 203 is used to decompose standard wind power output power data into multi-frequency components using the SGMD decomposition algorithm, identify multi-time-scale power fluctuations based on frequency components, and simultaneously extract fluctuation parameters and integrate them to generate multi-time-scale fluctuation feature data. The capacity configuration optimization module 204 is used to set energy storage response time constraints based on fluctuation characteristic data, and to optimize energy storage capacity using an immune hybrid particle swarm algorithm. It prioritizes meeting grid connection standards and reference power curve requirements, and secondly maximizes economic benefits while adhering to relevant constraints. The simulation verification module 205 is used to perform annual full life cycle simulation on the optimized energy storage capacity configuration, verify the effect of smoothing fluctuations and grid connection compliance at multiple time scales, and evaluate the economics to determine the optimal energy storage capacity configuration scheme.

[0066] above Figure 2 This application describes a multi-timescale hybrid energy storage capacity configuration system from the perspective of modular functional entities. The following describes a multi-timescale hybrid energy storage capacity configuration device 300 from the perspective of hardware processing.

[0067] Reference Figure 3 This application also provides a multi-timescale hybrid energy storage capacity configuration device 300, which can be a server, and its internal structure can be as follows: Figure 3 As shown. The multi-timescale hybrid energy storage capacity configuration device includes a processor 302, a memory 303, a display screen 304, an input device 305, a network interface 306, and a database 307 connected via a system bus 301. The processor 302, designed as a computer, provides computing and control capabilities. The memory 303 of the multi-timescale hybrid energy storage capacity configuration device includes a non-volatile storage medium 3031 and internal memory 3032. The non-volatile storage medium 3031 stores the operating system and computer programs. The internal memory 3032 provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database 307 of the multi-timescale hybrid energy storage capacity configuration device stores the data corresponding to this embodiment. The network interface 306 of the multi-timescale hybrid energy storage capacity configuration device is used for communication with external terminals via a network connection. The computer program, executed by the processor, can implement the above-described method.

[0068] Those skilled in the art will understand that Figure 3The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on a multi-timescale hybrid energy storage capacity configuration device on which the present application is applied.

[0069] This application also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the multi-timescale hybrid energy storage capacity configuration method.

[0070] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0071] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, 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 includes several instructions to cause a multi-timescale hybrid energy storage capacity configuration device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0072] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for configuring hybrid energy storage capacity across multiple time scales, characterized in that, Includes the following steps: Step S101: Based on the data collected from the construction status of the wind farm, preprocess the collected data to obtain standard wind power output data; Step S102: Obtain the load curve or grid dispatch demand data of the local power grid, calculate the difference between the standard wind power output power data and the load curve or grid dispatch demand data, and generate the reference power curve of the energy storage system. Step S103: The standard wind power output power data is decomposed into multiple frequency components using the SGMD decomposition algorithm. Based on the frequency components, power fluctuations at different time scales, such as second, minute, hour, day, week and above, are identified to generate multi-time scale fluctuation feature data. Step S104: Based on the multi-timescale fluctuation characteristic data, set energy storage response time constraints, and use the immune hybrid particle swarm algorithm to optimize the capacity configuration of each energy storage device, so as to prioritize meeting the national standards for wind power grid connection and the fluctuation requirements of the reference power curve, and secondly maximize economic benefits, while following the constraints of energy storage charging and discharging power, energy storage capacity, state of charge and response time. Step S105: Perform annual full life cycle simulation on the optimized energy storage capacity configuration to verify its effectiveness in smoothing load fluctuations and meeting peak shaving requirements at different time scales, as well as its compliance with national standards for wind power grid connection. At the same time, evaluate the economics to determine the optimal energy storage capacity configuration scheme.

2. The method for configuring multi-timescale hybrid energy storage capacity according to claim 1, characterized in that, Step S101 includes: Determine the construction status of the wind farm, which includes both completed and planned status; If the wind farm is already built, the actual output power data generated during its operation is directly collected. The actual output power data is then cleaned, smoothed, and outlier-handled to obtain the standard wind power output power data. If the project is in the planning stage, local wind resource data and wind turbine model data from the tender are collected. Based on the local wind resource data and wind turbine model data from the tender, the wind power output power is fitted using the LSTM algorithm to obtain the standard wind power output power data of the proposed wind farm.

3. The method for configuring multi-timescale hybrid energy storage capacity according to claim 1, characterized in that, The load curve data includes daily periodic fluctuation data and seasonal fluctuation data. The daily periodic fluctuation data reflects load changes at different times of the day, and the seasonal fluctuation data reflects the overall load level differences in different seasons. Step S102 includes: Calculate the difference between the standard wind power output data and the load curve data or the grid dispatch demand data at each time step; The differences are integrated in a time series to form the reference power curve, which reflects the power gap between wind power output and grid load or dispatch demand.

4. The method for configuring multi-timescale hybrid energy storage capacity according to claim 1, characterized in that, Step S103 includes: Using the standard wind power output as the decomposition object, ensure that its time dimension matches the subsequent multi-time scale identification requirements, covering time spans of seconds, minutes, hours, days, weeks and above. The SGMD decomposition algorithm is used to decompose the standard wind power output signal, splitting the single power signal into multiple sub-signals with different frequency characteristics. Each sub-signal corresponds to a power component with a different fluctuation frequency. Based on the frequency characteristics of each sub-signal, the corresponding time scale fluctuation characteristics are determined. Among them, high-frequency sub-signals correspond to second-level power fluctuations, mid-to-high-frequency sub-signals correspond to minute-level power fluctuations, mid-to-low-frequency sub-signals correspond to hour-level and daily-level power fluctuations, and low-frequency sub-signals correspond to week-level and above power fluctuations. Simultaneously, the amplitude, duration, and period parameters of each time scale fluctuation are extracted. The identified fluctuation features at various time scales are integrated to generate the multi-time scale fluctuation feature data.

5. The method for configuring multi-timescale hybrid energy storage capacity according to claim 4, characterized in that, The synchronous extraction of the amplitude, duration, and period parameters of fluctuations at each time scale includes: Taking the sub-signals corresponding to each time scale as the extraction objects, first determine the power reference value corresponding to each sub-signal. The power reference value is the average power value of the corresponding sub-signal within the complete fluctuation period. The absolute value of the difference between the peak power of the sub-signal and the power reference value is calculated as the amplitude of the fluctuation on the corresponding time scale. The monitoring sub-signal power is monitored for the complete time interval from the first deviation from the power reference value to the return to the power reference value. The duration of the time interval is the duration of the fluctuation on the corresponding time scale. For each sub-signal, identify the time points of two consecutive power peaks or two consecutive power valleys, calculate the time difference between two adjacent peaks or two adjacent valleys, and take the average of multiple calculation results as the period parameter of the corresponding time scale fluctuation. After extracting the amplitude, duration, and period parameters, each parameter is compared with the historical fluctuation parameters at the corresponding time scale. If the amplitude deviation exceeds the first preset value, the duration deviation exceeds the second preset value, or the period deviation exceeds the third preset value, the sub-signal decomposition results are re-examined and the parameters are extracted again to ensure that the extracted parameters conform to the actual characteristics of the fluctuation at the corresponding time scale.

6. The method for configuring multi-timescale hybrid energy storage capacity according to claim 1, characterized in that, Step S104 includes: Based on the multi-timescale fluctuation characteristic data, M particles are generated by logical mapping. Each particle is a matrix containing the capacity and power of each energy storage device. At the same time, initial antibodies are generated, which correspond to high-frequency or low-frequency compensation power and reference power for energy storage operation, respectively. Based on the primary objective of meeting the national standards for wind power grid connection and the secondary objective of maximizing economic benefits, the fitness of each particle is calculated by combining the maximum allowable fluctuation after smoothing, the time vector and the whole life cycle cost formula, and the discounted present value of operation and maintenance. The concentration selection mechanism selects N particles with suitable fitness from M particles, and combines them with the N particles newly generated by the logical mapping to form a population of M+N particles, thus avoiding the algorithm from getting trapped in local optima. The position and velocity of particles are updated based on particle fitness, the capacity and power parameters of each energy storage device are adjusted, and the optimization direction is updated synchronously with local and global optimal solutions. If the maximum number of iterations is reached or the global optimal position is obtained, the iteration stops and the optimal capacity configuration results of each energy storage device are output. If not, the iteration returns to the fitness calculation step to continue. The entire configuration process must comply with constraints on energy storage charging and discharging power, energy storage capacity, state of charge, and response time.

7. The method for configuring multi-timescale hybrid energy storage capacity according to claim 6, characterized in that, Step S105 includes: Based on the optimal capacity configuration results, a simulation model covering the entire life cycle of the energy storage system is constructed. The model input includes the multi-timescale fluctuation characteristic data, the reference power curve, and the charging and discharging efficiency and lifespan degradation parameters of the energy storage device. The functionality and compliance are verified through annual simulations. The functionality is verified by testing the effect of the optimal capacity configuration on smoothing load fluctuations at time scales of seconds, minutes, hours, days, weeks and above, to determine whether it meets the grid peak-shaving requirements. The compliance is verified by comparing the smoothed power curve of energy storage obtained from the simulation with the national standard for wind power grid connection to confirm whether it meets the grid connection requirements. The costs of the optimal capacity configuration result over its entire life cycle are calculated, including initial investment costs, operation and maintenance costs, energy storage replacement costs, and energy storage disposal costs. At the same time, the revenues such as peak shaving revenue and policy operation subsidies are also calculated to determine the economic benefits of the optimal capacity configuration result over its entire life cycle. By comparing the simulation verification results and economic evaluation data of various solutions based on the optimal capacity configuration, the solution that meets peak shaving requirements, complies with national grid connection standards, and has the best economic benefits is selected as the optimal energy storage capacity configuration solution.

8. A multi-timescale hybrid energy storage capacity configuration system, characterized in that, A method for configuring multi-timescale hybrid energy storage capacity as described in any one of claims 1-7, wherein the multi-timescale hybrid energy storage capacity configuration system comprises: The data acquisition and preprocessing module is used to collect data based on the construction status of wind farms, clean, smooth and handle outliers of actual output power data of completed wind farms, and use the LSTM algorithm to fit the power of proposed wind farms based on wind resources and wind turbine model data to obtain standard wind power output power data. The difference calculation module is used to obtain the local power grid load curve or dispatch demand data, calculate the difference between the standard wind power output power data and the data, and integrate the difference according to the time series to generate the reference power curve of the energy storage system. The power decomposition module is used to decompose standard wind power output power data into multi-frequency components using the SGMD decomposition algorithm, identify power fluctuations at multiple time scales based on frequency components, and simultaneously extract fluctuation parameters and integrate them to generate multi-time scale fluctuation feature data. The capacity configuration optimization module is used to set energy storage response time constraints based on fluctuation characteristic data, and to optimize energy storage capacity using an immune hybrid particle swarm optimization algorithm. It prioritizes meeting grid connection standards and reference power curve requirements, and secondly maximizes economic benefits while adhering to relevant constraints. The simulation verification module is used to perform annual full life cycle simulations on the optimized energy storage capacity configuration, verify the effect of smoothing fluctuations and grid connection compliance at multiple time scales, and evaluate the economics to determine the optimal energy storage capacity configuration scheme.

9. A multi-timescale hybrid energy storage capacity configuration device, characterized in that, It includes a memory and a processor, the memory storing a computer program that can run on the processor, and the processor executing the computer program to implement a multi-timescale hybrid energy storage capacity configuration method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, it causes the processor to execute a multi-timescale hybrid energy storage capacity configuration method as described in any one of claims 1 to 7.

Images