Energy storage power station energy efficiency optimization control method and system

Abstract

The present application relates to the technical field of energy storage control, and discloses an energy storage power station energy efficiency optimization control method and system, the method comprising collecting original data of the energy storage system, obtaining low-frequency profile and high-frequency features through preprocessing, signal decomposition and reconstruction; inputting the low-frequency profile into a constraint model, constructing an energy trend sequence through recursive estimation and fluctuation constraint, generating a corrected energy sequence by fusing smooth high-frequency disturbances when the deviation exceeds the threshold; combining battery information to do temperature compensation and power limitation on the corrected energy sequence, matching to obtain an initial power instruction set, and obtaining an optimized power instruction set through state of charge matching degree optimization; converting it into a control sequence, issuing it to the execution layer after stability verification compensation, and then constructing a response error vector according to the feedback electric parameters of the execution layer, and updating the constraint model parameters to obtain an enhanced constraint model. The method can realize accurate generation and dynamic optimization of the power instruction of the energy storage power station, and improve the energy efficiency control precision and system operation stability.

Description

Technical Field

[0001] This invention relates to the field of energy storage control technology, and in particular to a method and system for optimizing energy efficiency control of energy storage power stations. Background Technology

[0002] Against the backdrop of new power system construction and large-scale grid connection of new energy sources, energy storage power stations are core facilities for mitigating energy fluctuations and ensuring grid stability. Their operational efficiency directly affects energy utilization efficiency and grid operational reliability. The development of data acquisition and control systems provides technical support for the refined management and control of energy storage power stations. The accuracy of data acquisition and control response capabilities directly determine the operational efficiency and stability level of the energy storage system.

[0003] While current energy efficiency control in energy storage power stations relies on data acquisition and control systems to achieve basic regulation of charging and discharging power, significant technical shortcomings remain. The raw data such as current and voltage acquired by the data acquisition and control system are susceptible to interference from multiple sources, including grid fluctuations, equipment operating noise, and instantaneous load disturbances. Traditional processing methods struggle to effectively separate interference from valid signals, leading to distorted energy characteristic extraction. Furthermore, existing control methods lack sufficient coordination with the data acquisition and control system, failing to adapt to the nonlinear and time-varying operating characteristics of energy storage systems and hindering precise power regulation by incorporating constraints such as battery state of charge and temperature.

[0004] This type of control mode is prone to problems such as fluctuations in charging and discharging power and mismatch in equipment operating status, which reduces the overall operating efficiency of the energy storage power station and makes it difficult to meet the high precision and high stability requirements of the new power system for the operation of the energy storage power station. Summary of the Invention

[0005] This invention provides an energy efficiency optimization control method and system for energy storage power stations, which enables accurate processing of energy storage system operation data and effective extraction of energy characteristics. It combines multiple operating constraints to generate optimized power commands, dynamically updates the control model, improves the control accuracy and operational stability of energy storage power stations, and optimizes overall energy efficiency.

[0006] Firstly, in order to solve the above-mentioned technical problems, the present invention provides an energy efficiency optimization control method for energy storage power stations, comprising: The raw data of the energy storage system is collected according to the preset current sampling frequency and voltage sampling frequency. After preprocessing, the low-frequency coefficients are extracted, and then the low-frequency profile is obtained through signal reconstruction method. The low-frequency profile is input into a preset constraint model to obtain an initial state sequence. After applying Kalman filtering and fluctuation constraints to the initial state sequence, an energy trend sequence is constructed. Historical power curves are acquired and their deviation from the energy trend sequence is calculated. If the deviation continuously exceeds a preset deviation threshold, high-frequency disturbances are separated and smoothed to obtain smoothed high-frequency information. The smoothed high-frequency information and the energy trend sequence are then weighted and fused to generate a corrected energy sequence. Obtain the current battery temperature range, perform temperature adaptation calculation and charge / discharge capacity limitation on the corrected energy sequence, eliminate the part that exceeds the actual operating capacity of the battery, obtain the real-time power demand, and match the real-time power demand in the preset demand mapping matrix to obtain the initial power instruction set; Obtain the current state of charge range, calculate the matching degree with the initial power command set, and if the matching degree is lower than the preset matching threshold, adjust the amplitude and compensate the load of the initial power command set to obtain an optimized power command set. The optimized power instruction set is converted into an initial control sequence, and stability verification and compensation are performed. The target control sequence is then generated and executed.

[0007] Secondly, the present invention provides an energy efficiency optimization control system for an energy storage power station, comprising: The data acquisition and decomposition module is used to acquire the raw data of the energy storage system according to the preset current sampling frequency and voltage sampling frequency, extract the low frequency coefficients after preprocessing, and then obtain the low frequency profile through the signal reconstruction method. An energy trend construction module is used to input the low-frequency profile into a preset constraint model to obtain an initial state sequence, and then construct an energy trend sequence after applying Kalman filtering and fluctuation constraints to the initial state sequence. An energy sequence correction module is used to acquire historical power curves and calculate the deviation from the energy trend sequence. If the deviation continuously exceeds a preset deviation threshold, high-frequency disturbances are separated and smoothed to obtain smoothed high-frequency information. The smoothed high-frequency information and the energy trend sequence are then weighted and fused to generate a corrected energy sequence. The initial instruction generation module is used to obtain the battery temperature range at the current moment, perform temperature adaptation conversion and charge / discharge capacity limitation on the corrected energy sequence, eliminate the part that exceeds the actual operating capacity of the battery, obtain the real-time power demand, and match the real-time power demand in the preset demand mapping matrix to obtain the initial power instruction set. The power command optimization module is used to obtain the current state of charge range, calculate the matching degree with the initial power command set, and if the matching degree is lower than the preset matching threshold, adjust the amplitude and compensate the load of the initial power command set to obtain an optimized power command set. The module below the control sequence is used to convert the optimized power instruction set into the initial control sequence, perform stability verification and compensation, generate the target control sequence, and issue it for execution. The feedback error construction module is used to obtain the electrical parameters fed back by the execution layer, perform coordinate transformation and error calculation, and construct a response error vector. The model parameter update module is used to update the parameters of the constraint model according to the response error vector to obtain the enhanced constraint model.

[0008] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention achieves accurate extraction of low-frequency energy trends and effective filtering of high-frequency disturbances through a two-layer processing strategy of preprocessing the original data, wavelet decomposition and reconstruction combined with recursive estimation of the constraint model. At the same time, when the deviation exceeds the threshold, the smoothed high-frequency information is fused to correct the energy sequence, which effectively solves the problem of energy feature extraction distortion under complex working conditions.

[0009] (2) By constructing a full-process energy efficiency control system, this invention optimizes power commands layer by layer by combining multiple constraints such as battery temperature, charge and discharge rate and state of charge, and generates control sequences through extreme working condition stability verification and compensation, which greatly improves the executability of power commands and the stability of energy storage system operation.

[0010] (3) The present invention constructs an error vector by executing the feedback of the electrical parameters of the execution layer, and obtains an enhanced constraint model by iteratively updating the constraint model parameters based on the error characteristics, forming a closed-loop linkage mechanism of instruction execution and model optimization, thereby realizing the dynamic improvement and long-term optimization of energy efficiency control accuracy. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of a method for optimizing energy efficiency control of an energy storage power station according to the first embodiment of the present invention; Figure 2 This is a schematic diagram of the energy efficiency optimization control system for an energy storage power station provided in the second embodiment of the present invention. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] Reference Figure 1 The first embodiment of the present invention provides an energy efficiency optimization control method for an energy storage power station, including steps S11 to S16: S11: Collect raw data of the energy storage system according to the preset current sampling frequency and voltage sampling frequency, extract low-frequency coefficients after preprocessing, and then obtain the low-frequency profile through signal reconstruction method. S12, the low-frequency profile is input into a preset constraint model to obtain an initial state sequence. After performing Kalman filtering and fluctuation constraints on the initial state sequence, an energy trend sequence is constructed. S13, acquire the historical power curve and calculate the deviation from the energy trend sequence. If the deviation continuously exceeds the preset deviation threshold, separate and smooth the high-frequency disturbance to obtain smoothed high-frequency information. Weightedly fuse the smoothed high-frequency information with the energy trend sequence to generate a corrected energy sequence. S14, obtain the battery temperature range at the current moment, perform temperature adaptation calculation and charge / discharge capacity limitation on the corrected energy sequence, eliminate the part that exceeds the actual operating capacity of the battery, obtain the real-time power demand, and match the real-time power demand in the preset demand mapping matrix to obtain the initial power instruction set; S15, obtain the current state of charge range, calculate the matching degree with the initial power command set, and if the matching degree is lower than the preset matching threshold, adjust the amplitude and compensate the load of the initial power command set to obtain an optimized power command set. S16 converts the optimized power instruction set into an initial control sequence, performs stability verification and compensation, generates the target control sequence, and issues it for execution.

[0014] In step S11, the raw data of the energy storage system needs to be collected according to the preset current sampling frequency and voltage sampling frequency. After preprocessing, low-frequency coefficients are extracted, and then the low-frequency profile is obtained through signal reconstruction, including: The raw data of the energy storage system is collected according to the preset current sampling frequency and voltage sampling frequency, and the raw data is time-aligned to obtain a synchronous operation sequence; The synchronized running sequence is decomposed into low-frequency coefficients by using preset wavelet parameters at multiple scales. If the absolute value of the low-frequency coefficient is greater than the preset low-frequency threshold, then the low-frequency coefficient is subjected to mean filtering to obtain the target approximation coefficient; if the absolute value of the low-frequency coefficient is not greater than the preset low-frequency threshold, then it is directly used as the target approximation coefficient. The target approximation coefficients are subjected to inverse wavelet transform to obtain the low-frequency profile.

[0015] First, it should be noted that the raw data acquisition of the energy storage system relies on high-precision sensors deployed in the core loop. These core sensors include current and voltage sensors: the current sensors are installed at the battery cluster output and the power conversion unit (PCS) interface, employing Hall effect sensors with a range of 0-2000A, a measurement accuracy of ±0.2%FS, and a preset sampling frequency of 100Hz (set according to the millisecond-level response requirements of the energy storage system's power regulation; 100Hz can fully capture the current transients during charge-discharge switching, avoiding signal distortion); the voltage sensors are deployed at both ends of the DC bus and battery modules, employing voltage divider sensors with a range of 0-1000V, a measurement accuracy of ±0.1%FS, and a preset sampling frequency of 50Hz (voltage signals are relatively stable; 50Hz can reduce redundancy while ensuring data integrity, and subsequent interpolation can achieve synchronization with current data). Both types of sensors synchronously record microsecond-level timestamps, which are uploaded to the data processing unit in real time via industrial Ethernet to ensure timely and synchronized acquisition.

[0016] After the acquisition process is started, the raw data is summarized according to "timestamp - current value - voltage value". However, due to the difference in sampling frequency (100Hz vs. 50Hz), the timestamps of the data points cannot be directly matched and time alignment is required: taking the current sampling timestamp as the reference (one node every 10ms), the voltage sequence is resampled using linear interpolation, the voltage estimate of each node on the reference time axis is calculated, and a synchronous running sequence is generated.

[0017] When performing multi-scale decomposition on the synchronization sequence, the preset wavelet basis function is "db4" (adapting to the non-stationary and abrupt characteristics of energy storage signals, with balanced time and frequency domain resolution). The decomposition layer is 3 (based on 100 sets of data tests, the test data covers typical operating conditions of energy storage power stations such as morning peak, evening peak, and light load, as well as the entire temperature range of -10℃ to 55℃: layer 1 can only separate high-frequency noise, layer 2 is insufficient for separating mid-frequency fluctuations, and layer 3 can accurately separate the 0-1Hz low-frequency trend, 1-10Hz mid-frequency fluctuations, and 10-50Hz high-frequency noise, balancing accuracy and computational efficiency). Using the Mallat algorithm, the sequence is decomposed into low-frequency approximation coefficients (CA3) and high-frequency detail coefficients (CD1, CD2, CD3): CA3 characterizes the core trend of the slow change in battery state of charge (SOC), CD1 contains 10-50Hz high-frequency noise (sensor electronic noise, circuit ripple), CD2 contains 1-10Hz mid-frequency fluctuations (load switching disturbances), and CD3 contains near-low-frequency small fluctuations (battery internal resistance changes).

[0018] The preset low-frequency threshold is 5% of the system's average power (average power is calculated from the average of the current and voltage products; during steady-state operation, low-frequency coefficient fluctuations are typically ≤3%, and the 5% threshold effectively distinguishes long-term load trends from meaningless spikes, avoiding the false filtering of valid trends). CA3 is evaluated point-by-point: if the absolute value is greater than the threshold, it indicates a trend fluctuation caused by long-term load, and a 5-point mean filter is used (5 sampling points correspond to 50ms, smoothing fluctuations without losing the trend) to obtain the target approximation coefficient; if it is not greater than the threshold, it is directly retained as the target approximation coefficient.

[0019] Finally, the inverse Mallat algorithm corresponding to the decomposition is used to reconstruct the target approximation coefficients and the high-frequency feature sequences of each layer using inverse wavelet transform. For example, CA3 is [12.5, 12.7, 12.6, ...]. By substituting the high-frequency feature sequences of each layer into the transform formula, the reconstructed low-frequency energy profile is output (e.g., [12.545, 12.788, 12.695, ...]), clearly showing the smooth trend of net energy change within the observation period.

[0020] In step S12, the low-frequency profile needs to be input into a preset constraint model to obtain an initial state sequence. After performing Kalman filtering and fluctuation constraints on the initial state sequence, an energy trend sequence is constructed, including: The low-frequency profile is input into a preset constraint model to obtain an initial state sequence; The Kalman filter algorithm is applied to the initial state sequence to obtain the intermediate trend vector; If the fluctuation amplitude of the intermediate trend vector is greater than the preset fluctuation threshold, then the intermediate trend vector is subjected to a moving average filter to obtain a long-term energy representation; if the fluctuation amplitude of the intermediate trend vector is not greater than the preset fluctuation threshold, then it is directly used as the long-term energy representation. By integrating the long-term energy representations in chronological order along the timeline, an energy trend sequence is obtained.

[0021] First, it should be noted that the preset constraint model is a multi-dimensional constraint framework built on the physical characteristics and safe operation boundaries of the core component of the energy storage system (lithium-ion battery). Its core function is to perform "boundary verification + anomaly correction" on the low-frequency energy profile to ensure that the output initial state sequence not only matches the actual operating capability of the system, but also avoids the risk of equipment damage.

[0022] The constraint model incorporates three core, insurmountable constraints: First, the safe state of charge (SOC) range, strictly limited to 10%–90% (determined based on the electrochemical characteristics of lithium-ion batteries; SOC below 10% leads to lithium plating on the negative electrode and permanent capacity decay, while SOC above 90% easily triggers electrolyte decomposition and thermal runaway risks; this range maximizes battery cycle life while ensuring safety); Second, the maximum charge / discharge power constraint, set at ±110% of the system's rated power (taking a 1000kW rated power system as an example, the constraint range is -1100kW to +1100kW, with some reserves). The system includes three main components: a 10% redundancy to handle short-term grid fluctuations while avoiding exceeding hardware limits; a temperature-adaptive power constraint, dynamically adjusting the allowable power through a built-in temperature-power mapping table calibrated by high and low temperature environment experiments (e.g., power limit drops to 60% of rated power at -10~0℃; to 80% at 0~10℃; 10~40℃ is the optimal operating range, maintaining 100% power; and to 70% at 40~55℃). The core function is to compensate for the impact of temperature on battery ion conductivity and polarization impedance, avoiding performance degradation caused by high-power charging and discharging under extreme temperatures. After inputting the low-frequency profile into the constraint model, the system verifies the profile values ​​time-by-time: if it exceeds the SOC range, it corrects according to the boundary value (e.g., SOC=8% corrected to 10%); if it exceeds the temperature-adapted power constraint, it truncates according to the corresponding upper limit; if there are numerical jumps (e.g., power difference between adjacent times > 200kW), linear interpolation is used to fill the gap, ultimately outputting a continuous and compliant initial state sequence.

[0023] The initial state sequence is recursively estimated using the Kalman filter algorithm. The core objective is to filter out short-term random disturbances caused by real-time load fluctuations and sensor electronic noise, while preserving the long-term energy change trend of the system. Kalman filtering is achieved through a two-step iterative process of "prediction-update": In the prediction stage, based on the posterior state estimate of the previous moment and the preset state transition matrix, the prior state estimate of the current moment is calculated. The state transition matrix is ​​a 2×2 matrix (the dimension corresponds to "power amplitude - rate of change"), calibrated to [[0.995, 0.003], [0.002, 0.994]] by 100 sets of steady-state operation data of the energy storage system. The diagonal elements are close to 1 to ensure trend continuity, and the off-diagonal elements are small to avoid state coupling distortion. At the same time, the initial value of the state prediction error covariance matrix is ​​set to [[0.1, 0], [0, 0.08]]. In the update stage, the Kalman gain is calculated by combining the observed values ​​of the initial state sequence at the current moment and the observation noise covariance matrix (the observation noise covariance matrix is ​​determined based on the statistical analysis of sensor measurement errors, and the initial value is set to [[0.5, 0], [0, 0.2]], reflecting the uncertainty of current and voltage measurements). Furthermore, the state prediction error covariance matrix is ​​updated according to (which is the identity matrix, The observation matrix is ​​[[1,0],[0,1]]. The prior estimates are weighted and corrected using the gain matrix to obtain the posterior state estimates. The above recursive process is repeated time-by-time, and finally a smooth and continuous intermediate trend vector is output. This vector has initially removed short-term disturbances and is close to the true energy change trajectory of the system. For example, the initial state sequence is [520kW, 550kW, 510kW, 530kW, 560kW], and the intermediate trend vector obtained after Kalman filtering is [520.5kW, 532.3kW, 528.1kW, 529.5kW, 535.2kW]. The short-term fluctuation amplitude is reduced from 40kW to 7.1kW, and the trend stability is significantly improved.

[0024] The preset fluctuation threshold is set based on the steady-state operating characteristics of the energy storage system and is defined as the upper limit of the power change rate between adjacent time nodes. The specific value is 5kW / 10ms (i.e., 500kW / min, determined in combination with a 100Hz sampling frequency: 10ms is a single sampling interval. When there is no sudden load disturbance, the steady-state power change rate of the system usually does not exceed 3kW / 10ms. Setting it to 5kW / 10ms can effectively distinguish between "normal slow trend changes" and "abnormal drastic fluctuations that need to be corrected"). The system calculates the fluctuation amplitude of the intermediate trend vector in real time between adjacent moments and compares it with a preset fluctuation threshold: if the fluctuation amplitude is greater than 5kW / 10ms, it indicates that there is a violent disturbance that has not been completely filtered out during this period, and a second smoothing is required using a moving average filter with a window length of 5 (the window length is set according to the sampling frequency, with 5 sampling points corresponding to 50ms. After testing, its smoothing effect on violent fluctuations is 25% better than that of the 3-point moving average, and the trend lag time is controlled within 20ms) to obtain the long-term energy representation; if the fluctuation amplitude is no greater than 5kW / 10ms, the intermediate trend vector of this segment is directly used as the long-term energy representation, preserving the true slow trend. For example, the values ​​of adjacent moments in the intermediate trend vector are 535.2kW and 557.2kW, with a fluctuation range of 22kW / 10ms (far exceeding the threshold). After filtering with a 5-point moving average, the long-term energy representation for this period is corrected to [535.2kW, 542.6kW, 548.3kW, 553.1kW, 557.2kW], with the fluctuation gradually smoothing out and the trend transition becoming more gradual.

[0025] Finally, following the chronological order of the sampling timeline, the long-term energy representations for all time periods within the monitoring period are sequentially integrated and serialized to construct a continuous, smooth energy trend sequence without abnormal fluctuations. This sequence fully depicts the long-term net energy change pattern of the energy storage system within the observation period. For example, after integrating the long-term energy representations for 10 consecutive time periods, the final energy trend sequence is [518.5kW, 520.5kW, 523.2kW, 528.1kW, 529.5kW, 535.2kW, 538.7kW, 542.3kW, 545.1kW, 547.6kW].

[0026] In step S13, historical power curves are acquired and their deviation from the energy trend sequence is calculated. If the deviation continuously exceeds a preset deviation threshold, high-frequency disturbances are separated and smoothed to obtain smoothed high-frequency information. The smoothed high-frequency information and the energy trend sequence are then weighted and fused to generate a corrected energy sequence, including: Obtain the historical power curve, calculate the amplitude difference between the historical power curve and the energy trend sequence within the same time window, and obtain the deviation sequence; If the deviation sequence is greater than the preset deviation threshold within a preset continuous time window, the corresponding original data is extracted and the corresponding energy trend sequence is subtracted to obtain the corresponding high-frequency perturbation; the energy spectral density of the high-frequency perturbation is calculated by Fourier transform, and the energy distribution range is extracted. The high-frequency disturbances corresponding to the energy distribution range are smoothed by a preset filter to obtain smooth high-frequency information; The smoothed high-frequency information and the energy trend sequence are weighted and fused in the time domain to generate a corrected energy sequence.

[0027] First, it should be noted that the historical power curve is the actual charge and discharge power sequence recorded within a complete operating cycle (e.g., 24 hours) of the energy storage system. The data sampling frequency is consistent with the energy trend sequence (100Hz), and it is divided according to a uniform time granularity (every 15 minutes is a time window) to ensure the spatiotemporal consistency of the comparison process. The setting of this time window is determined by comprehensively considering the operating characteristics of the energy storage system, the validity of data statistics, and industry standards: it can fully capture the short- and medium-term fluctuations of grid load and renewable energy output (typical cycle 5-30 minutes), and the statistical reliability of deviation calculation is ensured by 90,000 sampling points (100Hz × 900 seconds). At the same time, it matches the commonly used time units of grid dispatch, making the analysis results more consistent with engineering practice. The historical power curve is stored in the system database, containing real operating data of the energy storage system under similar operating conditions (e.g., the same time period, similar weather conditions), and similar battery states (SOC, temperature). It serves as a verification benchmark for the energy trend sequence, and deviation analysis can identify whether the trend sequence has missed real short- and medium-term fluctuations due to excessive smoothing.

[0028] When calculating the deviation sequence, the amplitude difference between the energy trend sequence and the historical power curve is calculated precisely window by window, using a 15-minute time window as the unit. For all sampling points within each window, the absolute difference between the corresponding time value of the energy trend sequence and the historical power curve value is calculated one by one. Then, the arithmetic mean of all absolute differences within the window is taken to obtain the deviation value of that window. The deviation values ​​of all windows are arranged in chronological order to form a complete deviation sequence. For example, within a certain 15-minute time window, the set of absolute differences calculated point by point is [42kW, 45kW, 43kW, ...] (a total of 90,000 data points). After taking the average, the deviation value of that window is 44kW. If the deviation values ​​of three consecutive adjacent windows are 48kW, 52kW, and 46kW, respectively, then the corresponding segment of the deviation sequence is [48kW, 52kW, 46kW], which intuitively and quantitatively reflects the degree of continuous deviation between the energy trend sequence and the historical actual operating state.

[0029] The preset deviation threshold is set based on the allowable range of steady-state operating error of the energy storage system, and is taken as 5% of the system's rated power (taking a rated power of 1000kW as an example, the threshold is set to 50kW). This value is determined by calibrating and determining the historical deviation data under 50 similar operating conditions: when the system is operating normally, the deviation between the energy trend sequence and the historical power curve usually does not exceed 3% of the rated power. Setting the threshold to 5% can effectively distinguish between "significant and continuous deviations that need correction" and "random errors that do not need to be handled", balancing control accuracy and system overhead. If the deviation value of three or more consecutive time windows in the deviation sequence is greater than 50kW (three consecutive windows correspond to 45 minutes, which can effectively exclude instantaneous interference and ensure that it is a continuous deviation), it is determined that the energy trend sequence has missed key short- and medium-term fluctuations, and the high-frequency disturbance separation process should be initiated immediately; if only a single window or a non-continuous window deviation exceeds the threshold, it is regarded as a random error and no additional processing is required.

[0030] The separation of high-frequency disturbances specifically involves subtracting the corresponding values ​​of the energy trend sequence from the original synchronous operation sequence (the power sequence obtained by calculating current and voltage after time alignment) moment by moment, thus extracting the high-frequency components suppressed by trend smoothing. These components are the high-frequency disturbances. Physically, they represent short- to medium-term fluctuations that exist in the actual operation of the energy storage system but are not covered by the long-term energy trend, such as power changes caused by intermittent industrial load switching, small fluctuations in grid voltage, and fluctuations in renewable energy output. For example, if the power value of the original synchronous operation sequence at a certain moment is 580kW, and the corresponding value of the energy trend sequence at that moment is 525kW, then the high-frequency disturbance separated at that moment is 55kW. After completing the calculation point by point along the time axis, a complete and continuous high-frequency disturbance sequence is obtained [52kW, 48kW, 55kW, 51kW, ...].

[0031] A Fast Fourier Transform (FFT) is performed on the separated high-frequency disturbance sequence to convert the time-domain signal into a frequency-domain signal, and then the energy spectral density (unit: kW² / Hz) is calculated. By analyzing the peak distribution of the energy spectral density curve, the effective energy distribution range of the high-frequency disturbance is extracted—the frequency range corresponding to the concentration of energy spectral density peaks is the effective fluctuation range. This can accurately eliminate ultra-high frequency sensor electronic noise above 100Hz and low-frequency trend residual components below 0.5Hz. For example, after Fourier transform and energy spectral density analysis, it was found that the energy of the high-frequency disturbance is mainly concentrated in the 2.5Hz to 8Hz range, and the energy in this range accounts for more than 85% of the total energy. Therefore, 2.5Hz to 8Hz is determined as the effective energy distribution range.

[0032] A finite-length impulse response (FIR) bandpass filter is used to smooth high-frequency disturbances within the effective energy distribution range. The filter parameters are optimized through engineering practice: the passband strictly matches the effective energy distribution range (2.5Hz~8Hz), the stopband attenuation is set to 40dB (to ensure strong suppression of noise and clutter outside the passband), and the filter order is set to 16 (determined by compromising computational complexity and filtering accuracy; 16th order ensures a flat frequency response within the passband while avoiding signal delay and computational resource waste caused by excessively high order). For example, the original high-frequency disturbance sequence is [52kW, 68kW, 48kW, 72kW, 55kW]. After processing by the FIR bandpass filter, the smoothed high-frequency information is [51.8kW, 62.3kW, 49.1kW, 65.7kW, 54.5kW], with the fluctuation amplitude tending to be stable and no obvious spike noise.

[0033] Finally, the smoothed high-frequency information and the energy trend sequence are weighted and fused in the time domain. The weight allocation follows the core principle of "trend-driven, fluctuation-supplemented": the weight coefficient of the energy trend sequence is set to 0.85, and the weight coefficient of the smoothed high-frequency information is 0.15. The 85% trend weight ensures that the long-term energy change pattern is not disrupted, guaranteeing the stability of the energy storage system; the 15% fluctuation weight accurately supplements the missing details of real short- and medium-term fluctuations. The energy trend sequence value and the smoothed high-frequency information are multiplied by their respective weights and then added together to obtain the corrected energy sequence value. For example, if the energy trend sequence value at a certain moment is 525kW, and the corresponding smoothed high-frequency information value at that moment is 62.3kW, then the corrected value at that moment is 531.3kW. After fusion is completed point by point along the time axis, a complete corrected energy sequence is generated.

[0034] In step S14, it is necessary to obtain the current battery temperature range, perform temperature adaptation calculations and charge / discharge capacity limits on the corrected energy sequence, eliminate the portion exceeding the actual operating capacity of the battery, obtain the real-time power demand, and match the real-time power demand in a preset demand mapping matrix to obtain an initial power instruction set, including: Obtain the current battery temperature range and look up the corresponding temperature decay factor in the preset temperature mapping table; The amplitude of the corrected energy sequence is calculated based on the temperature decay factor to obtain the compensated energy sequence; Obtain the maximum charge / discharge rate of the current battery, and construct a charge / discharge rate limit boundary for the compensated energy sequence; If the value of the compensation energy sequence exceeds the charge / discharge rate limit boundary, it is truncated to obtain the real-time power requirement; The initial power instruction set is obtained by matching the real-time power demand in a preset demand mapping matrix.

[0035] First, it should be noted that the core battery information at the current moment includes the battery temperature range and the maximum charge / discharge rate, both of which are acquired in real time by the Battery Management System (BMS): The battery temperature is collected using distributed temperature sensors, with 8 temperature measurement points deployed on the surface of different modules and cells in the battery cluster. The average value of all temperature measurement points is taken as the current battery temperature, and then the temperature range is determined according to preset range division rules (-10~0℃, 0~10℃, 10~25℃, 25~40℃, 40~55℃), with a measurement accuracy of ±0.5℃; The maximum charge / discharge rate is dynamically calculated based on the battery's current state of charge (SOC), cycle count, and state of health (SOH). The BMS outputs the maximum charge / discharge rate that the current battery can safely withstand (such as 1C, 1.2C, 0.8C, etc.) by monitoring the cell voltage, current, and historical cycle data in real time, ensuring that the power limit boundary is consistent with the battery's real-time health status.

[0036] The preset temperature mapping table is based on the electrochemical characteristics of lithium-ion batteries. It is obtained by testing the charge and discharge capabilities of batteries in the full temperature range of -10 to 55℃ under rated operating conditions of 0 to 2000 cycles and SOC of 20% to 80% for matching lithium-ion batteries with nominal capacity. Its core function is to match the corresponding temperature decay factor according to the temperature range to compensate for the influence of temperature on the battery's charge and discharge capabilities. Low temperatures will reduce the battery's ionic conductivity and increase polarization resistance, while high temperatures will accelerate electrolyte decomposition and electrode aging. Both of these require the energy sequence amplitude to be reduced through the decay factor to avoid damaging the battery. The specific settings for the temperature mapping table are as follows: In the temperature range of -10 to 0℃, the attenuation factor is 0.6 (ion migration ability decreases significantly at low temperatures, only 60% of the rated capacity can be output); in the range of 0 to 10℃, the attenuation factor is 0.8 (ionic conductivity gradually recovers, and charge / discharge capacity improves); in the range of 10 to 25℃, the attenuation factor is 1.0 (the optimal operating temperature range for the battery, no attenuation required); in the range of 25 to 40℃, the attenuation factor is 0.9 (high temperatures begin to affect battery stability, resulting in slight attenuation); in the range of 40 to 55℃, the attenuation factor is 0.7 (risk increases at high temperatures, further limiting power). For example, if the current battery temperature is calculated to be 8℃, falling within the 0 to 10℃ range, the corresponding temperature attenuation factor found in the temperature mapping table is 0.8.

[0037] The amplitude of the corrected energy sequence is converted according to the temperature decay factor. The compensated energy sequence value is obtained by multiplying the corrected energy sequence value by the temperature decay factor. The compensated energy sequence is then calculated point by point along the time axis. For example, if the corrected energy sequence value at a certain moment is 531.3kW and the temperature decay factor is 0.8, the corresponding compensated energy sequence value is 425.04kW. After point-by-point conversion along the time axis, a complete compensated energy sequence is formed.

[0038] After obtaining the current battery's maximum charge / discharge rate, the charge / discharge rate limit boundary is obtained by multiplying the battery's rated capacity by the maximum charge / discharge rate. The upper limit is a positive value during discharge, and the lower limit is a negative value during charging. For example, if the current battery's maximum charge / discharge rate is 1.2C and its rated capacity is 1000kWh, the calculated charge / discharge rate limit boundary is -1200kW (upper charging limit) to 1200kW (lower discharging limit), thus constructing a bidirectional power limiting range.

[0039] The compensated energy sequence is truncated moment by moment: if the value of the compensated energy sequence at a certain moment is within the charge / discharge rate limit boundary (i.e., ≥-1200kW and ≤1200kW), then the value is directly retained as the real-time power demand; if the value exceeds the boundary (e.g., reaching 1300kW during discharge, or -1300kW during charging), then it is truncated according to the boundary value, and the excess part is forcibly corrected to the corresponding boundary value to avoid the instantaneous power demand exceeding the battery hardware's capacity. For example, if the compensated energy sequence value at a certain moment is 1250kW (exceeding the 1200kW discharge boundary), then the truncated real-time power demand is 1200kW; if the value at a certain moment is -1280kW (exceeding the -1200kW charging boundary), then the truncated value is -1200kW. After point-by-point processing, a complete real-time power demand sequence is obtained.

[0040] The pre-defined demand mapping matrix is ​​a two-dimensional matrix constructed based on the control strategy and equipment operating characteristics of the energy storage system. The row dimension represents the range of real-time power demand (e.g., 0~200kW, 200~400kW, ..., 1000~1200kW; the charging range is similarly represented by a negative range). The column dimension represents the corresponding control mode (constant power control, constant current control, constant voltage control) and command parameters (e.g., current threshold, voltage threshold, switching conditions, etc.). The mapping matrix is ​​based on the PCS factory rated parameters and the physical characteristics of the battery clusters, combined with over six months of actual operating data from the power station. It has been constructed through multiple rounds of control strategy trial runs and verification to ensure that each power demand range can be matched with the optimal control command that maximizes charging and discharging efficiency and minimizes battery loss. Based on the specific value of the real-time power demand, the corresponding row dimension range is found in the mapping matrix, and the corresponding control mode and parameters are extracted and combined to form the initial power command set. For example, the real-time power demand is 425.04kW, which falls within the discharge range of 400~600kW. The "constant power discharge control" mode is matched from the mapping matrix, and the corresponding instruction parameters are "discharge power 425kW, voltage upper limit 750V, current upper limit 567A, SOC lower limit 10%". These instructions are combined to form the initial power instruction set.

[0041] In step S15, it is necessary to obtain the current state of charge range and calculate the matching degree with the initial power command set. If the matching degree is lower than a preset matching threshold, the initial power command set is adjusted in amplitude and the load is compensated to obtain an optimized power command set, including: Obtain the current state of charge range of the energy storage system, and calculate the matching degree between the initial power command set and the current state of charge range through a preset matching calculation model; If the matching degree is lower than the preset matching threshold, the initial power instruction set is initially corrected according to the preset instruction adjustment range to obtain the corrected power instruction set; if the matching degree is higher than the preset matching threshold, it is directly used as the corrected power instruction set. The instantaneous fluctuation characteristics of the load are extracted from the real-time operation data of the energy storage system and the fluctuation intensity is quantified. The corrected power instruction set is weighted and compensated according to the fluctuation intensity by matching dynamic weights to generate a fluctuation compensation instruction set. The fluctuation compensation instruction set is subjected to boundary verification by a preset multi-constraint filter to determine the constraint satisfaction sequence; The modified power instruction set is fine-tuned based on the constraint satisfaction sequence to obtain the optimized power instruction set.

[0042] First, it should be noted that the current state of charge (SOC) range is calculated in real time by the battery management system (BMS) using an ampere-hour integration method combined with open-circuit voltage calibration: the initial SOC value is calculated by integrating the charge and discharge current, and calibration is performed hourly using the mapping relationship between the open-circuit voltage and SOC when the battery is at rest, ensuring a measurement accuracy of ±2%. The SOC range is divided into [10%, 90%] according to the safe operating interval, and the current SOC value and fluctuation range are output in real time (e.g., if the current SOC is 65%, the fluctuation range is ±3%, that is, the actual SOC range is 62% to 68%). This range directly determines the remaining energy that the battery can release or absorb.

[0043] The preset matching calculation model is a multi-dimensional quantitative model. It calculates the matching degree by integrating three key indicators: SOC margin, power command duration, and battery health (SOH), with values ​​ranging from 0 to 1 (1 for complete matching and 0 for complete mismatch). The weight coefficients of the three indicators are set according to the operating principle of energy storage systems: "safety first, feasibility second, and lifespan considered." The SOC margin coefficient has a weight of 0.6 (as a core prerequisite for command adaptation, directly determining the basis for the battery's safe execution of commands), the duration adaptation coefficient has a weight of 0.3 (the matching between energy demand and available energy can be corrected through parameter adjustment), and the SOH adaptation coefficient has a weight of 0.1 (affecting only the long-term cycle life of the battery). First, the SOC margin coefficient for discharging (or charging) is obtained by dividing the difference between the current SOC and the lower discharge limit by the difference between the upper charge limit and the lower discharge limit (or by dividing the difference between the upper charge limit and the current SOC by the difference between the upper charge limit and the lower discharge limit). Then, the duration matching coefficient is obtained by dividing the product of the command duration and the command power by the available energy (obtained by multiplying the battery rated capacity by the SOC margin), and then subtracting 1. The absolute value of the result is then subtracted by 1 to obtain the duration matching coefficient. The current SOH is directly used as the SOH matching coefficient. Finally, the SOC margin coefficient, duration matching coefficient, and SOH matching coefficient are multiplied by their corresponding weights and then added together to obtain the matching degree. For example, the initial power command set is "constant power discharge 425kW, lasting 2 hours", the current SOC range is 62% to 68% (average 65%), the rated capacity is 1000kWh, and the SOH is 95%. The SOC margin coefficient is (65%-10%) / (90%-10%) = 0.6875; the available energy is 1000×55%=550kWh, the commanded energy consumption is 425×2=850kWh, the duration adaptation coefficient is 1 - |850 / 550 - 1|≈0.3529; the SOH adaptation coefficient is 0.95, and the final matching degree is 0.54.

[0044] The preset matching threshold is set to 0.7 (determined by statistically analyzing command execution data from 100 sets of different SOC states: when the matching degree is ≥0.7, the battery will not trigger SOC protection during command execution, and the operational stability is improved by more than 90%; when it is below 0.7, it is easy to cause power limiting or shutdown due to insufficient energy). If the calculated matching degree is below 0.7, it indicates that the initial power command set is not well adapted to the current SOC range, and a correction process needs to be initiated; if it is ≥0.7, the initial power command set is directly used as the optimized power command set without additional processing.

[0045] The preset command adjustment range is a step-by-step adjustment rule based on the SOC margin. This rule is set according to the ratio of the battery's available energy to the commanded energy demand. It was calibrated by statistically analyzing multiple sets of operating data matching different SOCs and power commands. The core is to correct the commanded power in reverse according to the "available energy" to avoid energy overdraft: when the matching degree is 0.5 to 0.7, the power adjustment range is 20% of the initial power; when the matching degree is 0.3 to 0.5, the adjustment range is 40%; when the matching degree is <0.3, the adjustment range is 60%. For example, with a current matching degree of 0.54 and an initial power of 425kW, a 20% reduction results in a preliminary corrected power command of 340kW, forming a corrected power command set of "constant power discharge 340kW, lasting 2 hours, voltage upper limit 750V, current upper limit 453A", which initially solves the energy matching problem.

[0046] The instantaneous fluctuation characteristics of the load are extracted from the real-time operating data of the energy storage system. The standard deviation and peak factor (peak value divided by RMS value) of the load current are calculated using a sliding window method (window length 50ms, corresponding to 5 sampling points) to quantify the fluctuation intensity: cases where both standard deviation < 20A and peak factor < 1.5 are classified as weak fluctuations; cases that do not meet the weak fluctuation condition but simultaneously meet the standard deviation ≤ 50A and peak factor ≤ 2.5 are classified as moderate fluctuations; and all remaining operating conditions not covered by weak or moderate fluctuations (i.e., cases where standard deviation > 50A or peak factor > 2.5) are uniformly classified as strong fluctuations. Dynamic weights are assigned based on fluctuation intensity: strong fluctuations have a weight of 1.15, moderate fluctuations have a weight of 1.05, and weak fluctuations have a weight of 0.95, allowing correction commands to adapt to dynamic load changes and avoiding power gaps caused by sudden load changes. For example, real-time load data is extracted and the standard deviation is calculated to be 35A and the peak factor is 2.0. It is judged as medium fluctuation and a dynamic weight of 1.05 is assigned. The power correction command is 340kW×1.05=357kW, and the fluctuation compensation command set is generated as "constant power discharge 357kW, lasting for 2 hours, voltage limit 750V, current limit 476A".

[0047] The preset multi-constraint filter incorporates three types of hard constraints: first, the individual cell voltage constraint (lithium-ion battery cell voltage 2.8V~4.2V, corresponding to a total battery cluster voltage of 448V~672V); second, the transient current ramp-up rate constraint (≤5A / ms, to avoid damage to power devices from sudden current changes); and third, the battery temperature rise constraint (temperature rise ≤5℃ / h during charging and discharging, to prevent thermal runaway). During constraint verification, the execution process of the fluctuation compensation instruction set is simulated, and the above constraints are checked at each moment: if the total voltage corresponding to the instruction at a certain moment exceeds the range of 448V~672V, or the current ramp-up rate is >5A / ms, or the predicted temperature rise is >5℃ / h, it is marked as "constraint not satisfied"; otherwise, it is marked as "constraint satisfied", ultimately forming a constraint satisfaction sequence (e.g., [satisfied, satisfied, not satisfied, satisfied, ...]). For example, the current corresponding to the fluctuation compensation command of 357kW is 476A. Simulation verification shows that the current ramp rate is 3.2A / ms (≤5A / ms), the total voltage is 620V (within the range of 448V to 672V), and the predicted temperature rise is 3.8℃ / h (≤5℃ / h). The entire time period is marked as "constraint satisfied", and the constraint satisfaction sequence is a fully satisfied state.

[0048] The power instruction set is fine-tuned based on the constraint satisfaction sequence: if all constraints in the sequence are "satisfied," the fluctuation compensation instruction set is directly used as the optimized power instruction set; if there are "unsatisfied" nodes, the parameters are adjusted for the type of constraint that is not satisfied (e.g., if the voltage exceeds the upper limit, the voltage threshold is lowered; if the current ramp rate exceeds the limit, the soft-start time is increased). For example, since all constraints are satisfied, the final optimized power instruction set is "constant power discharge 357kW, lasting 2 hours, upper voltage limit 750V, upper current limit 476A, current ramp rate ≤5A / ms, temperature rise ≤5℃ / h."

[0049] In step S16, the optimized power instruction set needs to be converted into an initial control sequence, and stability verification and compensation need to be performed. The target control sequence is then generated and issued for execution, including: The optimized power instruction set is converted into the corresponding duty cycle using a space vector pulse width modulation algorithm to obtain the initial control sequence. Simulate the operating state of the initial control sequence under the boundary condition of the battery's maximum charge-discharge rate, and extract transient voltage fluctuation characteristics; The transient voltage fluctuation characteristics are input into a preset phase trajectory analysis model to obtain the sequence stability value; If the sequence stability value is greater than the preset stability threshold, then dead time compensation is performed on the initial control sequence to obtain the target control sequence; The target control sequence is output to the execution layer of the energy storage device to drive the power conversion unit to operate.

[0050] First, it should be noted that the Space Vector Pulse Width Modulation (SVPWM) algorithm is the core control algorithm of the power conversion unit (PCS). This algorithm first converts the three-phase power command from the optimized power command set into direct-axis and quadrature-axis voltage reference values ​​in the dq rotating coordinate system using Clark / Park transformation. Then, it obtains the desired voltage vector through voltage space vector synthesis. Subsequently, it determines the sector where the desired voltage vector is located, calculates its duration of action on adjacent basic voltage vectors, and finally combines it with a triangular carrier wave to generate the PWM dynamic signals for the six switches of the inverter, achieving precise conversion of power command to switch conduction timing. During the conversion, the system uses the target power, target voltage, and target current from the optimized power command set as input parameters. Through coordinate transformation (Clark and Park transformations), it converts the command in the three-phase stationary coordinate system into DC values ​​in the rotating coordinate system, calculates the conduction time ratio of the upper and lower bridge arm power transistors, i.e., the duty cycle signal. The duty cycle value ranges from 0 to 1, directly determining the output amplitude and frequency of the inverter. The duty cycle sequence generated at each time step is the initial control sequence. For example, the optimized power command is a 357kW discharge, corresponding to a DC bus voltage of 719V. After SVPWM conversion, the duty cycle of the output initial control sequence is [0.42, 0.43, 0.42, 0.44, ...]. To verify the reliability of the initial control sequence, its operating state under the boundary condition of the battery's maximum charge / discharge rate needs to be simulated in a virtual simulation environment. This condition is defined as the extreme operating state where the battery rate reaches 1.2C and the current is close to the rated limit. The simulation process collects the cell voltage, bus voltage, and output current waveforms in real time at a sampling frequency of 10kHz, focusing on extracting transient voltage fluctuation characteristics, including voltage peak value, voltage valley value, peak-to-peak fluctuation value, and fluctuation recovery time. For example, in the boundary condition simulation, the peak-to-peak value of the transient voltage fluctuation driven by the initial control sequence is 18.5V, and the fluctuation recovery time is 12ms. This feature directly reflects the stability of the control sequence under extreme conditions.

[0051] The phase trajectory analysis model is a dynamic stability assessment model based on the state space of a second-order nonlinear system. It uses the differential value (dv / dt) of the DC bus voltage fluctuation of the energy storage system as the horizontal axis and the actual voltage fluctuation value (Δv) as the vertical axis. Phase trajectory curves are constructed by real-time acquisition of voltage fluctuation data under boundary conditions. The convergence, closure, and enclosing area of ​​the curve are used as the criteria for stability judgment. The degree of trajectory convergence is the sequence stability value, ranging from 0 to 1, with values ​​closer to 1 indicating better stability. The model calculates the ratio of the actual trajectory enclosing area to the preset unstable critical trajectory area, and obtains the stability value after normalization. This objectively identifies different dynamic states such as oscillation, divergence, and decay. For example, after inputting the aforementioned transient voltage fluctuation characteristics into the phase trajectory model, the calculated actual trajectory enclosing area is 18% of the critical area. The ratio calculation yields a sequence stability value of 0.82.

[0052] The preset stability threshold is set to 0.9, a value determined through extensive PCS hardware-in-the-loop testing. When the stability value is ≥0.9, the control system exhibits no significant oscillations, and voltage and current fluctuations are ≤5%, meeting the grid-connected operation standards. When the stability value is <0.9, it indicates slight oscillations or dynamic response deviations, requiring dead-time compensation. Dead-time compensation is used to offset waveform distortion and voltage deviations caused by power transistor switching delays, with a compensation amount set to 2.5μs (determined based on IGBT switching characteristics and drive circuit delay). The compensation method involves superimposing a compensation increment into the duty cycle of the initial control sequence, causing the switching action to be advanced or delayed, thus suppressing transient fluctuations. For example, if the original stability value of 0.82 is lower than the threshold, after 2.5μs dead-time compensation, the duty cycle sequence is corrected to [0.418, 0.432, 0.423, 0.441, ...].

[0053] Finally, the target control sequence is output as a high-speed pulse signal to the execution layer of the energy storage device. The execution layer includes hardware units such as the PCS driver board, IGBT power modules, and filter circuits. The target control sequence directly drives the power conversion unit to achieve precise charge and discharge control, outputting a stable, oscillation-free, and overshoot-free power waveform. This ensures that the energy storage system operates strictly according to the optimized instructions, while guaranteeing long-term safety and efficiency. For example, after the target control sequence is issued, the actual output power of the PCS is 356.8kW, with a deviation of only 0.2kW from the command value, and voltage fluctuations are controlled within 3.2V.

[0054] After generating the target control sequence and issuing it for execution in step S16, the method further includes obtaining the electrical parameters fed back by the execution layer, performing coordinate transformation and error calculation, constructing a response error vector, updating the parameters of the constraint model based on the response error vector, and obtaining an enhanced constraint model, including: The instantaneous values ​​of the three-phase current and the DC bus voltage are obtained from the feedback of the execution layer using high-precision sensors. The instantaneous values ​​of the three-phase currents are subjected to Clarke transform and Park transform to obtain the direct-axis current components and quadrature-axis current components in the rotating coordinate system. Based on the optimized power instruction set and the instantaneous value of the DC bus voltage, calculate the direct-axis current reference value and the quadrature-axis current reference value; The differences between the direct-axis current component and the direct-axis current reference value, and between the quadrature-axis current component and the quadrature-axis current reference value are calculated respectively to obtain the direct-axis current error and the quadrature-axis current error; Vector synthesis of the direct-axis current error and the quadrature-axis current error is used to construct a response error vector; The parameters of the constraint model are updated based on the response error vector to obtain the enhanced constraint model.

[0055] First, it should be noted that the electrical parameter acquisition for the execution layer relies on high-precision sensors deployed at the output of the power conversion unit (PCS) and the DC bus to ensure the real-time performance and accuracy of the data feedback: the instantaneous values ​​of the three-phase current are acquired through closed-loop Hall current sensors with a range of 0-800A, a measurement accuracy of ±0.1%FS, and a sampling frequency of 10kHz synchronized with the control sequence (which can accurately capture transient changes in current and avoid phase lag); the instantaneous values ​​of the DC bus voltage are acquired through differential voltage sensors with a range of 0-1000V, a measurement accuracy of ±0.05%FS, and a sampling frequency of 10kHz, which synchronously records the sampling timestamp (accurate to the microsecond level) and is strictly aligned with the control sequence time axis.

[0056] After acquiring feedback data in real time through sensors, press "timestamp - three-phase current (, , DC bus voltage The dimensions of ")" are summarized, for example, the instantaneous value of the three-phase current at a certain sampling moment is I. a =456.2A、 =-228.1A、 =-228.1A, instantaneous value of DC bus voltage =718.5V, which directly reflects the actual operating status of the execution layer.

[0057] The core purpose of the Clark transform on the instantaneous values ​​of three-phase current is to convert the alternating current in a stationary three-phase coordinate system into a direct current in a stationary two-phase coordinate system, thereby eliminating three-phase coupling. The Clark transform uses the constant amplitude transformation formula: , For example, substituting the above three-phase current values ​​into the calculation: It is 456.2A. Given 0A, the current component in the αβ coordinate system is obtained as [456.2A, where 0A is].

[0058] Subsequently, the Park transformation is performed to convert the current component in the αβ coordinate system into a component in the dq coordinate system (direct-axis to quadrature-axis coordinate system) that rotates synchronously with the grid voltage, thereby achieving decoupled control of the current—the direct-axis current component ( The corresponding active power related current, quadrature axis current component ( The corresponding reactive power-related current. The Parker transformation formula is: , .in The real-time grid voltage phase angle is obtained in real-time by a phase-locked loop (PLL) (detection accuracy ±0.1°). For example, currently... =0° (zero-crossing point of grid voltage), substitute into , Value calculated It is 456.2A. It is 0A.

[0059] Calculate the direct-axis current reference value based on the optimized power command set and the instantaneous value of the DC bus voltage. ) and cross-axis current reference value ( ): The optimized power instruction set includes active power instructions ( ) and reactive power command ( The active power directly determines the direct-axis current reference value, while the reactive power determines the quadrature-axis current reference value. The calculation formula is derived based on the power balance principle of the PWM inverter. , For example, optimizing the active power of the power instruction set. =357kW, reactive power =0kVar (pure active power output), substitute with DC bus voltage =718.5V Calculation: It is 333.5A. It is 0A.

[0060] The difference between the actual current component and the reference value is calculated separately to obtain the direct-axis current error ( ). ) and cross-axis current error ( The error value directly quantifies the degree of deviation between the actual output of the execution layer and the instruction requirements: , For example, substituting the above numerical calculations... It is 122.7A. The value is 0A, meaning that the direct-axis current has a positive deviation, while the quadrature-axis current has no deviation.

[0061] Finally, the direct-axis current error and the quadrature-axis current error are vector-synthesized to construct a response error vector. The synthesis method is based on the vector characteristics of the dq coordinate system, using the direct-axis error as the real-axis component and the quadrature-axis error as the imaginary-axis component. The magnitude and direction of the error are fully characterized by the vector magnitude and phase angle. For example, substituting... =122.7A =0A calculation yields a response error vector magnitude of 122.7A and a phase angle of 0°.

[0062] After constructing the response error vector, the parameters of the constraint model need to be updated based on the response error vector to obtain the enhanced constraint model, including: Extract the magnitude deviation and phase deviation of the response error vector; Calculate the recursive gain matrix based on the preset system noise covariance matrix and measurement noise covariance matrix; The amplitude deviation and phase deviation are linearly transformed by the recursive gain matrix to obtain the state correction value; The battery temperature range and the current state of charge range are used to match the basic state transition matrix; The modified state transition matrix is ​​obtained by weighted summing of the nonlinear coupling elements in the basic state transition matrix using the state correction values. The parameters of the constraint model are updated based on the modified state transition matrix to obtain the enhanced constraint model.

[0063] First, it should be noted that the response error vector contains both amplitude and phase information, which is the core basis for updating the constraint model parameters. The amplitude deviation is the difference between the magnitude of the response error vector and the preset error tolerance. The preset error tolerance is set at 5% of the rated current based on the control accuracy requirements of the energy storage system (for example, with an 800A rated current, the tolerance is 40A), used to quantify the severity of the error. The phase deviation is the difference between the actual phase angle of the response error vector and the ideal phase angle (0°, i.e., the error exists only on the direct axis), reflecting the spatial distribution characteristics of the error. For example, if the magnitude of the response error vector obtained above is 122.7A and the phase angle is 0°, then the amplitude deviation is 82.7A and the phase deviation is 0°, indicating that the current error is mainly due to excessive amplitude on the direct axis, with no phase shift.

[0064] The preset system noise covariance matrix and measurement noise covariance matrix are fixed matrices statistically calibrated based on historical operating data, used to characterize internal disturbances and sensor measurement uncertainties. The system noise covariance matrix focuses on internal disturbances such as battery internal resistance fluctuations and temperature drift, and is determined to be [[0.02, 0], [0, 0.01]] after statistical analysis of 1000 sets of steady-state operating data; the measurement noise covariance matrix addresses the measurement errors of current and voltage sensors, and is calibrated to be [[0.05, 0], [0, 0.03]]. Both types of matrices together provide uncertainty boundaries for recursive gain calculation.

[0065] The recursive gain matrix is ​​derived through the gain calculation logic of Kalman filtering. Its core function is to balance the sensitivity and stability of error correction. The larger the value of this matrix, the stronger the correction force of the model to the current error. For example, the recursive gain matrix [[0.31, 0], [0, 0.26]] is calculated. The state correction value is obtained by linearly transforming the amplitude deviation and phase deviation through the recursive gain matrix, which essentially maps the error information into the correction amount of the model parameters. For example, with the recursive gain matrix [[0.31, 0], [0, 0.26]], amplitude deviation 82.7A, and phase deviation 0°, the state correction value [25.637A, 0°]ᵀ is calculated.

[0066] The basic state transition matrix is ​​a core parameter of the constraint model, used to describe the evolution of the energy storage system's state over time. It has a 2×2 dimension (corresponding to the direct and quadrature axes), and its elements are strongly correlated with the battery temperature range and the current state of charge (SOC) range—different temperature and SOC combinations correspond to different internal resistances and polarization characteristics, requiring differentiated transition matrices. The system has multiple built-in basic state transition matrices, matched via a lookup table: for the temperature range of 0~10℃ and SOC of 60%~70%, the basic state transition matrix is ​​preset to [[0.98, 0.01], [0.005, 0.97]]. This matrix has been calibrated through high and low temperature experiments and cyclic aging tests, accurately characterizing the system's dynamic characteristics under corresponding operating conditions.

[0067] The state correction values ​​are applied to the nonlinear coupling elements of the basic state transition matrix using a weighted summation method. The corrected coupling elements are obtained by multiplying the sum of the normalized correction coefficients plus 1 with the original coupling elements. The weighting coefficients are the normalized results of the state correction values ​​(the amplitude correction coefficient is obtained by dividing the amplitude correction by the rated current, and the phase correction coefficient is obtained by dividing the phase correction by 90°), used to avoid over-correction that could lead to model instability. For example, the amplitude correction coefficient is 0.032, the phase correction coefficient is 0, and the nonlinear coupling elements (off-diagonal elements) in the basic state transition matrix are 0.01 and 0.005, respectively. After correction, they are 0.0103 and 0.0052, respectively. The diagonal elements remain unchanged (to ensure the stability of the basic trend), and the corrected state transition matrix is ​​finally obtained as [[0.98, 0.0103], [0.0052, 0.97]].

[0068] Finally, the modified state transition matrix replaces the original basic state transition matrix in the constraint model, completing the model parameter update and obtaining the enhanced constraint model. The updated model can specifically compensate for the current error: by adjusting the coupling elements to enhance the control sensitivity of the direct-axis current, the error correction in the next sampling period is more accurate. For example, in subsequent operation, the prediction deviation of the direct-axis current by the enhanced constraint model will decrease from 122.7A to within 35A, and the amplitude deviation will be controlled below the preset tolerance of 40A, significantly improving the control accuracy and response stability of the energy storage system.

[0069] In summary, this invention discloses an energy efficiency optimization control method for energy storage power stations. Through a two-layer processing strategy combining raw data preprocessing, wavelet decomposition and reconstruction, and recursive estimation of the constraint model, it achieves accurate extraction of low-frequency energy trends and effective filtering of high-frequency disturbances. Simultaneously, when the deviation exceeds a threshold, it integrates smoothed high-frequency information to correct the energy sequence, effectively solving the problem of distorted energy feature extraction under complex operating conditions. By constructing a full-process energy efficiency control system, it optimizes power commands layer by layer based on multiple constraints such as battery temperature, charge / discharge rate, and state of charge. Then, it generates control sequences through extreme operating condition stability verification and compensation, significantly improving the executability of power commands and the stability of the energy storage system. By constructing an error vector through feedback of electrical parameters at the execution layer, and iteratively updating the constraint model parameters based on error characteristics to obtain an enhanced constraint model, it forms a closed-loop linkage mechanism between command execution and model optimization, achieving dynamic improvement and long-term optimization of energy efficiency control accuracy.

[0070] Reference Figure 2 The second embodiment of the present invention provides an energy efficiency optimization control system for an energy storage power station, comprising: The data acquisition and decomposition module is used to acquire the raw data of the energy storage system according to the preset current sampling frequency and voltage sampling frequency, extract the low frequency coefficients after preprocessing, and then obtain the low frequency profile through the signal reconstruction method. An energy trend construction module is used to input the low-frequency profile into a preset constraint model to obtain an initial state sequence, and then construct an energy trend sequence after applying Kalman filtering and fluctuation constraints to the initial state sequence. An energy sequence correction module is used to acquire historical power curves and calculate the deviation from the energy trend sequence. If the deviation continuously exceeds a preset deviation threshold, high-frequency disturbances are separated and smoothed to obtain smoothed high-frequency information. The smoothed high-frequency information and the energy trend sequence are then weighted and fused to generate a corrected energy sequence. The initial instruction generation module is used to obtain the battery temperature range at the current moment, perform temperature adaptation conversion and charge / discharge capacity limitation on the corrected energy sequence, eliminate the part that exceeds the actual operating capacity of the battery, obtain the real-time power demand, and match the real-time power demand in the preset demand mapping matrix to obtain the initial power instruction set. The power command optimization module is used to obtain the current state of charge range, calculate the matching degree with the initial power command set, and if the matching degree is lower than the preset matching threshold, adjust the amplitude and compensate the load of the initial power command set to obtain an optimized power command set. The module below the control sequence is used to convert the optimized power instruction set into the initial control sequence, perform stability verification and compensation, generate the target control sequence, and issue it for execution. The feedback error construction module is used to obtain the electrical parameters fed back by the execution layer, perform coordinate transformation and error calculation, and construct a response error vector. The model parameter update module is used to update the parameters of the constraint model according to the response error vector to obtain the enhanced constraint model.

[0071] It should be noted that the energy storage power station energy efficiency optimization control system provided in this embodiment of the invention is used to execute all the process steps of the energy storage power station energy efficiency optimization control method in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0072] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0073] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. The energy efficiency optimization control method for an energy storage power station, characterized in that, include: The raw data of the energy storage system is collected according to the preset current sampling frequency and voltage sampling frequency. After preprocessing, the low-frequency coefficients are extracted, and then the low-frequency profile is obtained through signal reconstruction method. The low-frequency profile is input into a preset constraint model to obtain an initial state sequence. After applying Kalman filtering and fluctuation constraints to the initial state sequence, an energy trend sequence is constructed. Historical power curves are acquired and their deviation from the energy trend sequence is calculated. If the deviation continuously exceeds a preset deviation threshold, high-frequency disturbances are separated and smoothed to obtain smoothed high-frequency information. The smoothed high-frequency information and the energy trend sequence are then weighted and fused to generate a corrected energy sequence. Obtain the current battery temperature range, perform temperature adaptation calculation and charge / discharge capacity limitation on the corrected energy sequence, eliminate the part that exceeds the actual operating capacity of the battery, obtain the real-time power demand, and match the real-time power demand in the preset demand mapping matrix to obtain the initial power instruction set; Obtain the current state of charge range, calculate the matching degree with the initial power command set, and if the matching degree is lower than the preset matching threshold, adjust the amplitude and compensate the load of the initial power command set to obtain an optimized power command set. The optimized power instruction set is converted into an initial control sequence, and stability verification and compensation are performed. The target control sequence is then generated and executed.

2. The energy efficiency optimization control method for energy storage power stations according to claim 1, characterized in that, The process of acquiring raw data from the energy storage system based on preset current and voltage sampling frequencies, preprocessing the data to extract low-frequency coefficients, and then obtaining the low-frequency profile through signal reconstruction includes: The raw data of the energy storage system is collected according to the preset current sampling frequency and voltage sampling frequency, and the raw data is time-aligned to obtain a synchronous operation sequence; The synchronized running sequence is decomposed into low-frequency coefficients by using preset wavelet parameters at multiple scales. If the absolute value of the low-frequency coefficient is greater than the preset low-frequency threshold, then the low-frequency coefficient is subjected to mean filtering to obtain the target approximation coefficient; if the absolute value of the low-frequency coefficient is not greater than the preset low-frequency threshold, then it is directly used as the target approximation coefficient. The target approximation coefficients are subjected to inverse wavelet transform to obtain the low-frequency profile.

3. The energy efficiency optimization control method for energy storage power stations according to claim 1, characterized in that, The step of inputting the low-frequency profile into a preset constraint model to obtain an initial state sequence, and then constructing an energy trend sequence after applying Kalman filtering and fluctuation constraints to the initial state sequence includes: The low-frequency profile is input into a preset constraint model to obtain an initial state sequence; The Kalman filter algorithm is applied to the initial state sequence to obtain the intermediate trend vector; If the fluctuation amplitude of the intermediate trend vector is greater than the preset fluctuation threshold, then the intermediate trend vector is subjected to a moving average filter to obtain a long-term energy representation; if the fluctuation amplitude of the intermediate trend vector is not greater than the preset fluctuation threshold, then it is directly used as the long-term energy representation. By integrating the long-term energy representations in chronological order along the timeline, an energy trend sequence is obtained.

4. The energy efficiency optimization control method for energy storage power stations according to claim 1, characterized in that, The process of acquiring historical power curves and calculating their deviation from the energy trend sequence, and if the deviation continuously exceeds a preset deviation threshold, involves separating and smoothing high-frequency disturbances to obtain smoothed high-frequency information, and then weightedly fusing the smoothed high-frequency information with the energy trend sequence to generate a corrected energy sequence, including: Obtain the historical power curve, calculate the amplitude difference between the historical power curve and the energy trend sequence within the same time window, and obtain the deviation sequence; If the deviation sequence is greater than the preset deviation threshold within a preset continuous time window, the corresponding original data is extracted and the corresponding energy trend sequence is subtracted to obtain the corresponding high-frequency perturbation; the energy spectral density of the high-frequency perturbation is calculated by Fourier transform, and the energy distribution range is extracted. The high-frequency disturbances corresponding to the energy distribution range are smoothed by a preset filter to obtain smooth high-frequency information; The smoothed high-frequency information and the energy trend sequence are weighted and fused in the time domain to generate a corrected energy sequence.

5. The energy efficiency optimization control method for energy storage power stations according to claim 1, characterized in that, The process involves obtaining the current battery temperature range, performing temperature-adaptive calculations and charge / discharge capacity limits on the corrected energy sequence, eliminating portions exceeding the battery's actual operating capacity, obtaining the real-time power demand, and matching this real-time power demand against a preset demand mapping matrix to obtain an initial power instruction set, including: Obtain the current battery temperature range and look up the corresponding temperature decay factor in the preset temperature mapping table; The amplitude of the corrected energy sequence is calculated based on the temperature decay factor to obtain the compensated energy sequence; Obtain the maximum charge / discharge rate of the current battery, and construct a charge / discharge rate limit boundary for the compensated energy sequence; If the value of the compensation energy sequence exceeds the charge / discharge rate limit boundary, it is truncated to obtain the real-time power requirement; The initial power instruction set is obtained by matching the real-time power demand in a preset demand mapping matrix.

6. The energy efficiency optimization control method for energy storage power stations according to claim 1, characterized in that, The process of obtaining the current state of charge range, calculating the matching degree with the initial power command set, and if the matching degree is lower than a preset matching threshold, adjusting the amplitude and compensating the load of the initial power command set to obtain an optimized power command set includes: Obtain the current state of charge range of the energy storage system, and calculate the matching degree between the initial power command set and the current state of charge range through a preset matching calculation model; If the matching degree is lower than the preset matching threshold, the initial power instruction set is initially corrected according to the preset instruction adjustment range to obtain the corrected power instruction set; if the matching degree is higher than the preset matching threshold, it is directly used as the corrected power instruction set. The instantaneous fluctuation characteristics of the load are extracted from the real-time operation data of the energy storage system and the fluctuation intensity is quantified. The corrected power instruction set is weighted and compensated according to the fluctuation intensity by matching dynamic weights to generate a fluctuation compensation instruction set. The fluctuation compensation instruction set is subjected to boundary verification by a preset multi-constraint filter to determine the constraint satisfaction sequence; The modified power instruction set is fine-tuned based on the constraint satisfaction sequence to obtain the optimized power instruction set.

7. The energy efficiency optimization control method for energy storage power stations according to claim 1, characterized in that, The process of converting the optimized power instruction set into an initial control sequence, performing stability verification and compensation, generating a target control sequence, and issuing it for execution includes: The optimized power instruction set is converted into the corresponding duty cycle using a space vector pulse width modulation algorithm to obtain the initial control sequence. Simulate the operating state of the initial control sequence under the boundary condition of the battery's maximum charge-discharge rate, and extract transient voltage fluctuation characteristics; The transient voltage fluctuation characteristics are input into a preset phase trajectory analysis model to obtain the sequence stability value; If the sequence stability value is greater than the preset stability threshold, then dead time compensation is performed on the initial control sequence to obtain the target control sequence; The target control sequence is output to the execution layer of the energy storage device to drive the power conversion unit to operate.

8. The energy efficiency optimization control method for energy storage power stations according to claim 7, characterized in that, After generating the target control sequence and issuing it for execution, the process further includes obtaining the electrical parameters fed back by the execution layer, performing coordinate transformation and error calculation, constructing a response error vector, and updating the parameters of the constraint model based on the response error vector to obtain an enhanced constraint model, including: The instantaneous values ​​of the three-phase current and the DC bus voltage are obtained from the feedback of the execution layer using high-precision sensors. The instantaneous values ​​of the three-phase currents are subjected to Clarke transform and Park transform to obtain the direct-axis current components and quadrature-axis current components in the rotating coordinate system. Based on the optimized power instruction set and the instantaneous value of the DC bus voltage, calculate the direct-axis current reference value and the quadrature-axis current reference value; The differences between the direct-axis current component and the direct-axis current reference value, and between the quadrature-axis current component and the quadrature-axis current reference value are calculated respectively to obtain the direct-axis current error and the quadrature-axis current error; Vector synthesis of the direct-axis current error and the quadrature-axis current error is used to construct a response error vector; The parameters of the constraint model are updated based on the response error vector to obtain the enhanced constraint model.

9. The energy efficiency optimization control method for energy storage power stations according to claim 8, characterized in that, The step of updating the parameters of the constraint model based on the response error vector to obtain the enhanced constraint model includes: Extract the magnitude deviation and phase deviation of the response error vector; Calculate the recursive gain matrix based on the preset system noise covariance matrix and measurement noise covariance matrix; The amplitude deviation and phase deviation are linearly transformed by the recursive gain matrix to obtain the state correction value; The battery temperature range and the current state of charge range are used to match the basic state transition matrix; The modified state transition matrix is ​​obtained by weighted summing of the nonlinear coupling elements in the basic state transition matrix using the state correction values. The parameters of the constraint model are updated based on the modified state transition matrix to obtain the enhanced constraint model.

10. The energy efficiency optimization control system for an energy storage power station, characterized in that, include: The data acquisition and decomposition module is used to acquire the raw data of the energy storage system according to the preset current sampling frequency and voltage sampling frequency, extract the low frequency coefficients after preprocessing, and then obtain the low frequency profile through the signal reconstruction method. An energy trend construction module is used to input the low-frequency profile into a preset constraint model to obtain an initial state sequence, and then construct an energy trend sequence after applying Kalman filtering and fluctuation constraints to the initial state sequence. An energy sequence correction module is used to acquire historical power curves and calculate the deviation from the energy trend sequence. If the deviation continuously exceeds a preset deviation threshold, high-frequency disturbances are separated and smoothed to obtain smoothed high-frequency information. The smoothed high-frequency information and the energy trend sequence are then weighted and fused to generate a corrected energy sequence. The initial instruction generation module is used to obtain the battery temperature range at the current moment, perform temperature adaptation conversion and charge / discharge capacity limitation on the corrected energy sequence, eliminate the part that exceeds the actual operating capacity of the battery, obtain the real-time power demand, and match the real-time power demand in the preset demand mapping matrix to obtain the initial power instruction set. The power command optimization module is used to obtain the current state of charge range, calculate the matching degree with the initial power command set, and if the matching degree is lower than the preset matching threshold, adjust the amplitude and compensate the load of the initial power command set to obtain an optimized power command set. The control sequence delivery module is used to convert the optimized power instruction set into an initial control sequence, perform stability verification and compensation, generate the target control sequence, and deliver it for execution.

Images