A wind power prediction-based energy storage dynamic allocation system

Abstract

The application provides a kind of energy storage dynamic deployment system based on wind power prediction, by obtaining real-time and historical power data of wind farm and forming standardized time series, combined with trend extraction structure to generate long-term change curve and local disturbance signal;Energy analysis mechanism is used to process the disturbance signal in segments, extract instantaneous energy features;According to the instantaneous energy features and set conditions to generate event trigger state;Combining trend changes and event trigger results to form intermediate power instructions, realize dynamic deployment of energy storage power;In the process of energy storage device operation, parameters such as state of charge, current, voltage and temperature are collected, and control parameters are adjusted in real time according to these parameters;Finally, the intermediate power instruction is amplitude, change rate and state of charge constraint, and the energy storage deployment control instruction is output to drive the energy storage device to perform charge and discharge behavior, so as to form the energy storage deployment process of wind power trend driving and disturbance response collaborative operation.

Description

Technical Field

[0001] This invention relates to the field of new energy power generation and energy storage control technology, and in particular to a dynamic energy storage allocation system based on wind power prediction. Background Technology

[0002] With the rapid growth of new energy installed capacity, the volatility and uncertainty of wind power output have placed higher demands on the safe and stable operation of the power grid. Currently, wind farms generally rely on energy storage devices for power smoothing and frequency regulation / peak shaving. However, existing energy storage dispatching methods are mostly based on simple threshold triggering or fixed strategy control, making it difficult to simultaneously take into account the trend changes and rapid disturbance characteristics of wind power. In practical applications, on the one hand, traditional trend extraction algorithms are not accurate enough in characterizing the low- and medium-frequency changes of wind power sequences and are easily affected by abrupt changes, noise, and sampling anomalies, resulting in a lack of foresight in energy storage dispatching. On the other hand, disturbance analysis methods based on single-point energy or fixed windows are difficult to accurately identify the energy structure of wind power fluctuations, and are prone to false triggering or missed triggering during event triggering, making the energy storage response unable to match the real dynamics of wind power disturbances.

[0003] Furthermore, traditional energy storage dispatch strategies generally lack decoupling between trend-driven and event-driven designs. Energy storage power commands are prone to drastic jumps under abrupt changes, increasing the stress load on energy storage devices. Simultaneously, control parameters are mostly statically set, failing to consider dynamic changes in operating parameters such as state of charge, voltage, current, and temperature, leading to a decline in control performance as operating conditions change. Moreover, existing command constraint mechanisms often only perform simple limiting, lacking coordinated constraints on ramp rate, descent rate, and charge boundary, easily resulting in command overshoot or limit exceedance risks.

[0004] Therefore, how to provide a dynamic energy storage allocation system based on wind power prediction is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a dynamic energy storage allocation system based on wind power prediction. This system is developed by constructing a dual-channel feature extraction mechanism for trends and disturbances, an event-triggered model driven by instantaneous energy, a power command generation strategy that integrates trends and events, a parameter update method that adapts to the operating state, and a command correction process that coordinates multiple constraints.

[0006] According to an embodiment of the present invention, an energy storage dynamic dispatch system based on wind power prediction includes the following steps: The data processing module acquires real-time and historical power data from the wind farm and converts the power data into a standardized time series. The trend extraction module receives standardized time series data and decomposes the power data using an H-V adaptive two-layer smoothing model to generate trend baseline signals and residual signals. The energy analysis module receives the residual signal and extracts instantaneous energy indices using the Teager–Kaiser energy operator. The event judgment module receives instantaneous energy indicators, determines whether the triggering conditions are met based on the set energy threshold and hysteresis band parameters, and generates an event trigger signal. The power command generation module receives the trend baseline signal and the event trigger signal, and generates intermediate power control commands based on the trend baseline signal and the event trigger status. The energy storage system module performs energy storage power control operations and outputs operating status parameters in real time. The parameter adaptive module receives the operating status parameters output by the energy storage system module and adjusts the control parameters in the power command generation module according to the operating status. The control command constraint module receives intermediate power control commands and performs constraint processing on the intermediate power control commands according to the power limitation conditions of the energy storage system module, generating energy storage dispatch control commands.

[0007] Optionally, the trend extraction module includes: using a standardized time series as the input signal of an H-V adaptive two-layer smoothing model, wherein the H-V adaptive two-layer smoothing model takes a standardized power series as input and outputs a trend baseline signal expressing long-term trend changes and a residual signal expressing local perturbation differences.

[0008] Optionally, the energy analysis module includes: The residual signals are written into the residual sequence buffer in chronological order; an independent residual slot is set at each time index position in the residual sequence buffer; According to the preset segment length L, the residual sequence buffer is divided into a group of continuous, fixed-length and non-overlapping residual segments starting from the starting index position; each residual segment contains L residual slots, and a segment energy buffer, segment direction buffer, segment energy concentration buffer and segment stability buffer corresponding to the residual segment are established in memory. When processing any residual segment, each index position is processed sequentially according to the time order within the residual segment. When processing the current index position, the following operations are performed: Read the residual value at the current index position from the current residual segment slot as the current residual value; Read the residual value of the previous index position from the current residual segment slot as the forward residual value; when the position is at the beginning of the segment, read the residual value of the residual slot before the starting index position of the residual segment from the residual sequence buffer as the forward residual compensation value for that position. Read the residual value of the next index position from the current residual segment slot as the backward residual value; When the position is at the end of a segment, the residual value of the residual slot after the end index position of the residual segment is read from the residual sequence buffer as the backward residual compensation value for that position. Multiply the current residual value by the current residual value and write the result as the current energy base value; Multiply the forward residual value by the backward residual value, and write the result as the neighborhood coupling energy value; Subtract the neighborhood coupled energy value from the current energy base value, and write the difference into the current index position slot of the segmented energy buffer; After writing to all L positions of the segmented energy buffer, directional analysis and directional consistency identification are performed on all segmented instantaneous energy within the residual segment in chronological order, specifically including: Within the residual segment, the difference operation is performed sequentially on the instantaneous energy of every two adjacent segments, and the sign of the difference is written into the segment direction buffer. In the segmented direction buffer, starting from the starting index position, search for adjacent intervals with the same symbol, and record all consecutive intervals with the same symbol as the direction-consistent area; Write the number of indexes in each direction-consistent region to the direction-consistent region length field, and write the direction-consistent region length field to the segmented energy concentration buffer. After performing direction consistency identification, energy concentration calculation is performed on the residual segments, specifically including: For each directional consistent region, sum the instantaneous energy of all segments within it sequentially, divide by the length of the directional consistent region, and write the quotient into the corresponding slot of the segmented energy concentration buffer. Find the maximum energy concentration value in the all-directional consistent region and write the maximum energy concentration value into the segmented energy concentration maximum value slot; After obtaining the maximum value of the segmented energy concentration, cross-segment stability identification is performed, which specifically includes: Perform a difference operation on the maximum value of segmented energy concentration of the residual segment and the previous residual segment, and write the difference into the segmented stability buffer. When the absolute value of the difference is less than the preset stability threshold S, the residual segment is marked as a stable segment; when the absolute value is greater than the stability threshold S, the residual segment is marked as a sudden change segment. Write the maximum segment energy, maximum segment energy concentration, and segment stability flag of each residual segment into the energy structure output buffer according to the time coverage range corresponding to the residual segment. The three fields in the energy structure output buffer are combined into the final instantaneous energy index.

[0009] Optionally, the event determination module includes: The instantaneous energy index at each time index position is written sequentially into the input buffer structure of the event judgment module; Configure energy threshold parameters With hysteresis band parameters The event status register is set to record whether the event is in a "triggered state" or a "non-triggered state". The register is initialized to the "non-triggered state". Construct a sliding energy window of fixed length L for the current index position. The sliding energy window contains The energy amplitude and energy symbol value corresponding to each consecutive time index position are used as the local behavior region; When the event status register is in the "not triggered" state, the following three decisions are made at the current time index position: Criterion 1: The energy amplitude at the current index position satisfies... ; Criterion 2: Exceeding within the sliding energy window An energy direction value with the same sign as the current index; Criterion 3: The average value of all energy amplitudes within the sliding energy window is greater than... ; If all three conditions are met, the event state register is updated to "trigger state", and an event trigger signal value "1" is generated at the current time index position and written to the event trigger signal buffer. When the event status register is in "triggered state", the following two decisions are made at the current index position: The energy amplitude at the current index position satisfies ; The average energy amplitude within the sliding energy window is less than ; If both conditions are met, the event status register is updated to "not triggered" and an event trigger signal value of "0" is generated at the current time index position and written to the event trigger signal buffer. When the event status register is in "triggered state" and the cancellation condition is not met, the event trigger signal value "1" is continuously generated at the current index position and written to the event trigger signal buffer until the cancellation condition is met. After traversing all time index positions, the event trigger signals arranged in chronological order in the event trigger signal buffer are output.

[0010] Optionally, the power command generation module includes: The trend baseline signal is written to the trend baseline buffer in time index order. Each slot in the trend baseline buffer stores only the trend baseline value corresponding to the current time index. The event trigger signals are written to the event trigger buffer in time index order. Each slot in the event trigger buffer stores a binary result indicating whether the current time index is in the trigger state. The trigger state is "1" and the non-trigger state is "0". Three types of buffer slots are established within the power command generation module: trend power slot, event modulation slot, and intermediate command slot; Before the traversal begins, set the initial value of all event modulation slots to zero and the initial value of all intermediate instruction slots to zero. When processing any time index position, the trend baseline value corresponding to the time index is read from the trend baseline buffer, the trend baseline value is multiplied by a preset trend ratio coefficient, and the result is written into the trend power slot to form the trend power value of the time index position. At the same time index position, the trigger signal is read from the event trigger buffer. When the trigger signal is "1", the trend power value is read from the trend power slot. The trend power value is multiplied by the preset event gain coefficient. The result is written into the event modulation slot as the event modulation start value at the time index position. When the event trigger signal at N consecutive time index positions is "1", the event modulation value is read from the event modulation slot at the previous time index position. The value is then subjected to a linear weighted operation with the event modulation start value. The weighting ratio uses a preset event ramp-up coefficient. The weighted operation result is written to the event modulation slot at the current time index position, so that the event modulation value is written in a continuous ramp-up form along the time axis. When the trigger signal at the current time index position is "0" and the trigger signal at the previous time index position is "1", the event modulation value is read from the previous time index position, the event modulation value is multiplied by the preset attenuation coefficient, the result is written to the current event modulation slot, and the same attenuation operation is performed on the event modulation slot at each subsequent time index position until the event modulation value is equal to zero. At each time index position, the trend power value and event modulation value are read from the trend power slot and the event modulation slot respectively. The trend power value and the event modulation value are linearly summed according to the proportional coefficient. The linear summation result is written into the intermediate instruction slot as the intermediate power control instruction for the current time index position. Once all trend power slots, event modulation slots, and intermediate instruction slots at all time index positions have been written, all instructions in the intermediate instruction slot will be output in chronological order.

[0011] Optionally, the energy storage system module includes: sequentially writing and reading energy storage allocation control commands on the time axis; using the commands and DC bus voltage together to obtain the current reference value; driving the converter unit to output the corresponding charging and discharging current with current closed-loop control; periodically collecting battery cluster operating parameters using current, voltage, and temperature sensors and performing current integration to update the state of charge and terminal voltage consistency correction; and packaging the state of charge, current, voltage, and temperature into an operating status parameter frame according to the time index and outputting it.

[0012] Optionally, the parameter adaptive module includes: writing the running status parameter frame into the parameter input buffer according to the time index and sequentially completing the safe range judgment of the state of charge, temperature and power to generate corresponding tags; constructing and fusing the state response matrix to obtain the adjustment command level field; retrieving three types of adjustment coefficients, namely trend ratio, event gain and event ramp rate, from the parameter mapping table and writing them into the parameter update buffer; and synchronously outputting the updated control parameters.

[0013] Optionally, the control command constraint module includes: writing intermediate power control commands into a time-continuous intermediate command buffer, and performing amplitude comparison, rate limiting, and state-of-charge direction verification in conjunction with the maximum charging power, maximum discharging power, maximum climbing speed, maximum descent speed, and state-of-charge safety range; generating energy storage allocation control commands that meet the requirements of power boundary, rate of change constraint, and state-of-charge protection at each time index position and writing them into the result slot.

[0014] The beneficial effects of this invention are: This invention constructs a trend-disturbance decoupled wind power analysis structure and uses an H-V adaptive double-layer smoothing model to decompose wind power at multiple scales, effectively extracting long-term variation trends and short-term disturbance components. This enables the energy storage control system to have long-term dispatch foresight while responding quickly to sudden disturbances, thereby improving the system's dispatch sensitivity and anti-disturbance capability.

[0015] This invention introduces the Teager-Kaiser energy operator to construct an instantaneous energy index, and combines an event judgment mechanism based on energy threshold and hysteresis band to achieve precise triggering of disturbance events, effectively avoiding resource waste and control instability caused by misjudgment of fluctuations or frequent start-stops.

[0016] This invention introduces a trend power and event modulation fusion mechanism in the power command generation stage. By driving stable operation through trend and achieving dynamic response through event modulation, the generated power commands are smoother and more controllable, significantly reducing the impact of power fluctuations in energy storage systems on equipment lifespan and operational safety. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This invention proposes a dynamic energy storage dispatch system based on wind power prediction. Figure 2 This is a schematic diagram of the H–V adaptive two-layer smoothing model processing flow proposed in this invention. Detailed Implementation

[0018] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0019] refer to Figure 1 - Figure 2 A dynamic energy storage dispatch system based on wind power prediction includes the following steps: The data processing module acquires real-time and historical power data from the wind farm and converts the power data into a standardized time series. The trend extraction module receives standardized time series data and decomposes the power data using an H-V adaptive two-layer smoothing model to generate trend baseline signals and residual signals. The energy analysis module receives the residual signal and extracts instantaneous energy indices using the Teager–Kaiser energy operator. The event judgment module receives instantaneous energy indicators, determines whether the triggering conditions are met based on the set energy threshold and hysteresis band parameters, and generates an event trigger signal. The power command generation module receives the trend baseline signal and the event trigger signal, and generates intermediate power control commands based on the trend baseline signal and the event trigger status. The energy storage system module performs energy storage power control operations and outputs operating status parameters in real time, including state of charge, current, voltage, or temperature. The parameter adaptive module receives the operating status parameters output by the energy storage system module and adjusts the control parameters in the power command generation module according to the operating status. The control command constraint module receives intermediate power control commands and performs constraint processing on the intermediate power control commands according to the power limitation conditions of the energy storage system module, generating energy storage dispatch control commands.

[0020] In this embodiment, the trend extraction module includes: Obtain a standardized time series. The standardized time series is obtained by performing unified unit conversion, time resampling, missing point completion, outlier handling and numerical normalization on the original wind power data to obtain a power data sequence that meets the requirements of time continuity and amplitude constraints, thus forming a structured input sequence based on equal interval time index. The standardized time series is used as the input signal of the H-V adaptive two-layer smoothing model, which takes the standardized power series as input and outputs a set of trend baseline signals to express long-term trend changes and a set of residual signals to express local perturbation differences. The trend baseline signal is a continuous curve of the same length as the input time series, used to reflect the stationary component of wind power under low and medium frequency variations, and has the characteristics of smoothness and time differentiability; The residual signal is a point-to-point difference sequence between the input power sequence and the trend baseline signal, used to characterize the perturbation component in the original power relative to the trend change; The trend baseline signal and residual signal output by the trend extraction module are sent to the power command generation module and the energy analysis module, respectively, forming a trend-disturbance decoupled input structure.

[0021] In this embodiment, the H–V adaptive two-layer smoothing model includes an outer Hemlert filtering layer and an inner Vondrák smoothing layer: The outer Hemlert filter layer includes: The standardized power time series is read point by point in chronological order, and a storage location for intermediate frequency residual signal, trend baseline signal, and disturbance signal is set for each time position. When processing any time position, with the time position as the center, actual samples within the width range are sequentially extracted from the standardized power time series according to a fixed left-right symmetrical width to form a neighborhood window; When the neighborhood window moves to the left or right beyond the effective range of the time series, the missing position of the neighborhood window is filled back by mirroring the sample value at the boundary of the time series. That is, the actual sample corresponding to the boundary point is copied to the missing position of the neighborhood window according to the symmetrical relationship between the boundary point and the boundary point. When processing any time position, the standardized power sample value of the time position is read, and the intermediate frequency residual signal written at the previous time position is read as a reference value. The absolute value of the difference between the standardized power sample value and the reference value is taken. The absolute value is repeatedly multiplied according to the preset nonlinear intensity parameter to form a power difference. The power difference is divided according to the preset smoothing scale parameter. The resulting ratio is used as the independent variable of the exponential decay function to perform exponential decay operation, and the output of the exponential decay is used as the center weight of the time position. Within the neighborhood window, the distance between each sample location and the center location is read; the distance decay coefficient corresponding to the distance is selected from the distance decay table, and the distance decay coefficient is multiplied by the center weight. The product is used as the unnormalized weight of the sample location. Add all the unnormalized weights one by one to get the total weight; then divide each unnormalized weight by the total weight, so that the sum of all weights equals one, and use the weights after division as the normalized weight sequence. The normalized weights are multiplied by the actual values ​​of each sample in the neighborhood window, and then all the products are added together in sequence. The final result of the sum is written into the intermediate frequency residual signal storage location at that time position. At the same time, the intermediate frequency residual signal is written as the reference value for the next time position. The inner Vondrák smoothing layer includes: In the inner Vondrák smoothing layer, a trend baseline slot is created for each time position, so that all trend baseline slots are linearly arranged according to the time order of the standardized power time series. An approximation term is placed at each trend baseline slot, which is equal to the intermediate frequency residual signal at the time position corresponding to that slot; Define continuity conditions for the trend baseline slot sequence: The difference between two consecutive trend slots is taken as a first difference term; The weighted combination of the current trend slot, the next trend slot, and the previous trend slot is used as a second-order difference term, where the trend values ​​of the next slot and the previous slot are each weighted by one, and the current slot is weighted by two and has the opposite sign. For each trend trough, a trend equation is established, which consists of two types of terms: The difference between the approximation term and the trend slot is multiplied once and used as the approximation part of the trend equation; The second-order difference term is multiplied once by the square of the trend smoothing factor and used as the smoothing part of the trend equation. The approximation part and the smoothed part are combined into a single trend equation by adding each term. According to the trend slots, all trend equations are written into a trend equation matrix in sequence. The trend equation matrix has non-zero parameters only at the main diagonal position, the first diagonal position, and the last diagonal position at each trend slot, thus forming a tri-diagonal structure. At the same time, all approximation terms are written into the right-hand vector in the order of the trend slots. Starting from the first row of the trend equation matrix, linear cancellation operations are performed sequentially on the corresponding columns from the second row to the last row. By multiplying and subtracting the parameters at the main diagonal position as coefficients in the next row, the parameters in that column of the next row are made zero. This operation is performed on all rows in order from top to bottom until the trend equation matrix is ​​transformed into an upper triangular structure that retains only its main diagonal and the diagonal above the main diagonal. Starting from the last row of the trend equation matrix, divide the right-hand vector value of the last row by the parameters of the main diagonal to obtain the trend baseline value of the last trend slot. Then, process the upper triangular structure in reverse order row by row. In each row, multiply the obtained subsequent trend slot values ​​by the parameters of the corresponding positions in the upper triangle, subtract the product from the right-hand term of the current row, and finally divide by the parameters of the main diagonal of the current row to obtain the trend baseline values ​​of all trend slots. Write all the obtained trend baseline values ​​into the trend baseline signal storage location in chronological order; For each time position, the intermediate frequency residual signal and the trend baseline signal are extracted in sequence, the intermediate frequency residual signal is subtracted from the trend baseline signal, and the resulting difference is written into the disturbance signal storage position of that time position.

[0022] In this embodiment, the energy analysis module includes: The residual signals are written into the residual sequence buffer in chronological order; an independent residual slot is set at each time index position in the residual sequence buffer to store the residual sample value at that position, while maintaining the integrity of the time index structure of the residual signals. According to the preset segment length L, the residual sequence buffer is divided into a group of continuous, fixed-length and non-overlapping residual segments starting from the starting index position; each residual segment contains L residual slots, and a segment energy buffer, segment direction buffer, segment energy concentration buffer and segment stability buffer corresponding to the residual segment are established in memory to store the energy structure inside the residual segment. When processing any residual segment, each index position is processed sequentially according to the time order within the residual segment. When processing the current index position, the following operations are performed: Read the residual value at the current index position from the current residual segment slot as the current residual value; Read the residual value of the previous index position from the current residual segment slot as the forward residual value; When the position is at the beginning of a segment, the residual value of the residual slot preceding the starting index position of the residual segment is read from the residual sequence buffer as the forward residual compensation value for that position. Read the residual value of the next position after the current index position from the current residual segment slot as the backward residual value; when the position is at the end of the segment, read the residual value of the residual slot after the end index position of the residual segment from the residual sequence buffer as the backward residual compensation value of that position. Multiply the current residual value by the current residual value and write the result as the current energy base value; Multiply the forward residual value by the backward residual value, and write the result as the neighborhood coupling energy value; Subtract the neighborhood coupled energy value from the current energy base value, and write the difference into the current index position slot of the segmented energy buffer; When processing any residual segment, each index position is processed sequentially according to the time order within that residual segment. The corresponding residual sample value is extracted from the current index position, the previous index position, and the next index position, respectively. The residual sample value at the current index position is denoted as... The residual sample value at the previous position is denoted as The residual sample value at the next position is denoted as The current residual sample value Multiplying by itself yields the energy base value, which is then used to calculate the forward residual sample value. Compared with backward residual sample values The neighborhood coupling value is obtained by multiplying the two values. The difference between the energy base value and the neighborhood coupling value is used as the segmented instantaneous energy value at the current index position and written into the segmented energy buffer. The corresponding relationship is expressed as follows: ; in, For the current time index position Based on residual sequences The calculated segmented instantaneous energy index; After writing to all L positions of the segmented energy buffer, directional analysis and directional consistency identification are performed on all segmented instantaneous energy within the residual segment in chronological order, specifically including: Within the residual segment, the difference operation is performed sequentially on the instantaneous energy of every two adjacent segments, and the sign of the difference is written into the segment direction buffer. In the segmented direction buffer, starting from the starting index position, search for adjacent intervals with the same symbol, and record all consecutive intervals with the same symbol as the direction-consistent area; Write the number of indexes in each direction-consistent region to the direction-consistent region length field, and write the direction-consistent region length field to the segmented energy concentration buffer. After performing directional consistency identification, energy concentration calculation is performed on the residual segment, specifically including: For each directional consistent region, sum the instantaneous energy of all segments within it sequentially, divide by the length of the directional consistent region, and write the quotient into the corresponding slot of the segmented energy concentration buffer. Find the maximum energy concentration value in the all-directional consistent region and write the maximum energy concentration value into the segmented energy concentration maximum value slot; After obtaining the maximum value of the segmented energy concentration, cross-segment stability identification is performed, which specifically includes: Perform a difference operation on the maximum value of segmented energy concentration of the residual segment and the previous residual segment, and write the difference into the segmented stability buffer. When the absolute value of the difference is less than the preset stability threshold S, the residual segment is marked as a stable segment; when the absolute value is greater than the stability threshold S, the residual segment is marked as a mutation segment. The maximum segment energy, maximum segment energy concentration, and segment stability flag of each residual segment are written into the energy structure output buffer according to the time coverage range corresponding to the residual segment, so that the energy structure output buffer remains continuous and without gaps throughout the entire time index. The three fields in the energy structure output buffer are combined into the final instantaneous energy index according to the following rules: Read the maximum segmented energy value corresponding to the time index position; Read the maximum value of segmented energy concentration at this time index position; Read the stability flag corresponding to the time index position and assign fixed weighting coefficients according to the different stable segments or abrupt segments; The above three values ​​are summed in a linear combination manner, and the result is written to the corresponding position in the instantaneous energy index buffer. Output the instantaneous energy index buffer in chronological order.

[0023] In this embodiment, the event determination module includes: The instantaneous energy index value at each time index position is written into the input buffer structure in sequence. The input buffer structure contains the corresponding energy amplitude slot and energy direction slot, which store the absolute value and sign value respectively, to ensure that the data corresponds one-to-one with the index position. Configure energy threshold parameters for the event detection module. With hysteresis band parameters An event status register is set up to record whether an event is in a "triggered state" or a "non-triggered state". This register is initialized to the "non-triggered state". Construct a sliding energy window of fixed length L for the current index position. The sliding energy window contains The energy amplitude and energy symbol value corresponding to each consecutive time index position are used as the local behavior region; When the event status register is in the "not triggered" state, the following three decisions are made at the current time index position: The energy amplitude at the current index position satisfies ; More than in the sliding energy window An energy direction value with the same sign as the current index; The average value of all energy amplitudes within the sliding energy window is greater than ; If all three conditions are met, the event state register is updated to "trigger state", and an event trigger signal value "1" is generated at the current time index position and written to the event trigger signal buffer. When the event status register is in "triggered state", the following two decisions are made at the current index position: The energy amplitude at the current index position satisfies ; The average energy amplitude within the sliding energy window is less than ; If both conditions are met, the event status register is updated to "not triggered" and an event trigger signal value of "0" is generated at the current time index position and written to the event trigger signal buffer. When the event status register is in "triggered state" and the cancellation condition is not met, the event trigger signal value "1" is continuously generated at the current index position and written to the event trigger signal buffer until the cancellation condition is met. After traversing all time index positions, the event trigger signals arranged in chronological order in the event trigger signal buffer are output as the final result.

[0024] In this embodiment, the power command generation module includes: The trend baseline signal is written into the trend baseline buffer in time index order. Each slot in the trend baseline buffer only stores the trend baseline value corresponding to the current time index, keeping the input sequence consistent with the original time axis. The event trigger signals output by the event judgment module are written into the event trigger buffer in time index order. Each slot in the event trigger buffer stores a binary result indicating whether the current time index is in the trigger state. The trigger state is "1" and the non-trigger state is "0". Three types of buffer slots are established: trend power slot, event modulation slot, and intermediate instruction slot.

[0025] Before the traversal begins, set the initial value of all event modulation slots to zero and the initial value of all intermediate instruction slots to zero. When processing any time index position, the trend baseline value corresponding to that time index is read from the trend baseline buffer, the trend baseline value is multiplied by the preset trend ratio coefficient, and the result is written into the trend power slot to form the trend power value of that time index position. At the same time index position, the trigger signal is read from the event trigger buffer. When the trigger signal is "1", the trend power value is read from the trend power slot. The trend power value is multiplied by the preset event gain coefficient. The result is written to the event modulation slot as the event modulation start value at that time index position. When the event trigger signal at multiple consecutive time index positions is "1", the event modulation value is read from the event modulation slot at the previous time index position. This value is then subjected to a linear weighted operation with the event modulation start value. The weighting ratio uses a preset event ramp-up coefficient. The weighted operation result is written to the event modulation slot at the current time index position, so that the event modulation value is written in a continuous ramp-up form along the time axis. When the trigger signal at the current time index position is "0" and the trigger signal at the previous time index position is "1", the event modulation value is read from the previous time index position, the event modulation value is multiplied by the preset attenuation coefficient, the result is written to the current event modulation slot, and the same attenuation operation is performed on the event modulation slot at each subsequent time index position until the event modulation value is equal to zero. At each time index position, the trend power value and event modulation value are read from the trend power slot and the event modulation slot respectively. The trend power value and the event modulation value are linearly summed according to the proportional coefficient. The linear summation result is written into the intermediate instruction slot as the intermediate power control instruction for the current time index position. Once all trend power slots, event modulation slots, and intermediate instruction slots at all time index positions have been written, all instructions in the intermediate instruction slot will be output in chronological order.

[0026] In this embodiment, the energy storage system module includes: Obtain energy storage allocation and control commands, and write the active power command value corresponding to each time index position into the power command buffer; The active power command value is read in the power command buffer in chronological order and together with the current DC bus voltage sample value, it participates in the current reference value calculation process. The current reference value is written into the current reference buffer so that the current reference buffer is consistent with the power command buffer in terms of time index. The current reference value in the current reference buffer is sent to the current closed-loop controller of the energy storage converter. The controller drives the switching devices of the energy storage converter through internal adjustment calculations to form a charging current or discharging current that matches the current reference value on the AC or DC side. Current sensors, voltage sensors, and temperature sensors are arranged at the terminals of energy storage cells or energy storage battery clusters. The terminal current, terminal voltage, and critical node temperature of the battery cluster are collected at fixed sampling periods, and the collected results are written into the current sampling buffer, voltage sampling buffer, and temperature sampling buffer. In each sampling period, the current current value is read from the current sampling buffer, the state of charge value of the previous period is read from the state of charge storage unit of the previous period, the state of charge value of the current period is updated by current integration, and the updated state of charge is written into the state of charge storage unit so that the state of charge is continuously updated over time. Consistency detection is performed between the terminal voltage value in the voltage sampling buffer and the state of charge value in the state of charge storage unit. When the deviation between the two exceeds the preset tolerance range, the state of charge value is corrected and the corrected state of charge value is rewritten into the state of charge storage unit. The current period's state of charge value, current sample value, voltage sample value, and temperature sample value are packaged into a running status parameter frame according to the time index, written into the running status output buffer, and the running status parameter frames in the running status output buffer are output in chronological order.

[0027] In this embodiment, the parameter adaptive module includes: Write the state of charge, current, voltage and temperature values ​​corresponding to each time index position in the running status parameter frame into the parameter input buffer in sequence, keeping the fields aligned according to the time index structure. Read the state of charge (SOC) value at the current time index position from the parameter input buffer, and compare it with the set lower and upper limits of SOC safety. If the SOC value is lower than the lower limit, generate a SOC reduction flag; if the SOC value is higher than the upper limit, generate a SOC suppression flag; otherwise, generate a SOC permission flag. Read the temperature value at the current time index position from the parameter input buffer and compare it with the set maximum temperature threshold. If the temperature value is higher than the threshold, generate a temperature weakening flag; otherwise, generate a normal temperature flag. Read the current and voltage values ​​at the current time index from the parameter input buffer, calculate the current power value, and compare the power value with the set maximum power threshold. If the current power value is greater than the threshold, generate a power reduction flag; otherwise, generate a power normal flag. The charge, temperature, and power tags are combined to form a state response matrix, and priority sorting and state fusion processing are performed on the matrix to generate a regulation command level field. This field is represented by three binary bits, which correspond to the constraint states of the three types of tags: charge, temperature, and electric power. Based on the adjustment instruction level field, the corresponding trend ratio adjustment coefficient, event gain adjustment coefficient, and event ramp rate coefficient are retrieved from the preset parameter mapping table, and the retrieved coefficient values ​​are written into the parameter update buffer. The three control parameter values ​​in the parameter update buffer are synchronously transmitted to the control parameter input port of the power command generation module, overwriting the current parameter setting value, and the control command is synthesized according to the new parameters in the next control cycle; Repeat the above steps to dynamically update the control parameter values ​​according to the operating status of the energy storage system, and continuously write them into the control parameter record table in chronological order.

[0028] In this embodiment, the control command constraint module includes: Intermediate power control instructions are written to the intermediate instruction buffer in time index order. Each slot in the intermediate instruction buffer stores the intermediate power control instruction value corresponding to the current time index, keeping the index structure continuous and consistent. The maximum charging power, maximum discharging power, maximum climbing speed, maximum descent speed, and state of charge safety range are read from the energy storage system module, and the above limiting parameters are written into the power limiting parameter buffer in a fixed order. When processing any time index position, the current intermediate power control command value is read from the intermediate command buffer, and the maximum charging power and maximum discharging power are read from the power limit parameter buffer. The current intermediate power control command value is compared with the upper and lower bounds of the two power limits. If the current command value is greater than the maximum discharging power, the limited discharging power value is written to the limiting result slot; if the current command value is less than the negative value of the maximum charging power, the limited charging power value is written to the limiting result slot; if it is between the two, the current intermediate power control command value is written to the limiting result slot. The energy storage allocation control command value of the previous cycle is read from the previous time index position, and the difference is calculated with the value of the current limiting result slot. This difference is compared with the maximum climbing speed and the maximum descent speed. If the difference is greater than the maximum climbing speed, the command value of the previous cycle plus the maximum climbing speed is written into the climbing and descent slot. If the difference is less than the negative value of the maximum descent speed, the command value of the previous cycle minus the maximum descent speed is written into the climbing and descent slot. Otherwise, the limiting result slot value is written into the climbing and descent slot. The system reads the state of charge (SOC) value at the current time index position from the energy storage system module and compares it with the safe SOC range. When the SOC value is higher than the upper limit of safety, the power value of the climbing and ramping slot is compared with zero. If it is in the discharge direction, the value is kept; if it is in the charging direction, the power value is forcibly written to zero. When the SOC value is lower than the lower limit of safety, the power value of the climbing and ramping slot is compared with zero. If it is in the charging direction, the value is kept; if it is in the discharge direction, the power value is forcibly written to zero. When the SOC is within the safe range, the power value of the climbing and ramping slot remains unchanged. Write the power command constrained by the state of charge into the energy storage allocation control command slot as the final energy storage allocation control command at the current time index position. After traversing all time index positions, the power command sequence arranged in time order in the energy storage allocation control command slot is output to the energy storage system module. Example

[0029] To verify the practical application effect of the wind power prediction-based energy storage dynamic allocation system of this invention, it was deployed in a wind farm with a total installed capacity of 100MW and equipped with a 20MW / 40MWh lithium iron phosphate battery energy storage system. The region experiences severe wind speed fluctuations, large diurnal temperature variations, and frequent and significant short-term wind power fluctuations. Previous systems using rule-driven and simple threshold methods for energy storage regulation suffer from the following problems: frequent start-up and shutdown of energy storage, excessive battery temperature rise, severe state-of-charge exceeding limits, untimely response to sudden wind power changes, and large command fluctuations.

[0030] After deploying the system of this invention, the data processing module standardizes the historical and real-time power data of the wind farms. The system performs unified unit conversion, time resampling, and outlier removal on the raw power data, generating a continuous and amplitude-controlled standardized time series. This time series is input to the trend extraction module and processed by the H–V adaptive two-layer smoothing model. This model consists of a Hemlert filtering layer and a Vondrák smoothing layer. The Hemlert layer extracts the intermediate frequency residual signal using a nonlinear weighting method, while the Vondrák layer generates the trend baseline signal by solving a tridiagonal sparse matrix. Ultimately, two output paths are formed: the trend baseline signal and the residual signal, corresponding to long-term power variations and disturbance fluctuations, respectively.

[0031] The trend baseline signal is fed into the power command generation module, while the residual signal is input into the energy analysis module for further calculation of instantaneous energy indices. During energy analysis, the residual signal is divided into fixed-length residual segments. The system uses the Teager-Kaiser energy operator to calculate local energy at each time index, identifying the disturbance structure through directional consistency and concentration analysis. For continuous disturbance intervals, the system dynamically determines the disturbance by combining energy amplitude and sign information. The judgment logic is implemented through an event judgment module, which uses a dual-threshold strategy based on energy threshold and hysteresis band to issue an event trigger signal when the triggering conditions are met.

[0032] The power command generation module receives the trend baseline signal and the event trigger signal, and internally establishes trend power slots, event modulation slots, and intermediate command slots. The system linearly weights the trend power and event disturbance compensation by setting a trend scaling factor and event gain parameter, achieving integrated control of smooth scheduling and rapid disturbance response. For continuous disturbance scenarios, the event modulation value increases at a set ramp rate and then rapidly decreases after the disturbance ends with a set attenuation factor. The generated intermediate power control command is sent to the control command constraint module.

[0033] Based on the energy storage system's maximum charge / discharge capacity, power change rate, and current state of charge, this module sequentially applies amplitude limits, slope constraints, and charge boundary corrections to the intermediate power command, ultimately generating energy storage allocation control commands to drive the converter to perform actual charge / discharge behavior. After the energy storage system is operational, it collects terminal current, voltage, temperature, and SOC (state of charge) every minute, constructing an operating status parameter frame and feeding it back to the parameter adaptive module.

[0034] This module generates indicators such as charge reduction, temperature reduction, and power suppression based on whether the current state of charge exceeds a safety threshold, whether the voltage is abnormal, whether the temperature is too high, and whether the power exceeds limits. These indicators are then combined to form a state response matrix. The response matrix indicates the applicable control level for the current operating state and dynamically adjusts the trend ratio, event gain, and event ramp-up rate parameters through a parameter mapping table, thereby achieving automatic adjustment of power control parameters for the next round. Control commands can thus dynamically adapt to the energy storage system's state, significantly improving the system's self-stabilization capability.

[0035] Throughout the scheduling process, each module operates continuously on a cyclical basis. Power fluctuations are jointly regulated through three paths: trend guidance, disturbance drive, and status feedback, enabling the system to possess accurate prediction, rapid response, and self-protection capabilities. To verify the actual operating effect of the system, we selected a one-month operating cycle from May 1 to June 1, 2024, and compared it with the wind farm's original traditional control strategy.

[0036] The following is a real comparison data, which comes from the records of the SCADA system and EMS system of the wind farm operation and maintenance center: Table 1 Comparison of the operational performance of the system of the present invention with that of the traditional control system

[0037] The data in the table shows that the system of this invention significantly optimized the energy storage operation status and wind power grid connection quality within the same operating cycle: the total number of energy storage start-ups and shutdowns decreased from 312 to 138, a reduction of approximately 55.8%, indicating smoother power commands and a more stable strategy, avoiding unnecessary frequent charging and discharging; the cumulative duration of SOC out-of-bounds charges decreased from 512 minutes to 46 minutes, with a voltage drop of approximately 91%, indicating that the state-of-charge constraint and parameter adaptive mechanism effectively prevented overcharging and over-discharging, significantly improving battery safety margin; the maximum jump in a single power command decreased from 5.3MW to 2.1MW, a reduction of approximately 60.4%, combined with an increase in the intraday power fluctuation reduction rate from 31.6% to 57.9% (a relative increase of approximately 83.2%), indicating that the intermediate command after the superposition of trend baseline guidance and event modulation significantly suppressed drastic fluctuations while tracking changes in wind power; and the accuracy of energy event identification increased from 69.4%. The average event response time was reduced from 18.2 seconds to 4.6 seconds, reaching 93.8%, demonstrating a substantial improvement in both disturbance detection and response timeliness based on instantaneous energy structure. The accuracy and speed of energy storage participation in regulation were both enhanced. At the equipment operating condition level, the average battery temperature of the energy storage system decreased from 41.7℃ to 36.3℃, a reduction of approximately 12.9%, and the number of abnormal power alarms decreased from 29 to 4, a reduction of approximately 86.2%. This indicates that the invention is highly effective in reducing thermal stress and the probability of extreme operating conditions, thus extending equipment lifespan. On the grid side, the monthly wind power absorption rate increased from 88.2% to 96.5%, a relative increase of approximately 9.4%. This demonstrates that under the same wind resource conditions, more refined dynamic allocation of energy storage can convert more electricity that might otherwise be curtailed into absorbable output, achieving a simultaneous improvement in grid friendliness and power plant economics.

[0038] The wind farm operation and maintenance team stated that the control system of this invention significantly improved dispatch performance after deployment. Particularly during the consecutive days of strong winds from May 19th to 21st, 2024, the previously prevalent widespread and recurring State of Charge (SOC) oscillations were largely eliminated, and energy storage control exhibited smooth, flexible, and precise dynamic behavior. Grid-side feedback also showed reduced frequency fluctuations, faster voltage response, and a significant improvement in the overall grid connection quality of the power plant.

[0039] In summary, the system of this invention has comprehensive advantages in real wind farm environments, including reasonable structure, fast response, stable control, and safe operation. It not only effectively solves the technical problems of frequent start-stop, serious false triggering, delayed response, and overloading that exist in traditional strategies, but also significantly improves the grid integration and absorption level of wind power, providing solid support for achieving a high proportion of new energy grid connection.

[0040] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A dynamic energy storage dispatch system based on wind power prediction, characterized in that, include: The data processing module acquires real-time and historical power data from the wind farm and converts the power data into a standardized time series. The trend extraction module receives a standardized time series and decomposes the power data using an H-V adaptive two-layer smoothing model to generate a trend baseline signal and a residual signal. The H-V adaptive two-layer smoothing model includes an outer Hemlert filter layer and an inner Vondrák smoothing layer. The energy analysis module receives the residual signal and extracts instantaneous energy indices using the Teager–Kaiser energy operator. The event judgment module receives instantaneous energy indicators, determines whether the triggering conditions are met based on the set energy threshold and hysteresis band parameters, and generates an event trigger signal. The power command generation module receives the trend baseline signal and the event trigger signal, and generates intermediate power control commands based on the trend baseline signal and the event trigger status. The energy storage system module performs energy storage power control operations and outputs operating status parameters in real time. The parameter adaptive module receives the operating status parameters output by the energy storage system module and adjusts the control parameters in the power command generation module according to the operating status. The control command constraint module receives intermediate power control commands and performs constraint processing on the intermediate power control commands according to the power limitation conditions of the energy storage system module, generating energy storage dispatch control commands.

2. The energy storage dynamic dispatch system based on wind power prediction according to claim 1, characterized in that, The trend extraction module includes: using the standardized time series as the input signal of the H-V adaptive two-layer smoothing model, wherein the H-V adaptive two-layer smoothing model takes the standardized power series as input and outputs a trend baseline signal expressing long-term trend changes and a residual signal expressing local perturbation differences.

3. The energy storage dynamic dispatch system based on wind power prediction according to claim 1, characterized in that, The energy analysis module includes: The residual signals are written into the residual sequence buffer in chronological order; an independent residual slot is set at each time index position in the residual sequence buffer; According to the preset segment length L, the residual sequence buffer is divided into a group of continuous, fixed-length and non-overlapping residual segments starting from the starting index position; each residual segment contains L residual slots, and a segment energy buffer, segment direction buffer, segment energy concentration buffer and segment stability buffer corresponding to the residual segment are established in memory. When processing any residual segment, each index position is processed sequentially according to the time order within the residual segment. When processing the current index position, the following operations are performed: Read the residual value at the current index position from the current residual segment slot as the current residual value; Read the residual value of the previous position from the current residual segment slot as the forward residual value; when the position is at the beginning of the segment, read the residual value of the residual slot one position before the starting index position of the residual segment from the residual sequence buffer as the forward residual compensation value of the position. Read the residual value of the next index position from the current residual segment slot as the backward residual value; When the position is at the end of a segment, the residual value of a residual slot after the end index position of the residual segment is read from the residual sequence buffer as the backward residual compensation value of the position. Multiply the current residual value by the current residual value and write the result as the current energy base value; Multiply the forward residual value by the backward residual value, and write the result as the neighborhood coupling energy value; Subtract the neighborhood coupling energy value from the current energy base value, and write the difference into the current index position slot of the segmented energy buffer; After writing to all L positions of the segmented energy buffer, directional analysis and directional consistency identification are performed on all segmented instantaneous energy within the residual segment in chronological order, specifically including: Within the residual segment, the difference operation is performed sequentially on the instantaneous energy of every two adjacent segments, and the sign of the difference is written into the segment direction buffer. In the segmented direction buffer, starting from the starting index position, search for adjacent intervals with the same symbol, and record all consecutive intervals with the same symbol as the direction-consistent area; Write the number of indexes in each direction-consistent region to the direction-consistent region length field, and write the direction-consistent region length field to the segmented energy concentration buffer. After performing direction consistency identification, energy concentration calculation is performed on the residual segments, specifically including: For each directional consistent region, sum the instantaneous energy of all segments within it sequentially, divide by the length of the directional consistent region, and write the quotient into the corresponding slot of the segmented energy concentration buffer. Find the maximum energy concentration value in the all-directional consistent region and write the maximum energy concentration value into the segmented energy concentration maximum value slot; After obtaining the maximum value of the segmented energy concentration, cross-segment stability identification is performed, which specifically includes: Perform a difference operation on the maximum value of segmented energy concentration of the residual segment and the previous residual segment, and write the difference into the segmented stability buffer. When the absolute value of the difference is less than the preset stability threshold S, the residual segment is marked as a stable segment; when the absolute value is greater than the stability threshold S, the residual segment is marked as a sudden change segment. Write the maximum segment energy, maximum segment energy concentration, and segment stability flag of each residual segment into the energy structure output buffer according to the time coverage range corresponding to the residual segment. The three fields in the energy structure output buffer are combined into the final instantaneous energy index.

4. The energy storage dynamic dispatch system based on wind power prediction according to claim 1, characterized in that, The event determination module includes: The instantaneous energy index at each time index position is written sequentially into the input buffer structure of the event judgment module; Configure energy threshold parameters With hysteresis band parameters An event status register is set to record whether an event is in a "triggered state" or a "non-triggered state", and the register is initialized to the "non-triggered state". Construct a sliding energy window of fixed length L for the current index position. The sliding energy window contains The energy amplitude and energy symbol value corresponding to each consecutive time index position are used as the local behavior region; When the event status register is in the "not triggered" state, the following three decisions are made at the current time index position: Criterion 1: The energy amplitude at the current index position satisfies... ; Criterion 2: Exceeding within the sliding energy window An energy direction value with the same sign as the current index; Criterion 3: The average value of all energy amplitudes within the sliding energy window is greater than... ; If all three conditions are met, the event state register is updated to "trigger state", and an event trigger signal value "1" is generated at the current time index position and written to the event trigger signal buffer. When the event status register is in "triggered state", the following two decisions are made at the current index position: The energy amplitude at the current index position satisfies ; The average energy amplitude within the sliding energy window is less than ; If both conditions are met, the event status register is updated to "not triggered" and an event trigger signal value of "0" is generated at the current time index position and written to the event trigger signal buffer. When the event status register is in "trigger state" and the cancellation condition is not met, the event trigger signal value "1" is continuously generated at the current index position and written to the event trigger signal buffer until the cancellation condition is met. After traversing all time index positions, the event trigger signals arranged in chronological order in the event trigger signal buffer are output.

5. The energy storage dynamic dispatch system based on wind power prediction according to claim 1, characterized in that, The power command generation module includes: The trend baseline signal is written to the trend baseline buffer in time index order. Each slot in the trend baseline buffer stores only the trend baseline value corresponding to the current time index. The event trigger signals are written to the event trigger buffer in time index order. Each slot in the event trigger buffer stores a binary result indicating whether the current time index is in the trigger state. The trigger state is "1" and the non-trigger state is "0". Three types of buffer slots are established within the power command generation module: trend power slot, event modulation slot, and intermediate command slot; Before the traversal begins, set the initial value of all event modulation slots to zero and the initial value of all intermediate instruction slots to zero. When processing any time index position, the trend baseline value corresponding to the time index is read from the trend baseline buffer, the trend baseline value is multiplied by a preset trend ratio coefficient, and the result is written into the trend power slot to form the trend power value of the time index position. At the same time index position, the trigger signal is read from the event trigger buffer. When the trigger signal is "1", the trend power value is read from the trend power slot. The trend power value is multiplied by the preset event gain coefficient. The result is written into the event modulation slot as the event modulation start value at the time index position. When the event trigger signal at N consecutive time index positions is "1", the event modulation value is read from the event modulation slot at the previous time index position. The event modulation value and the event modulation start value are subjected to a linear weighting operation. The weighting ratio uses a preset event ramp-up coefficient. The weighting operation result is written to the event modulation slot at the current time index position, so that the event modulation value is written in a continuous ramp-up form along the time axis. When the trigger signal at the current time index position is "0" and the trigger signal at the previous time index position is "1", the event modulation value is read from the previous time index position, the event modulation value is multiplied by the preset attenuation coefficient, the result is written to the current event modulation slot, and the same attenuation operation is performed on the event modulation slot at each subsequent time index position until the event modulation value is equal to zero. At each time index position, the trend power value and event modulation value are read from the trend power slot and the event modulation slot respectively. The trend power value and the event modulation value are linearly summed according to the proportional coefficient. The linear summation result is written into the intermediate instruction slot as the intermediate power control instruction for the current time index position. Once all trend power slots, event modulation slots, and intermediate instruction slots at all time index positions have been written, all instructions in the intermediate instruction slot will be output in chronological order.

6. The energy storage dynamic dispatch system based on wind power prediction according to claim 1, characterized in that, The energy storage system module includes: sequentially writing and reading energy storage allocation control commands on the time axis; using the commands and DC bus voltage together to obtain the current reference value; driving the converter unit to output the corresponding charging and discharging current with current closed-loop control; periodically collecting battery cluster operating parameters using current, voltage and temperature sensors and performing current integration to update the state of charge and terminal voltage consistency correction; and packaging the state of charge, current, voltage and temperature into an operating status parameter frame according to the time index and outputting it.

7. The energy storage dynamic dispatch system based on wind power prediction according to claim 1, characterized in that, The parameter adaptive module includes: writing the operating status parameter frame into the parameter input buffer according to the time index and sequentially completing the safe range judgment of the state of charge, temperature and power to generate corresponding tags; constructing and fusing the state response matrix to obtain the adjustment command level field; retrieving three types of adjustment coefficients, namely trend ratio, event gain and event ramp rate, from the parameter mapping table and writing them into the parameter update buffer; and synchronously outputting the updated control parameters.

8. The energy storage dynamic dispatch system based on wind power prediction according to claim 1, characterized in that, The control command constraint module includes: writing intermediate power control commands into a time-continuous intermediate command buffer, and performing amplitude comparison, rate limiting, and state-of-charge direction verification in combination with the maximum charging power, maximum discharging power, maximum climbing speed, maximum descent speed, and state-of-charge safety range; generating energy storage allocation control commands that meet the requirements of power boundary, rate of change constraint, and state-of-charge protection at each time index position and writing them into the result slot.

Images