Method for suppressing aging of energy storage battery participating in frequency modulation

Abstract

The application provides a method for inhibiting aging of an energy storage battery participating in primary frequency modulation, and relates to the technical field of optimization of energy storage batteries. The control method comprises the following steps: when the frequency deviation of a power grid is located in a frequency modulation dead zone, using an LSTM to predict SOE data of the energy storage battery, and using two kinds of recovery power to recover the SOE of the energy storage battery according to different predicted SOE data; when the frequency deviation of the power grid is not located in the frequency modulation dead zone, the energy storage battery participates in primary frequency modulation, and a primary frequency modulation control mechanism adjusts the output power of the energy storage battery according to the frequency deviation of the power grid. The control method can reduce the impact of SOE recovery power on the power grid, and prolong the service life of the energy storage battery by reducing the number of recovery actions of the energy storage battery.

Description

Technical Field

[0001] This invention relates to the field of energy storage battery optimization technology, and in particular to a method for suppressing and controlling the aging of energy storage batteries involved in primary frequency regulation. Background Technology

[0002] In recent years, energy storage battery technology has developed rapidly. Its fast response characteristics and precise power point tracking capabilities give it significant advantages in participating in grid auxiliary frequency regulation tasks. When the grid frequency deviates from its rated value, the energy storage battery can automatically increase or decrease active power through primary frequency regulation control, limiting grid frequency changes and supporting grid frequency stability. However, irregular charging and discharging conditions and frequently changing operating conditions lead to severe battery energy loss and low utilization. With the increase in cycle number, battery life faces the risk of rapid degradation, and may even cause the battery to no longer be able to effectively provide various grid auxiliary services, affecting the safe and stable operation of the grid. Therefore, suppressing the aging of energy storage batteries can not only improve battery life and performance, reduce battery replacement and maintenance costs, and enhance the sustainability of energy storage systems, but also help improve battery safety and avoid potential accidents and failure risks.

[0003] Currently, the key technologies for suppressing the aging of energy storage batteries are as follows:

[0004] Accurate prediction of battery aging: By establishing a battery aging model and combining it with real-time battery operating data, the performance degradation and remaining lifespan during the battery aging process can be accurately predicted.

[0005] Dynamic power management: During actual operation, the charging and discharging power of the battery is adjusted in real time based on the battery's aging status and predictive information to minimize the aging rate.

[0006] Recovery control optimization: Adjust the battery SOE (State of Energy) and power command according to the recovery reference range to slow down the battery aging rate while ensuring that the battery can participate normally in power system tasks.

[0007] When implementing control methods to suppress the aging of energy storage batteries, the following main challenges are still faced:

[0008] Balancing safety and performance: Strategies to suppress battery aging need to be implemented while ensuring the safe operation of the power grid, effectively achieving a balance between mitigating aging and responding to power grid tasks.

[0009] Accuracy of battery aging model: Establishing an accurate battery aging model is crucial, and it is necessary to combine actual operating data and experimental verification to improve the accuracy of the model.

[0010] System feasibility and real-time performance: Algorithms and strategies for suppressing battery aging need to be verified through actual systems or aging assessment models with high accuracy, and must be real-time and feasible in order to play a role in practical applications. Summary of the Invention

[0011] To address the technical problems existing in the prior art, this invention provides a method for suppressing aging of energy storage batteries participating in primary frequency regulation. This method includes...

[0012] If the grid frequency deviation is within the frequency regulation dead zone, then SOE recovery is initiated:

[0013] The SOE data of the energy storage battery is predicted using LSTM (Long Short-Term Memory) network, and CDS (Charging or Discharging Segment) segments are extracted.

[0014] Calculate the change in SOE from the moment the grid frequency deviation returns to the frequency regulation dead zone t(k) to the end of the CDS segment.

[0015]

[0016] In the formula, t(m) and t(n) are the start and end times of the CDS segment, respectively; The change in SOE for CDS segments;

[0017] like If the value is less than a preset threshold, then the first recovery power P is applied. RSOE (t) Restore SOE:

[0018]

[0019] In the formula, μ(t) is the SOE of the energy storage battery at the current moment; μ ref This is the recovery reference value for SOE; E BESS Rated capacity of the energy storage battery, where Δt is the duration of charging or discharging;

[0020] like If it is not less than the preset threshold, then Adding the SOE at time t(k) yields

[0021] like Then charging or discharging will not occur;

[0022] like Then according to the second recovery power P′ RSOE(t) Restore SOE:

[0023]

[0024] In the formula, μ l To begin restoring the lower bound of SOE, μ h To begin restoring the SOE cap.

[0025] Correspondingly, if the grid frequency deviation is not located in the frequency regulation dead zone, the energy storage battery participates in the primary frequency regulation, and the primary frequency regulation control mechanism adjusts the output power of the energy storage battery according to the grid frequency deviation.

[0026] Preferably, the above-mentioned method of using LSTM to predict SOE data of energy storage batteries and extracting CDS segments includes: segmenting SOE data according to the direction of fluctuation, replacing segments with fluctuations less than a preset value with straight lines, and extracting CDS segments.

[0027] Preferably, using LSTM to predict SOE data of energy storage batteries specifically includes: inputting periodically iterated historical SOE data into a baseline LSTM for prediction to obtain reference predicted values ​​of SOE, and constructing SOE data polylines based on multiple reference predicted values.

[0028] Furthermore, the reference predicted value is compared with the real-time SOE to obtain the correction error. The correction error is input into the correction layer LSTM to obtain the predicted correction value of SOE at the next time step. The predicted value of SOE is obtained by combining the reference predicted value and the predicted correction value. SOE data polyline is constructed based on multiple predicted values.

[0029] As can be seen, this invention segments the SOE (Sequential Frequency Equilibrium) and uses charge / discharge power control for recovery within different SOE ranges, effectively reducing the impact of SOE recovery power on the power grid and decreasing the number of recovery actions, thus extending the battery's lifespan. Furthermore, the battery only participates in frequency regulation once when the grid frequency deviation is not within the frequency regulation dead zone, avoiding repeated recovery actions within the dead zone and achieving the technical effect of suppressing battery aging.

[0030] In addition to achieving the aforementioned beneficial effects, each preferred scheme also achieves the following further beneficial effects: During SOE data processing, replacing segments with fluctuations less than the preset value with straight lines facilitates later identification of CDS segments and improves work efficiency; Optimizing and adjusting the battery's SOE recovery strategy based on LSTM predicted SOE values ​​can fully utilize the charging and discharging trends of the energy storage battery during primary frequency regulation to recover SOE, avoiding excessively high or low SOE values, effectively mitigating accelerated aging of the energy storage battery during primary frequency regulation, and extending its lifespan; Furthermore, based on battery charging and discharging information predicted by dual-layer LSTM rolling, the SOE value of the energy storage battery can be predicted more accurately. Attached Figure Description

[0031] Figure 1This is a schematic diagram of the control process for suppressing the aging of energy storage batteries based on dual-layer LSTM in this invention.

[0032] Figure 2 This is a schematic diagram of the process of predicting the SOE of a battery based on a dual-layer LSTM in this invention.

[0033] Figure 3 This is a schematic diagram of the filtering process for SOE prediction values ​​in this invention.

[0034] Figure 4 This is a schematic diagram of the SOE prediction results of energy storage batteries based on single and double-layer LSTM in this invention.

[0035] Figure 5 This is a schematic diagram comparing the incremental aging of the energy storage battery in this invention. Detailed Implementation

[0036] The inventors will now provide a more detailed description of the invention in conjunction with the accompanying drawings and specific embodiments.

[0037] In existing technologies, the penetration rate of battery energy storage systems in the power grid is constantly increasing, requiring the provision of various types of ancillary services to the grid. Primary frequency regulation is a major scenario in which battery energy storage systems participate in grid regulation. However, when providing primary frequency regulation to the grid, energy storage batteries need to be frequently charged and discharged, which has a significant impact on the aging characteristics and cycle life of the batteries.

[0038] This invention proposes an aging suppression control method for energy storage batteries participating in primary frequency regulation, the control process of which is as follows: Figure 1 As shown, it includes:

[0039] First, such as Figure 2 As shown, the SOE of energy storage batteries is predicted based on long short-term memory networks:

[0040] Historical SOE data from periodic iterations are input into a baseline LSTM layer for prediction to obtain reference prediction values ​​for SOE. Based on multiple reference prediction values, SOE data polylines are constructed, and CDS segments are extracted.

[0041] Preferably, the reference predicted value can be compared with the real-time SOE to obtain the correction error. The correction error is input into the correction layer LSTM to obtain the predicted correction value of SOE at the next time step. The reference predicted value and the predicted correction value are added to obtain the predicted value of SOE. Based on multiple predicted values, SOE data polylines are constructed, and then CDS segments are extracted.

[0042] Secondly, preferably, the predicted SOE data should be preprocessed before being applied to the design of the recovery strategy. For example... Figure 3As shown, the SOE data is first segmented according to the direction of fluctuation, with the inflection points of the curves recorded as the endpoints of the segments. Then, segments with smaller SOE fluctuations are filtered according to a set threshold (fluctuations greater than the threshold are retained, while those less than the threshold are replaced with straight lines), and the corresponding CDS segments are extracted.

[0043] If the power grid frequency deviation is located in the frequency regulation dead zone δ DB Then, SOE recovery is initiated by calculating the change in SOE from the moment the grid frequency deviation returns to the frequency regulation dead zone t(k) to the end of the CDS segment.

[0044]

[0045] In the formula, t(m) and t(n) are the start and end times of the CDS segment, respectively; This represents the change in SOE segmentation of the CDS. If... If the value is less than the preset threshold ε, then the first recovery power P is applied. RSOE (t) Restore SOE:

[0046]

[0047] In the formula, μ(t) is the SOE of the energy storage battery at the current moment; μ ref The recovery reference value for SOE is, for example, 50%; E BESS The rated capacity of the energy storage battery, where Δt is the duration of charging or discharging.

[0048] like If it is not less than the preset threshold ε, then Adding the SOE at time t(k) yields like Then charging and discharging will not occur; if Then according to the second recovery power P′ RSOE (t) Restore SOE:

[0049]

[0050] In the formula, μ l To begin restoring the lower bound of SOE, μ h To begin restoring the SOE cap.

[0051] If the grid frequency deviation is not located in the frequency regulation dead zone δ DB If the energy storage battery participates in primary frequency regulation, the primary frequency regulation control mechanism adjusts the output power of the energy storage battery according to the grid frequency deviation. RSOE To set the battery energy storage system (BESS) used for SOE recovery to a positive output power, P BESS This is the overall output of the energy storage battery system.

[0052] Among them, the frequency modulation dead zone δ DB The frequency is typically 0.033 Hz. Under primary frequency regulation control, the change in the energy storage battery SOE (i.e., μ(t)) at time t is dynamically determined by the frequency deviation, and thus...

[0053]

[0054] In the formula, K B Δf is the frequency droop control coefficient, and Δf is the grid frequency deviation.

[0055] The present invention will be further described below through an embodiment.

[0056] Example 1

[0057] This invention verifies the proposed control method using Matlab software through an implementation example. This example uses frequency deviation data Δf from the first half of 2021 from an actual power grid for operational simulation (sampling interval of 5 minutes), and compares it with the current traditional frequency regulation dead zone SOE recovery control. The parameter settings for the two-layer LSTM rolling prediction model are shown in Table 1, and the SOE recovery control parameter settings are shown in Table 2.

[0058] Table 1. Parameter settings for the two-layer LSTM rolling prediction model

[0059]

[0060] Table 2. SOE Frequency Dead Zone Recovery Control Parameter Settings

[0061]

[0062] In SOE rolling prediction based on a two-layer LSTM, the prediction period can be adjusted according to the actual situation. In this embodiment, the prediction period is 24 hours (based on a 5-minute sampling frequency, the number of data samples in 24 hours is 288). The learning data for LSTM prediction comes from the real data in the previous prediction period, thus continuously updating the network parameters and improving prediction accuracy. Figure 4 As shown, the prediction results are displayed over a 10-day period, comparing the prediction results. Single-layer LSTM refers to an LSTM with only a baseline prediction layer, while double-layer LSTM refers to an LSTM containing both a baseline prediction layer and an error correction layer. Comparing with real data, it is clear that the prediction accuracy of double-layer LSTM is improved, effectively reducing prediction errors.

[0063] To demonstrate the effectiveness of this invention in suppressing battery aging, this embodiment sets up no SOE recovery control and traditional SOE recovery control as control groups for the SOE recovery control method proposed in this invention, and compares the differences in battery aging under the three methods.

[0064] Traditional SOE recovery strategies are based on two decision-making steps.

[0065] First, determine whether the frequency exceeds the frequency tuning dead zone. Similar to this invention, SOE recovery will only take effect when the frequency is within the dead zone range; otherwise, the energy storage battery should provide frequency control.

[0066] Second, determine whether the real-time SOE value is equal to the reference value μ. ref When the frequency is within the dead zone, if the real-time SOE value exceeds μ... ref Then to the reference value μ ref Perform discharge recovery; if the real-time SOE value is equal to μ ref The energy storage battery will not operate; if the real-time SOE value is lower than μ ref Then to the reference value μ ref Perform charging recovery. Specifically, perform charging and discharging recovery according to the first recovery power.

[0067] This invention employs an empirical model (B. Xu, A. Oudalov, A. Ulbig, et al., “Modeling of Lithium-Ion Battery Degradation for Cell Life Assessment,” IEEE Trans. Smart Grid, vol. 9, no. 2, pp. 1131–1140, 2018.) to evaluate battery capacity loss. This model considers both calendar aging and cycle aging, and comparison with experimental aging results shows that it has high accuracy.

[0068] Figure 5 The data shows the incremental aging of the energy storage battery as it provides primary frequency regulation to the grid over 181 days, comparing the results of a recovery strategy without SOE, a traditional SOE recovery strategy, and the recovery control method of this invention. Figure 5 As shown, both the traditional SOE recovery strategy and the recovery control of this invention can reduce the incremental aging of the battery, but the recovery control of this invention is more effective. The traditional strategy lacks flexibility and does not consider the SOE changes caused by the frequency modulation operation itself. If the battery recovers from a high SOE through frequency modulation dead-zone discharge and then continues to discharge at the same frequency, it increases the number of additional battery cycles and further reduces the SOE level, thus affecting the effectiveness of mitigating battery aging. Compared to the frequent operations of the traditional recovery strategy, the recovery control of this invention can estimate future SOE changes and adaptively improve the recovery battery SOE level using the primary frequency modulation trend, effectively reducing the number of battery cycles and mitigating battery aging.

[0069] As can be seen, this invention segments the SOE (Sequential Frequency Equilibrium) and uses charge / discharge power control for recovery within different SOE ranges, effectively reducing the impact of SOE recovery power on the power grid and decreasing the number of recovery actions, thus extending the battery's lifespan. Furthermore, the battery only participates in frequency regulation once when the grid frequency deviation is not within the frequency regulation dead zone, avoiding repeated recovery actions within the dead zone and achieving the technical effect of suppressing battery aging.

[0070] In addition to achieving the aforementioned beneficial effects, each preferred scheme also achieves the following further beneficial effects: During SOE data processing, replacing segments with fluctuations less than the preset value with straight lines facilitates later identification of CDS segments and improves work efficiency; Optimizing and adjusting the battery's SOE recovery strategy based on LSTM predicted SOE values ​​can fully utilize the charging and discharging trends of the energy storage battery during primary frequency regulation to recover SOE, avoiding excessively high or low SOE values, effectively mitigating accelerated aging of the energy storage battery during primary frequency regulation, and extending its lifespan; Furthermore, based on battery charging and discharging information predicted by dual-layer LSTM rolling, the SOE value of the energy storage battery can be predicted more accurately.

[0071] The above are merely preferred embodiments of the present invention and are not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. 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.

Claims

1. A method for suppressing aging of energy storage batteries participating in primary frequency regulation, characterized in that, include: If the grid frequency deviation is within the frequency regulation dead zone, then SOE recovery is initiated: LSTM is used to predict the SOE data of the energy storage battery and CDS segments are extracted. Calculate the change in SOE from the moment the grid frequency deviation returns to the frequency regulation dead zone t(k) to the end of the CDS segment. In the formula, t(m) and t(n) are the start and end times of the CDS segment, respectively; The change in SOE for CDS segments; like If it is less than the preset threshold, then according to the first recovery power P RSOE (t) Restore SOE: In the formula, μ(t) is the SOE of the energy storage battery at the current moment; μ ref This is the recovery reference value for SOE; E BESS Rated capacity of the energy storage battery, where Δt is the duration of charging or discharging; like If it is not less than the preset threshold, then Adding the SOE at time t(k) yields like Then charging or discharging will not occur; like Then according to the second recovery power P′ RSOE (t) Restore SOE: In the formula, μ l To begin restoring the lower bound of SOE, μ h To begin restoring the SOE cap; If the grid frequency deviation is not located in the frequency regulation dead zone, the energy storage battery participates in the primary frequency regulation, and the primary frequency regulation control mechanism adjusts the output power of the energy storage battery according to the grid frequency deviation.

2. The method for suppressing aging of energy storage batteries participating in primary frequency regulation as described in claim 1, characterized in that, The process of using LSTM to predict the SOE data of energy storage batteries and extracting CDS segments includes: The SOE data is segmented according to the direction of fluctuation. Segments with fluctuations less than a preset value are replaced with straight lines, and CDS segments are extracted.

3. The method for suppressing aging of energy storage batteries participating in primary frequency regulation as described in claim 1, characterized in that, The process of using LSTM to predict the SOE data of energy storage batteries and extracting CDS segments includes: Historical SOE data from periodic iterations are input into a baseline LSTM layer for prediction to obtain reference prediction values ​​for SOE. Based on multiple reference prediction values, SOE data polylines are constructed, and CDS segments are extracted.

4. The method for suppressing aging of energy storage batteries participating in primary frequency regulation as described in claim 3, characterized in that, Also includes: The correction error is obtained by comparing the reference prediction value with the real-time SOE. The correction error is then input into the correction layer LSTM to obtain the predicted correction value of SOE at the next time step. The reference prediction value and the predicted correction value are added together to obtain the predicted value of SOE. Based on multiple prediction values, SOE data polylines are constructed, and CDS segments are extracted.

Images