A dynamic capacity flexible regulation method of an energy storage system

Abstract

The application relates to the technical field of energy storage system control, and discloses a dynamic capacity flexible regulation method of an energy storage system. The method pre-processes historical operation data to form standardized operation index time series data, and constructs multiple capacity regulation characteristic profiles reflecting different operation states according to the time series data. In real-time application, the matching degrees of current operation data flow and the characteristic profiles are identified, and the target profile with the highest matching degree and the associated adjacent profiles are extracted. The core lies in that, instead of directly adopting the fixed strategy of the target profile, the capacity regulation strategy is dynamically generated according to the data distribution characteristics between the target profile and the adjacent profiles. Finally, the strategy is executed to realize flexible adjustment of the output capacity of the energy storage system. Through multi-profile matching and data distribution fusion, the application overcomes the limitations of traditional rigid regulation methods, and significantly enhances the self-adaptive ability of the system to complex operation conditions.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy storage system control, in particular to a dynamic capacity flexible regulation method of an energy storage system. BACKGROUND

[0002] The current energy storage system capacity regulation mainly adopts fixed threshold control or simple planning and scheduling method. The existing technology does not make full use of historical operation data, and fails to mine the characteristic rules under different operating conditions. The capacity regulation strategy is static, which cannot adapt to the real-time changing operating conditions. The feature extraction method is single, which is difficult to fully reflect the dynamic characteristics of the system. The matching recognition precision is low, and the target state positioning is not accurate. The strategy generation mechanism is fixed, and lacks consideration of the adjacent state transition characteristics. The existing method needs to solve the key technical problems of feature mining, state recognition and dynamic strategy generation.

[0003] The traditional energy storage regulation method has obvious deficiencies in adaptability and refinement. The historical data preprocessing method is simple, and the noise interference affects the data quality. The feature profile construction dimension is limited, and the key state characteristics are seriously lost. The real-time data matching algorithm has high complexity and large calculation delay. The adjacent profile selection standard is subjective, and the correlation judgment accuracy is poor. The distribution characteristic analysis is linear, which fails to capture the nonlinear change rule. The regulation strategy generation is empirical, which cannot realize flexible adjustment. The execution mechanism response lags, and the capacity adjustment precision is insufficient. SUMMARY

[0004] The purpose of the present application is to provide a dynamic capacity flexible regulation method of an energy storage system to solve the problems raised in the background art.

[0005] To achieve the above purpose, the present application provides a dynamic capacity flexible regulation method of an energy storage system, which comprises:

[0006] The historical operation data of the energy storage system is preprocessed to form standardized operation index time series data;

[0007] Based on the operation index time series data, a plurality of capacity regulation characteristic profiles reflecting different operating states of the system are constructed;

[0008] The current operation data stream of the energy storage system is collected in real time, and the matching degree of the current operation data stream and the plurality of capacity regulation characteristic profiles is identified;

[0009] The target capacity regulation characteristic profile with the highest matching degree and its associated adjacent capacity regulation characteristic profile are extracted from the plurality of capacity regulation characteristic profiles;

[0010] According to the data distribution characteristics between the target capacity regulation characteristic profile and the adjacent capacity regulation characteristic profile, a dynamic capacity regulation strategy is generated;

[0011] The dynamic capacity regulation strategy is executed to flexibly adjust the output capacity of the energy storage system.

[0012] Preferably, the preprocessing of the historical operation data of the energy storage system comprises:

[0013] The historical operation data is cleaned to eliminate abnormal data points that are obviously beyond a reasonable range;

[0014] The cleaned historical operation data is normalized to unify the numerical range thereof to between zero and one;

[0015] The normalized data is divided according to a preset length of a time window, and the data in each time window constitutes an operation index time series data segment.

[0016] Preferably, the multiple capacity regulation characteristic profiles reflecting different operation states of the system are constructed, comprising:

[0017] Feature extraction is performed on each operation index time series data segment to obtain a feature vector representing the operation state of the system in the time window corresponding to the operation index time series data segment;

[0018] Density peak clustering algorithm is used to perform clustering analysis on all the feature vectors, and each clustering center represents a typical operation state of the system;

[0019] All operation index time series data segments belonging to the same cluster are fused and averaged to form a capacity regulation characteristic profile corresponding to the cluster set obtained by the density peak clustering algorithm.

[0020] Preferably, the matching degree of the current operation data stream and the multiple capacity regulation characteristic profiles is identified, comprising:

[0021] The current operation data stream collected in real time is processed according to the same preprocessing manner as the historical data to obtain a current operation feature vector;

[0022] The Euclidean distance between the current operation feature vector and the feature vector corresponding to each capacity regulation characteristic profile is calculated;

[0023] The Euclidean distance is inversely converted to obtain a matching degree score between the current operation data stream and each capacity regulation characteristic profile.

[0024] Preferably, the target capacity regulation characteristic profile with the highest matching degree to the current operation data stream and the adjacent capacity regulation characteristic profiles associated therewith are extracted from the multiple capacity regulation characteristic profiles, comprising:

[0025] selecting a capacity regulation feature profile with the highest matching degree score as a target capacity regulation feature profile;

[0026] calculating distances between other capacity regulation feature profiles and the target capacity regulation feature profile in a feature space;

[0027] determining, as neighboring capacity regulation feature profiles, capacity regulation feature profiles with distances less than a dynamic similarity threshold, the dynamic similarity threshold being self-adaptively adjusted according to fluctuation of a recent operation data stream.

[0028] Preferably, the generating of the dynamic capacity regulation strategy comprises:

[0029] analyzing data distribution differences between the target capacity regulation feature profile and each neighboring capacity regulation feature profile in key operation indexes;

[0030] determining a direction and a range candidate of capacity adjustment of the energy storage system according to the data distribution differences;

[0031] determining a final capacity adjustment amount and an adjustment rate from the range candidate in combination with a real-time power grid frequency signal and a state of charge of the energy storage system, to form a complete dynamic capacity regulation strategy.

[0032] Preferably, the analyzing of the data distribution differences between the target capacity regulation feature profile and each neighboring capacity regulation feature profile in key operation indexes comprises:

[0033] calculating a mean difference and a variance ratio between the target capacity regulation feature profile and each neighboring capacity regulation feature profile in a power output index;

[0034] calculating a distribution overlap area between the target capacity regulation feature profile and each neighboring capacity regulation feature profile in an efficiency index;

[0035] weighting and combining the mean difference, the variance ratio and the distribution overlap area to obtain a comprehensive difference index.

[0036] Preferably, the determining of the direction and the range candidate of capacity adjustment of the energy storage system comprises:

[0037] if the comprehensive difference index shows that the target profile tends to have a higher power output compared with the neighboring profiles, setting a capacity up-regulation direction;

[0038] linearly mapping upper and lower limits of a capacity adjustment range according to a numerical value of the comprehensive difference index, to form the range candidate;

[0039] adjusting the range candidate by using a deviation between an actual range and an expected range of a successfully executed regulation strategy in the recent period as a correction term.

[0040] Preferably, the combination of real-time grid frequency signal and state of charge of the energy storage system includes:

[0041] Monitoring the deviation value of real-time grid frequency from the rated frequency;

[0042] Obtaining the state of charge value of the energy storage system at the current time;

[0043] Establishing a fuzzy logic rule table with frequency deviation and state of charge as input and capacity adjustment correction coefficient as output;

[0044] Using the fuzzy logic rule table to correct the preliminary adjustment amount selected from the candidate range.

[0045] Preferably, the execution of the dynamic capacity regulation strategy includes:

[0046] Packaging the final determined capacity adjustment amount, adjustment direction and adjustment rate into a standard control instruction;

[0047] Issuing the standard control instruction to the energy storage converter;

[0048] Monitoring the actual output power of the energy storage system after the execution of the instruction, and comparing it with the expected value of the strategy, and feeding back the comparison difference to the strategy generation stage for optimizing the generation of subsequent regulation strategy.

[0049] Compared with the prior art, the present application has the following beneficial effects:

[0050] The historical operation data of the energy storage system is preprocessed to form standardized operation index time series data. The preprocessing includes data cleaning, outlier removal and normalization operation. The operation index covers key indicators such as charging and discharging rate, state of charge, efficiency parameter, etc. The time series data is sorted by timestamp to maintain data continuity. Standardization processing eliminates the dimensional difference and improves the data comparability. Through data preprocessing, high-quality data foundation is provided for feature mining.

[0051] Based on the operation index time series data, a plurality of capacity regulation feature profiles reflecting different operating states of the system are constructed. The feature profiles are generated by clustering algorithm, and each profile represents a typical operating mode. The profile features include mean, variance, extreme value and other statistical quantities. The number of profiles is determined by optimizing the contour coefficient to ensure the rationality of classification. The profile data is dynamically updated to adapt to the changes of system characteristics. Through the construction of feature profiles, the standard quantification of operating state is realized.

[0052] The system collects real-time operational data streams from the energy storage system and identifies their matching degree with multiple capacity control characteristic profiles. Data streams are acquired in real-time via a sensor network to ensure data timeliness. Matching degree calculation employs a similarity measurement algorithm to quantify the degree of proximity between the current state and each profile. Matching results include confidence assessments to improve identification reliability. Through real-time matching, accurate classification of operational states is achieved.

[0053] The target profile with the highest matching degree and its associated neighboring profiles are extracted from multiple capacity regulation feature profiles. The target profile is selected based on the principle of highest matching degree to ensure representativeness. Neighboring profiles are determined according to feature space distance to reflect possible state transitions. Profile correlation is verified through topological analysis to ensure logical rationality. The extraction of target and neighboring profiles provides multi-dimensional references for strategy generation.

[0054] Dynamic capacity control strategies are generated based on the data distribution characteristics between the target feature profile and neighboring feature profiles. These distribution characteristics include statistical features such as probability density and correlation. Strategy generation considers state transition probabilities and system constraints. Control parameters are adaptively adjusted according to the distribution characteristics to achieve flexible control. The strategy optimization aims to balance system efficiency and equipment lifespan. Dynamic strategy generation improves the accuracy of control.

[0055] A dynamic capacity control strategy is implemented to flexibly adjust the output capacity of the energy storage system. The process employs a gradual adjustment strategy to avoid sudden power surges. The adjustment range is optimized in real time based on system status to ensure operational stability. Flexible adjustment balances response speed and equipment safety. Through precise execution, optimized control of capacity output is achieved. Attached Figure Description

[0056] Figure 1 This is a schematic diagram illustrating the working principle of the dynamic capacity flexible control method for the energy storage system described in this invention.

[0057] Figure 2 A flowchart for preprocessing historical operating data of an energy storage system;

[0058] Figure 3 A flowchart for identifying the matching degree between the current running data stream and the capacity control feature profile;

[0059] Figure 4 This is a monitoring diagram showing the power deviation distribution and allowable range of the energy storage system.

[0060] Figure 5 A trend chart showing the optimization effect of strategy feedback. Detailed Implementation

[0061] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0062] Please see Figure 1 This invention provides a dynamic capacity flexible control method for an energy storage system. The method includes: preprocessing historical operating data of the energy storage system, removing outliers and normalizing the data, and then dividing the data into standardized time-series operating indicators according to a preset time window; extracting feature vectors from these time-series data segments and clustering them using a density peak clustering algorithm, merging and averaging the data segments of each cluster to construct multiple capacity control feature profiles reflecting different operating states of the system; acquiring the current operating data stream of the energy storage system in real time, obtaining the current operating feature vector after the same preprocessing, calculating its Euclidean distance with the vectors corresponding to each feature profile, and obtaining the matching degree score through inverse proportional conversion; selecting the capacity control feature profile with the highest matching degree as the target profile, and determining its neighboring profiles based on a dynamic similarity threshold; analyzing the data distribution differences between the target profile and neighboring profiles on key operating indicators, and generating a dynamic control strategy for capacity adjustment direction, amplitude, and rate by combining the grid frequency signal and the state of charge of the energy storage system; finally, encapsulating the strategy into control commands and issuing them to the energy storage converter, and monitoring the actual output for feedback optimization.

[0063] Example 1: See Figure 2 In practical implementation, when preprocessing the historical operating data of the energy storage system, the historical operating data is cleaned to remove abnormal data points that are clearly outside the reasonable range, such as invalid data such as power values ​​exceeding the system's rated capacity or negative values. In practice, the cleaned historical operating data is then normalized to unify its numerical range to between zero and one. The normalization process uses a minimum-maximum scaling method, and its formula is expressed as:

[0064]

[0065] in: These are the original historical operation data values. It is the minimum value in the historical data sequence. It is the maximum value in the historical data sequence. These are normalized historical operating data values. In specific implementations, the normalized historical operating data is divided into time windows of preset length. The historical operating data within each time window constitutes a time-series data segment of the operating indicator. The length of the time window can be set to a fixed duration according to system characteristics. In some embodiments, the normalization process can also employ the Z-score standardization method to convert the historical operating data into a distribution with a mean of zero and a standard deviation of one.

[0066] In practical implementation, when constructing multiple capacity control feature profiles reflecting different system operating states, feature extraction is performed on each time-series data segment of operating indicators to obtain feature vectors representing the system operating state within the time window corresponding to the time-series data segment of operating indicators. Feature extraction includes calculating the mean, variance, and peak value of the time-series data segment of operating indicators. In practical implementation, a density peak clustering algorithm is used to perform cluster analysis on all feature vectors. The density peak clustering algorithm determines the cluster center by calculating the local density and relative distance of each feature vector, and each cluster center represents a typical system operating state. In some embodiments, feature extraction may also include time-domain features such as root mean square or frequency-domain features such as spectral components of the time-series data segment of operating indicators. All time-series data segments of operating indicators belonging to the same cluster are fused and averaged to form the capacity control feature profile corresponding to the cluster set obtained by the density peak clustering algorithm. The fusion average is the arithmetic mean of the indicator values ​​at each time point in the time-series data segment of operating indicators. It can be understood that the density peak clustering algorithm can automatically identify the number of clusters, avoiding the limitation of a preset number of categories. Optionally, the fusion averaging operation can also use the median instead of the arithmetic mean to enhance robustness.

[0067] Example 2: See Figure 3 In practical implementation, when real-time acquisition of the current operating data stream of the energy storage system and identification of the matching degree between the current operating data stream and multiple capacity control feature profiles, the real-time acquired current operating data stream is processed using the same preprocessing method as historical data. Preprocessing operations include cleaning the current operating data stream to remove abnormal data points exceeding a reasonable range; normalizing the cleaned current operating data stream to unify its numerical range to between zero and one; dividing the normalized current operating data stream according to a preset time window to generate current data segments with the same structure as historical operating indicator time-series data segments; extracting features from the current data segments to obtain the current operating feature vector; and calculating the Euclidean distance between the current operating feature vector and the feature vector corresponding to each capacity control feature profile, based on all dimensions of the feature vector. The Euclidean distance is then converted into a matching degree score using the following formula:

[0068]

[0069] in: Represents the match score. This represents the calculated Euclidean distance. The number of dimensions of the feature vector. This is an adjustable width parameter used to control the rate at which the score decays with distance. In some embodiments, the number of dimensions of the feature vector... This is a fixed value, determined by the number of features initially selected when constructing the capacity-controlled feature profile. Optional, the width parameter. It can be set to the statistical median of the Euclidean distance between all capacity regulation feature profiles.

[0070] In specific implementation, when extracting the target capacity control feature profile and its associated neighboring capacity control feature profiles with the highest matching degree to the current running data stream from multiple capacity control feature profiles, the capacity control feature profile with the highest matching degree score is selected as the target capacity control feature profile. In specific implementation, the distance between other capacity control feature profiles and the target capacity control feature profile in the feature space is calculated, and the distance metric uses the same Euclidean distance as in the matching degree calculation stage. Capacity control feature profiles with a distance less than the dynamic similarity threshold are determined as neighboring capacity control feature profiles. The dynamic similarity threshold is adaptively adjusted according to the fluctuation of the recent running data stream. The adjustment method is to calculate the mean standard deviation of multiple current running feature vectors obtained within a recent time window. The dynamic similarity threshold is directly proportional to this mean standard deviation. When the fluctuation of the recent running data stream increases, the dynamic similarity threshold is increased accordingly to include more neighboring capacity control feature profiles. In some embodiments, the distance calculation in the feature space can also use Mahalanobis distance to consider the correlation between feature dimensions. Optionally, the length of the time window used to calculate the mean standard deviation can be the same as the length of the time window used when constructing the capacity control feature profile. It is understandable that the adaptive adjustment mechanism of the dynamic similarity threshold allows the size of the neighbor profile set to flexibly respond to the changing trends of the system's operating state.

[0071] Example 3: In specific implementation, when generating a dynamic capacity control strategy based on the data distribution characteristics between the target capacity control feature profile and adjacent capacity control feature profiles, the differences in data distribution between the target capacity control feature profile and each adjacent capacity control feature profile on key operating indicators are analyzed. In specific implementation, the candidate range for the direction and magnitude of energy storage system capacity adjustment is determined based on the data distribution differences. The final capacity adjustment amount and adjustment rate are determined from the candidate range by combining the real-time grid frequency signal and the energy storage system's own state of charge, forming a complete dynamic capacity control strategy. When analyzing the differences in data distribution between the target capacity control feature profile and each adjacent capacity control feature profile on key operating indicators, the mean difference and variance ratio of the target capacity control feature profile and each adjacent capacity control feature profile on power output indicators are calculated. In specific implementation, the distribution overlap area of ​​the target capacity control feature profile and each adjacent capacity control feature profile on efficiency indicators is calculated. The distribution overlap area is obtained by integrating the overlapping part of the probability distribution functions of the two efficiency indicators. The mean difference, variance ratio, and distribution overlap area are weighted and combined to obtain a comprehensive difference index. The weighting combination process uses the formula:

[0072]

[0073] in: Represents the overall degree of difference. The average power output index representing the characteristic profile of target capacity regulation. The average power output index representing the characteristic profile of adjacent capacity regulation. This is the system power reference value. The variance of the power output index, representing the characteristic profile of the target capacity regulation, The variance of the power output index, representing the characteristic profile of adjacent capacity regulation, The overlapping area of ​​the distribution representing efficiency indicators. , , These are preset weighting coefficients. In some embodiments, the power reference value... The system's rated power or historical average power can be used. Optional, the overlapping area of ​​the distribution... The calculation can be achieved through numerical integration.

[0074] In practical implementation, when determining the direction and magnitude candidate range of energy storage system capacity adjustment based on data distribution differences, if the comprehensive difference index shows that the target capacity control characteristic profile tends to have higher power output compared to adjacent capacity control characteristic profiles, then the capacity adjustment direction is set upward. In practical implementation, the upper and lower limits of the capacity adjustment magnitude are linearly mapped based on the magnitude of the comprehensive difference index to form the magnitude candidate range. The linear mapping relationship is determined by predefined slope and intercept parameters. Combining the real-time grid frequency signal and the energy storage system's own state of charge (SOC), the deviation of the real-time grid frequency from the rated frequency is monitored to obtain the current SOC value of the energy storage system. In practical implementation, a fuzzy logic rule table is established with frequency deviation and SOC as input and capacity adjustment correction coefficient as output. The fuzzy logic rule table contains a series of "if-then" conditional statements. The fuzzy logic rule table is used to correct the initial adjustment amount selected from the candidate range to obtain the final adjustment amount. The correction process includes fuzzification input, rule reasoning, and defuzzification output steps. In some embodiments, the determination of the capacity adjustment direction can also comprehensively consider the changing trends of multiple key operating indicators. Optionally, the parameters of the linear mapping function for the magnitude candidate range can be obtained through offline training using historical data.

[0075] Example 4: In specific implementation, when determining the direction and candidate range of capacity adjustment for the energy storage system, if the comprehensive difference index shows that the target capacity control characteristic profile tends to have a higher power output compared to adjacent capacity control characteristic profiles, then the capacity adjustment direction is set upwards. In specific implementation, the upper and lower limits of the capacity adjustment range are linearly mapped based on the magnitude of the comprehensive difference index to form the candidate range of the range. The linear mapping process uses the following formula:

[0076]

[0077] in: This represents the upper or lower limit of the capacity adjustment range. It is a dimensionless proportionality constant. Represents the overall degree of difference. This is the system power baseline value. The deviation between the actual and expected amplitudes of recently successfully implemented control strategies is used as a correction term to fine-tune the amplitude candidate range. The correction term is obtained by calculating the moving average of the difference between the actual adjustment amount of the most recent control measures and the expected adjustment amount obtained by linear mapping from the comprehensive difference index, and this moving average is added to the boundary of the original amplitude candidate range.

[0078] In practical implementation, by combining the real-time grid frequency signal and the energy storage system's own state of charge (SOC), the deviation of the real-time grid frequency from the rated frequency is monitored to obtain the current SOC value of the energy storage system. In this implementation, a fuzzy logic rule table is established, with frequency deviation and SOC as inputs and capacity adjustment correction coefficients as outputs. The fuzzy logic rule table defines the output correction coefficients corresponding to different combinations of fuzzy sets of input variables. The fuzzy logic rule table is used to correct the initial adjustment amount selected from the candidate range. The correction process includes converting the precise frequency deviation and SOC values ​​into the membership degrees of the corresponding fuzzy sets, performing inference based on the fuzzy logic rules, and fuzzifying the resulting fuzzy output solution into precise correction coefficients. See Table 1, which shows a fuzzy logic rule table.

[0079] Table 1: Fuzzy Logic Rule Table

[0080] Frequency offset ambiguity set State of charge ambiguity set Correction factor ambiguity set Negative large Low Positive small Negative small Medium Positive medium Positive small High Negative small Positive large High Negative large

[0081] In some embodiments, the determination of the capacity adjustment direction can be based on a comparison between a comprehensive difference index and a fixed threshold. Optionally, a proportional coefficient... Pre-calibration can be achieved through regression analysis of historical operating data. It is understood that the moving average correction term can adaptively optimize the adjustment range based on recent control effects. In some embodiments, the fuzzy set partitioning of frequency deviation and state of charge can employ a triangular membership function. In specific implementations, the application of the triangular membership function is achieved by defining three key parameters for each fuzzy set of frequency deviation and state of charge: the start point, vertex, and end point of the membership function. The vertex corresponds to the core value of the fuzzy set, and the start and end points determine the range of membership degree variation from zero to one and back to zero. For frequency deviation fuzzy sets such as negative large, negative small, positive small, and positive large, the parameters of their triangular membership function are set based on the historical statistical range of the rated frequency deviation. For state of charge fuzzy sets such as low, medium, and high, the parameters of their triangular membership function are determined based on the safe operating range of the energy storage system's state of charge. The triangular membership function converts precise frequency deviation and state of charge values ​​into membership degrees for each fuzzy set. The membership degree value, between zero and one, indicates the degree to which the precise value belongs to the corresponding fuzzy set, thus providing input for subsequent fuzzy logic rule reasoning.

[0082] See Figure 4The graph visually presents the distribution characteristics of the deviation between actual and expected power at different time points and clarifies the allowable deviation range of ±3kW. Specifically, the yellow bars in the graph represent the power deviation (actual - expected) at each time point, while the red dashed lines mark the upper (+3kW) and lower (-3kW) limits of the allowable deviation, respectively. Analysis shows that the power deviation at most time points is within the allowable range, but at some moments (such as around time point 10), the deviation reaches -6kW, and around time point 80, the deviation reaches 8kW, exceeding the allowable range. These abnormal deviations can serve as one of the triggering criteria for dynamic capacity control strategies. In practical applications, the deviation data in this graph can be used to calibrate the comprehensive difference index of capacity control: by statistically analyzing the operating states corresponding to different deviation ranges, it helps optimize the linear mapping coefficient of the capacity adjustment amplitude. Simultaneously, the fluctuation characteristics of the deviation can also serve as the basis for adaptive adjustment of the dynamic similarity threshold.

[0083] Example 5: In a specific implementation, when executing the dynamic capacity control strategy, the final determined capacity adjustment amount, adjustment direction, and adjustment rate are encapsulated into standard control commands. The encapsulation process of the standard control commands follows the data format specified by the energy storage converter communication protocol. For example, the capacity adjustment amount, adjustment direction, and adjustment rate are mapped to predefined register addresses or data fields in the protocol. In a specific implementation, the encapsulation operation includes converting numerical parameters into binary encoding or string format, and adding necessary frame headers, frame trailers, and checksums to ensure the integrity of command transmission. In some embodiments, the format of the standard control commands can be compatible with multiple industrial communication standards, such as ModbusTCP or CANopen. Optionally, the encapsulation process may also include command sequence numbers and timestamp information for subsequent command tracking and logging.

[0084] In practical implementation, standard control commands are sent to the energy storage converter. This sending operation is achieved through a preset communication interface, which can be wired Ethernet, serial port, or wireless cellular network. The sending process employs a request-response mechanism; after sending the standard control command, the control unit waits for an acknowledgment signal from the energy storage converter. If no acknowledgment signal is received within the timeout period, a retransmission process is triggered. In some embodiments, a highly reliable communication link can be prioritized when sending commands to reduce the probability of transmission errors. Optionally, the communication interface parameters, such as baud rate or IP address, can be configured via a configuration file during system initialization.

[0085] In practice, the actual output power of the energy storage system after the command is executed is monitored and compared with the expected value of the strategy. The difference is compared using the formula:

[0086]

[0087] in: To indicate a comparison of differences, This indicates the actual output power detected. This represents the expected value set in the dynamic capacity control strategy. In specific implementations, the comparison difference is fed back to the strategy generation stage. The feedback data includes the difference value, timestamp, and associated strategy identifier. This data is stored in a historical operation database for subsequent analysis. In some embodiments, the monitoring of actual output power can utilize high-sampling-rate sensor data, and noise interference can be eliminated through filtering. Optionally, the feedback of the comparison difference can trigger real-time parameter adjustments to the strategy generation algorithm, such as modifying the matching weight or similarity threshold of the capacity control feature profile.

[0088] See Figure 5 This study quantifies the optimization effect of feedback adjustments on control errors. Specifically, the error value (left axis) includes the error before feedback (orange line) and the error after feedback (green line), while the error reduction (right axis, blue bars) is the difference between the two. During implementation, the error before feedback is calculated based on the initial expected value of the control strategy, and the error after feedback is the actual output power deviation after command execution. Within the strategy optimization cycle, the error before feedback shows a fluctuating downward trend, while the error after feedback continuously decreases with the progression of the cycle, reflecting the suppressive effect of the feedback mechanism on errors: the error after feedback is approximately 12kW in the initial cycle (e.g., cycle 1), decreasing to approximately 2.5kW in the final cycle (e.g., cycle 19), with the overall error reduction expanding with the progression of the cycle, peaking at approximately 12kW. At the parameter mapping level, the change in error reduction is directly related to the optimization logic of "comparing differences and feeding them back to the strategy generation stage" in the feedback phase. The feedback difference data corrects the matching weights and similarity thresholds of the capacity control feature profile, thereby improving the accuracy of the control strategy.

[0089] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0090] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for dynamic capacity flexible control of an energy storage system, characterized in that, The method includes: Historical operating data of energy storage systems are preprocessed to form standardized time-series data of operating indicators; Based on the time-series data of the aforementioned operating indicators, multiple capacity control feature profiles reflecting different operating states of the system are constructed. Real-time acquisition of the current operating data stream of the energy storage system, and identification of the matching degree between the current operating data stream and the multiple capacity control feature profiles; Extract the target capacity control feature profile that has the highest matching degree with the current running data stream and its associated neighboring capacity control feature profiles from the multiple capacity control feature profiles; A dynamic capacity control strategy is generated based on the data distribution characteristics between the target capacity control feature profile and the adjacent capacity control feature profiles. The dynamic capacity control strategy is implemented to flexibly adjust the output capacity of the energy storage system; The construction of multiple capacity regulation feature profiles reflecting different operating states of the system includes: Feature extraction is performed on each time series data segment of the operational indicator to obtain a feature vector representing the system's operational status within the time window corresponding to the time series data segment of the operational indicator; Density peak clustering algorithm is used to perform cluster analysis on all feature vectors, and each cluster center represents a typical system operating state; The time series data segments of all operating indicators belonging to the same cluster are fused and averaged to form the capacity regulation feature profile corresponding to the cluster set obtained by the density peak clustering algorithm. The dynamic capacity control strategy includes: Analyze the differences in data distribution of key operating indicators between the target capacity control feature profile and each adjacent capacity control feature profile; Based on the differences in data distribution, determine the direction and range of candidate values ​​for energy storage system capacity adjustment; By combining real-time grid frequency signals and the state of charge of the energy storage system itself, the final capacity adjustment amount and adjustment rate are determined from the candidate range to form a complete dynamic capacity control strategy. The analysis of the differences in data distribution of the target capacity control feature profile and each adjacent capacity control feature profile on key operating indicators includes: Calculate the mean difference and variance ratio of the target capacity control feature profile and each neighboring capacity control feature profile in terms of power output index; Calculate the area of ​​overlap between the target capacity regulation feature profile and each neighboring capacity regulation feature profile in terms of efficiency index; The mean difference, variance ratio, and distribution overlap area are weighted and combined to obtain a comprehensive difference index. The comprehensive dissimilarity index is obtained by weighting the mean difference, variance ratio, and distribution overlap area. The weighting process uses the following formula: in: Represents the overall degree of difference. The average power output index representing the characteristic profile of target capacity regulation. The average power output index representing the characteristic profile of adjacent capacity regulation. This is the system power reference value. The variance of the power output index, representing the characteristic profile of the target capacity regulation, The variance of the power output index, representing the characteristic profile of adjacent capacity regulation, The overlapping area of ​​the distribution representing efficiency indicators. These are preset weighting coefficients; The process of determining the candidate range for the direction and magnitude of energy storage system capacity adjustment includes: If the comprehensive difference index shows that the target profile tends to have higher power output compared to neighboring profiles, then the capacity adjustment direction is set; Based on the magnitude of the comprehensive difference index, the upper and lower limits of the capacity adjustment range are linearly mapped to form a candidate range for the range. The linear mapping process uses the following formula: in: This represents the upper or lower limit of the capacity adjustment range. It is a dimensionless proportionality constant. Represents the overall degree of difference. This is the system power reference value; The deviation between the actual and expected magnitudes of recently successfully implemented control strategies is used as a correction term to fine-tune the candidate range of magnitudes. The combination of real-time grid frequency signals and the energy storage system's own state of charge includes: Monitor the deviation of the real-time power grid frequency from the rated frequency; Obtain the current state of charge (SPO) value of the energy storage system; Establish a fuzzy logic rule table with frequency deviation and state of charge as inputs and capacity adjustment correction coefficient as output; The fuzzy logic rule table is used to correct the initial adjustment amount selected from the candidate range.

2. The dynamic capacity flexible control method for an energy storage system according to claim 1, characterized in that, The preprocessing of historical operating data of the energy storage system includes: Clean historical operational data and remove abnormal data points that are clearly outside the reasonable range; The cleaned historical operational data is normalized to unify its numerical range to between zero and one. The normalized data is divided into time windows of preset length, and the data in each time window constitutes a time series data segment of the operating indicator.

3. The dynamic capacity flexible control method for an energy storage system according to claim 1, characterized in that, The identification of the matching degree between the current running data stream and the multiple capacity control feature profiles includes: The real-time collected current running data stream is processed in the same preprocessing method as historical data to obtain the current running feature vector; Calculate the Euclidean distance between the current running feature vector and the feature vector corresponding to each capacity regulation feature profile; The Euclidean distance is inversely converted to obtain the matching score between the current running data stream and each capacity control feature profile.

4. The dynamic capacity flexible control method for an energy storage system according to claim 3, characterized in that, The step of extracting the target capacity control feature profile with the highest matching degree to the current running data stream and its associated neighboring capacity control feature profiles from the plurality of capacity control feature profiles includes: Select the capacity regulation feature profile with the highest matching score as the target capacity regulation feature profile; Calculate the distances between other capacity regulation feature profiles and the target capacity regulation feature profile in the feature space; Capacity control feature profiles whose distance is less than the dynamic similarity threshold are identified as neighboring capacity control feature profiles. The dynamic similarity threshold is adaptively adjusted according to the fluctuation of recent running data streams.

5. The dynamic capacity flexible control method for an energy storage system according to claim 1, characterized in that, The execution of the dynamic capacity control strategy includes: The final determined capacity adjustment amount, adjustment direction, and adjustment rate are encapsulated into standard control commands; The standard control commands are sent to the energy storage converter; The actual output power of the energy storage system after the monitoring command is executed is compared with the expected value of the strategy. The difference is fed back to the strategy generation stage to optimize the subsequent generation of control strategies.

Images