Intelligent energy coordination system and method for large energy storage system

Abstract

The invention relates to the technical field of energy management and control, in particular to an intelligent energy coordination system and method for a large energy storage system. Comprising a data acquisition module used for acquiring related energy consumption data of a shared energy storage power station adjacent to a plurality of energy storage containers; the dynamic feature extraction module is used for obtaining dynamic feature information according to the parasitic energy consumption data; the data acquisition and processing module is used for acquiring four parasitic energy consumption time curves of the energy storage unit and constructing a first analysis unit and a second analysis unit; and the main association unit determination module is used for determining the main association unit according to the four sort sequences of the adjacent energy storage units. According to the method, the limitation of traditional single threshold monitoring is broken through, and deep insight of the health state and the operation mode of the energy storage unit from macroscopic to microscopic is realized by constructing parasitic energy consumption time curves in four operation states and extracting fine features of non-stationary data segments.

Description

Technical Field

[0001] This invention relates to the field of energy management and control technology, and in particular to an intelligent energy coordination system and method for large-scale energy storage systems. Background Technology

[0002] With the increasing penetration of renewable energy and the growing demand for flexible power system regulation, large-scale energy storage systems, especially clusters composed of shared energy storage power stations and modular energy storage containers, have become key infrastructure supporting the stable operation of new power systems. In such large-scale deployment scenarios, due to their physical proximity, electrical coupling, and inter-condition linkage, the operating states of energy storage units are not independent of each other, but rather involve complex mutual influences.

[0003] Traditional energy management strategies for energy storage systems mostly focus on optimizing the scheduling of main charging and discharging power to achieve macro-level grid service goals such as peak shaving, valley filling, and frequency regulation. However, the internal auxiliary equipment necessary to maintain the controllability and availability of the energy storage system itself, such as the battery management system (BMS), thermal management system, and communication modules, generates continuous parasitic energy consumption. Although the value of this energy consumption at a single point is small, its cumulative effect is significant in large-scale clusters, directly affecting the system's net output efficiency and operational economy. Existing monitoring methods typically only measure the total energy consumption of energy storage units or simple zone energy consumption and issue threshold alarms, failing to deeply analyze the dynamic time-varying characteristics of parasitic energy consumption under different operating conditions.

[0004] More importantly, current technology lacks the ability to uncover the potential correlation mechanisms between parasitic energy consumption of different units within a cluster. When an energy storage unit experiences abnormal fluctuations in parasitic energy consumption, these fluctuations stem from its own performance degradation, as well as from related disturbances caused by the transmission of operating states from neighboring units. If these two causes cannot be accurately distinguished and the key influencing units identified, the formulation of coordinated control strategies will lack specificity, making it difficult to achieve the leap from passive response alarms to proactive prediction and collaborative suppression. Therefore, developing an energy management system capable of deeply analyzing the dynamic characteristics of parasitic energy consumption, accurately identifying the correlations between units, and making intelligent predictions and coordination based on this has become an urgent technical requirement for improving the overall energy efficiency and operational robustness of large-scale energy storage clusters. Summary of the Invention

[0005] To overcome the shortcomings of existing energy management methods for energy storage systems, which only monitor parasitic energy consumption at static threshold alarms, lack in-depth analysis of dynamic time-varying characteristics, and cannot identify the correlation and influence of parasitic energy consumption between units within a cluster, this invention provides an intelligent energy coordination system and method for large-scale energy storage systems.

[0006] The technical solution is as follows: An intelligent energy coordination system for large-scale energy storage systems, comprising: The data acquisition module is used to acquire relevant energy consumption data of the shared energy storage power station and multiple energy storage containers arranged adjacent to each other; The dynamic feature extraction module is used to obtain dynamic feature information based on parasitic energy consumption data; The data acquisition and processing module is used to acquire four parasitic energy consumption time curves of the energy storage unit and to construct the first analysis unit and the second analysis unit. The main associated unit determination module is used to identify the main associated units based on four sorting sequences of adjacent energy storage units; The association optimization module is used to perform association prediction and proactive optimization based on the main association units.

[0007] Preferably, the data acquisition module is used to acquire relevant energy consumption data of the shared energy storage power station and multiple energy storage containers arranged adjacent to each other, including: the relevant energy consumption data includes parasitic energy consumption data of the shared energy storage power station and parasitic energy consumption data of the energy storage containers; the parasitic energy consumption data is the total electrical energy consumed by the internal auxiliary equipment necessary for the energy storage system to maintain its own safety, controllability and availability in standby, charging, discharging or dormant states.

[0008] Preferably, the dynamic feature extraction module is used to obtain dynamic feature information based on parasitic energy consumption data, including: generating parasitic energy consumption time curves for each energy storage unit based on the acquired parasitic energy consumption data; detecting non-stationary data segments in the parasitic energy consumption time curves that exceed a preset steady-state threshold in real time; and extracting dynamic feature information of the non-stationary data segments, wherein the dynamic feature information includes the interval time and the duration of exceeding the threshold between adjacent non-stationary data segments within the same energy storage unit.

[0009] Preferably, the data acquisition and processing module is used to acquire four parasitic energy consumption time curves of the energy storage unit and construct a first analysis unit and a second analysis unit, including: establishing a standby state parasitic energy consumption time curve, a charging state parasitic energy consumption time curve, a discharging state parasitic energy consumption time curve, and a dormant state parasitic energy consumption time curve for each energy storage unit; detecting the time curves of each energy storage unit according to preset state thresholds, and marking the time points when the parasitic energy consumption exceeds the corresponding state threshold as threshold-exceeding events; the first analysis unit analyzes the performance degradation and fluctuation patterns of each energy storage unit based on the interval time in the dynamic feature information; the second analysis unit is used to perform cross-unit correlation analysis on the threshold-exceeding events of different energy storage units when the first energy storage unit detects a threshold-exceeding event in the charging state parasitic energy consumption time curve, and constructing a threshold-exceeding time difference matrix between each energy storage unit; the first energy storage unit is an energy storage unit that has an anomaly and needs to be diagnosed and protected.

[0010] Preferably, the first analysis unit analyzes the performance degradation and fluctuation patterns of the unit itself based on the interval time in the dynamic feature information, including: analyzing the parasitic energy consumption fluctuation characteristics of each energy storage unit based on the interval time, the analysis including identifying periodic over-threshold events, shortening interval time, unstable interval time, and synchronization of over-threshold events among multiple energy storage units; when the interval time is stable and periodic, it is determined that the corresponding energy storage unit has decreased temperature control efficiency or degraded auxiliary system performance; when the interval time gradually shortens, it is determined that the corresponding energy storage unit has signs of internal degradation such as increased battery internal resistance and frequent equalization; when the interval time fluctuates irregularly, it is determined that there is external disturbance or abnormal state switching; when the over-threshold events of multiple energy storage units occur synchronously within a preset time window, it is determined to be a system-level anomaly; when the first energy storage unit detects an over-threshold event in the parasitic energy consumption time curve of the charging state, it performs joint detection with the time curves of neighboring energy storage units.

[0011] Preferably, the second analysis unit is used to perform cross-unit correlation analysis on the threshold-exceeding events of different energy storage units when the first energy storage unit detects an over-threshold event in the parasitic energy consumption time curve of the charging state, and to construct a threshold-exceeding time difference matrix among the energy storage units, including: Analyze the standby parasitic energy consumption time curves of neighboring energy storage units. In the standby parasitic energy consumption time curves, within the time window centered on the threshold event, when the threshold event occurs and the parasitic energy consumption time curve of the first energy storage unit rises abnormally in the charging state, the threshold of the first energy storage unit is adjusted, the threshold duration of the threshold event within the time window is obtained, and the first neighboring energy storage units are sorted from largest to smallest according to the threshold duration to obtain the sorting sequence of the first neighboring energy storage units. Analyze the parasitic energy consumption time curve of the charging state of neighboring energy storage units. In the time window centered on the threshold event, when both the neighboring energy storage unit and the first energy storage unit experience a threshold event, implement an overall coordinated charging strategy, obtain the number of times the threshold event occurs within the time window, and sort the neighboring energy storage units from most to least frequent to obtain the second neighboring energy storage unit sorting sequence. Analyze the parasitic energy consumption time curve of the discharge state of the neighboring energy storage unit. In the time window centered on the threshold event, when the neighboring energy storage unit experiences a threshold event, a delay filter interval is added to the first energy storage unit. According to the pre-constructed threshold exceedance time difference matrix, the propagation delay time between the occurrence time of the threshold event and the occurrence time of the threshold event of the first energy storage unit within the time window is obtained as the time difference value. The third neighboring energy storage unit sorting sequence is obtained by sorting the time difference values ​​from smallest to largest. Analyze the parasitic energy consumption time curves of the dormant state of neighboring energy storage units. In the time window centered on the over-threshold event in the dormant state parasitic energy consumption time curve, when the energy consumption of neighboring energy storage units fluctuates, the over-threshold standard of the charging state of the first energy storage unit is increased; and the fourth neighboring energy storage unit sorting sequence is obtained by sorting the energy consumption fluctuations from largest to smallest.

[0012] Preferably, the main associated unit determination module is used to identify the main associated units based on four sorting sequences of neighboring energy storage units, including: calculating the number of intersections of neighboring energy storage units in the four sequences based on the first, second, third, and fourth neighboring energy storage unit sorting sequences; when the number of intersections of the target neighboring energy storage unit reaches a preset threshold, the target neighboring energy storage unit is determined as the main associated unit of the parasitic cause of the charging state of the first energy storage unit, and association prediction and active optimization are performed based on the main associated unit.

[0013] Preferably, the correlation optimization module is used to perform correlation prediction and active optimization based on the main correlation unit, including: analyzing the parasitic energy consumption time series curves generated by the main correlation unit and the first energy storage unit under various operating state combinations according to historical operating data, identifying statistically significant dynamic correlation feature patterns, constructing and training a prediction model based on the dynamic correlation feature patterns, the prediction model taking the curve shape features collected in real time as input to obtain the change path of parasitic energy consumption of the first energy storage unit within a specified time range, and simultaneously providing the confidence interval of the prediction path; and constructing an optimization function based on the prediction model to obtain the optimal coordinated control instruction set.

[0014] Preferably, the step of constructing an optimization function to obtain the optimal coordinated control instruction set based on the prediction model includes: constructing an optimization function based on the prediction model with the objective of minimizing the future parasitic energy consumption peak of the first energy storage unit and the total system disturbance; solving for the optimal coordinated control instruction set for the first energy storage unit and the main associated units; adjusting according to the optimal coordinated control instruction set; monitoring the actual parasitic energy consumption curve in real time; comparing the actual parasitic energy consumption curve with the predicted trajectory; and performing online adaptive calibration of the prediction model based on the comparison results.

[0015] Preferably, an intelligent energy coordination method for large-scale energy storage systems further includes: S1: Obtain relevant energy consumption data of shared energy storage power stations and multiple energy storage containers arranged adjacent to each other; S2: Obtain dynamic characteristic information based on parasitic energy consumption data; S3: Obtain the four parasitic energy consumption time curves of the energy storage unit, and construct the first analysis unit and the second analysis unit; S4: Identify the main associated units based on the four sorting sequences of adjacent energy storage units; S5: Perform association prediction and active optimization based on the main association units.

[0016] Compared with the prior art, the present invention has the following advantages: 1. This invention, at the state perception level, overcomes the limitations of traditional single-threshold monitoring. By constructing parasitic energy consumption time curves under four operating states and extracting fine features from non-stationary data segments, it achieves a deep understanding of the health status and operating mode of energy storage units from macro to micro levels. This enables the system not only to detect anomalies but also to identify anomaly patterns, providing a direct chain of data evidence for judging performance degradation or identifying external disturbances. 2. This invention innovatively designs a collaborative analysis mechanism consisting of a first analysis unit and a second analysis unit at the correlation analysis level. The former focuses on diagnosing fluctuation patterns within the unit, while the latter initiates a deep correlation scan across units when triggered by specific events. By constructing multiple sorting sequences based on different state curves and using their intersection to filter and determine the main correlated units, this scheme achieves precise localization of influence paths within complex clusters. This method effectively distinguishes between self-faults and correlated disturbances, transforming previously ambiguous system-level anomaly alarms into a clear and traceable inter-unit influence relationship map; 3. This invention, at the decision-making and control level, constructs a predictive model based on the identified dynamic correlation feature patterns, enabling forward extrapolation of the parasitic energy consumption evolution path. Based on this predictive capability, an optimization function aimed at minimizing local peak loads and overall disturbances can be dynamically solved, thereby generating a forward-looking and coordinated control command set. This transforms management strategies from lagging, isolated adjustments to proactive, coordinated optimization, significantly reducing the impact of peak loads and disorderly fluctuations in parasitic energy consumption on overall system efficiency. The combined effect of these factors leads to a substantial improvement in the robustness and economy of system-level operation. Through precise correlation prediction and proactive coordination, this solution can suppress the energy consumption amplification effect caused by adverse interactions between units and delay the equipment performance degradation caused by chain reactions. This not only improves the overall energy utilization efficiency of the energy storage cluster but also provides intelligent technical support for its long-term safe, stable, and economical operation. The problem is precisely identified, focusing not only on parasitic energy consumption but also delving into the deep-seated and often overlooked problem of cross-unit, dynamically propagating parasitic energy consumption correlations and disturbances caused by close proximity and thermal coupling. Attached Figure Description

[0017] Figure 1This is a schematic diagram of an intelligent energy coordination system for large-scale energy storage systems according to the present invention. Figure 2 This is a flowchart of an intelligent energy coordination method for large-scale energy storage systems according to the present invention. Detailed Implementation

[0018] 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.

[0019] Example 1: An intelligent energy coordination system for large-scale energy storage systems, such as Figure 1 As shown, it includes: The data acquisition module is used to acquire relevant energy consumption data of the shared energy storage power station and multiple energy storage containers arranged adjacent to each other; The dynamic feature extraction module is used to obtain dynamic feature information based on parasitic energy consumption data; The data acquisition and processing module is used to acquire four parasitic energy consumption time curves of the energy storage unit and to construct the first analysis unit and the second analysis unit. The main associated unit determination module is used to identify the main associated units based on four sorting sequences of adjacent energy storage units; The association optimization module is used to perform association prediction and proactive optimization based on the main association units.

[0020] The relevant energy consumption data includes parasitic energy consumption data of shared energy storage power stations and parasitic energy consumption data of energy storage containers; the parasitic energy consumption data is the total electrical energy consumed by the internal auxiliary equipment necessary for the energy storage system to maintain its own safety, controllability and availability in standby, charging, discharging or hibernation states.

[0021] Relying on high-precision power metering devices deployed on the auxiliary power supply circuits within the shared energy storage power station and each independent energy storage container, the continuously collected energy consumption data constitutes the relevant energy consumption data. Sensor nodes sample the instantaneous power consumed by their respective auxiliary equipment, with the sampling frequency set to 1Hz by default. The sampled data includes unit identifier, UTC timestamp, and instantaneous parasitic power consumption value. It is important to clarify that the relevant energy consumption data here is strictly defined in terms of its content as two independent and complementary parts: one is the parasitic energy consumption data of the shared energy storage power station itself, and the other is the parasitic energy consumption data of each individual energy storage container arranged adjacent to it; the parasitic energy consumption data has a clear physical and functional definition. In the context of this system, it does not refer to the losses generated during the main energy conversion of the energy storage unit, but specifically refers to the total electrical energy consumed by a series of auxiliary devices within the energy storage system to ensure that it remains in a safe, controllable, and available state under the four typical operating conditions of standby, charging, discharging, or hibernation. These auxiliary devices typically include battery management systems, thermal management circulation pumps and fans, safety monitoring sensors, internal control circuits, and communication modules. Therefore, this data essentially represents the basic energy expenditure required to maintain the system's vital signs. After collecting the aforementioned multi-source heterogeneous parasitic energy consumption data, the core processing task of the data acquisition module is to obtain the motion state curves of each shared energy storage power station and multiple energy storage containers based on the relevant energy consumption data. This process is achieved through time-series alignment and state labeling algorithms. The module synchronizes and matches the continuous time-series parasitic energy consumption data of each energy storage unit with the explicit operating condition signals obtained through the upper-level battery management system. Based on this, a motion state curve for each unit is plotted with time as the horizontal axis and parasitic energy consumption power or energy as the vertical axis, clearly marking the standby, charging, discharging, and hibernation state intervals. These four curves serve as the basic data carrier for all subsequent advanced analyses, intuitively reflecting the dynamic evolution of the energy consumption of the internal auxiliary systems of each unit under different working modes, providing a raw data framework for identifying abnormal modes and inter-unit correlations. A close proximity matrix is ​​constructed using the physical location or site topology information of the unit metadata. The close proximity determination rule is based on spatial distance thresholds or topological relationships, such as the same cabinet, the same column, or adjacent columns. The spatial distance threshold is obtained by statistically analyzing all recorded non-stationary energy consumption events that are confirmed to have causal or strong correlations through offline analysis or historical operating data accumulated during the initial operation phase. For each group of correlated events, the actual physical installation distance between the corresponding energy storage units is calculated, thus forming a correlation distance sample set. By performing statistical analysis on this set, the quantiles at a specific confidence level are taken. In this embodiment, the 95th percentile is taken to obtain a statistical distance threshold based on the historical actual correlation strength.

[0022] Based on the acquired parasitic energy consumption data, a parasitic energy consumption time curve for each energy storage unit is generated; non-stationary data segments in the parasitic energy consumption time curve that exceed a preset steady-state threshold are detected in real time; dynamic feature information of the non-stationary data segments is extracted, and the dynamic feature information includes the interval time and the duration of exceeding the threshold between adjacent non-stationary data segments within the same energy storage unit.

[0023] In this embodiment, this step is considered as receiving and confirming the raw data, ensuring that the curve used for feature extraction has undergone the necessary preprocessing, including data cleaning, noise filtering, and state segmentation. Then, the crucial operation of real-time detection of non-stationary data segments in the parasitic energy consumption time curve that exceed a preset steady-state threshold is performed. In this embodiment, the system presets one or more steady-state thresholds for parasitic energy consumption under each operating state. These thresholds are set based on the statistical characteristics of historical normal operating data of each unit, such as the mean plus three standard deviations or physical model estimates. The module compares the current energy consumption value with the corresponding state threshold in real time through a sliding time window. When the energy consumption values ​​of multiple consecutive data points exceed the threshold, the time period is determined to be a non-stationary data segment, indicating that the unit has experienced internal auxiliary energy consumption exceeding normal levels during this period. Feature extraction focuses on two time-series parameters with clear physical and diagnostic significance: first, the duration of exceeding the threshold, i.e., the length of time from the start to the end of a single non-stationary data segment, which reflects the persistence of abnormal energy consumption events; second, the interval time between adjacent non-stationary data segments within the same energy storage unit, i.e., the time difference between the start time of the current non-stationary event and the end time of the previous non-stationary event of the same type, which reveals the frequency and periodicity of abnormal events. These two features together constitute a preliminary quantitative description of the dynamic fluctuation pattern of parasitic energy consumption, providing standardized feature inputs for subsequent fault diagnosis and correlation analysis of the analyzed units.

[0024] For each energy storage unit, parasitic energy consumption time curves are established for standby, charging, discharging, and dormant states. Based on preset state thresholds, the time curves of each energy storage unit are detected, and time points where parasitic energy consumption exceeds the corresponding state threshold are marked as threshold-exceeding events. A first analysis unit analyzes the performance degradation and fluctuation patterns of each energy storage unit based on the interval time in the dynamic feature information. A second analysis unit performs cross-unit correlation analysis on threshold-exceeding events of different energy storage units when the first energy storage unit detects a threshold-exceeding event in its charging parasitic energy consumption time curve, constructing a threshold-exceeding time difference matrix between energy storage units. The first energy storage unit is the energy storage unit that has experienced an anomaly and requires diagnosis and protection.

[0025] First, based on the operating status records of the energy storage unit, parasitic energy consumption time curves are constructed for standby, charging, discharging, and dormant states. In actual operation, after receiving the real-time parasitic energy consumption sample value from the energy storage unit, the data acquisition and processing module synchronously reads the corresponding operating status identifier, thereby accurately classifying the data at each time point into the corresponding curve among the four types of parasitic energy consumption time curves.

[0026] Based on the constructed curves, the data acquisition and processing module performs state threshold detection on each parasitic energy consumption time curve. The system pre-sets state thresholds for different operating states, such as a steady-state threshold for standby, a heat dissipation load threshold for charging, a drive load threshold for discharging, and a minimum auxiliary energy consumption threshold for hibernation. When the parasitic energy consumption value of a certain energy storage unit exceeds the corresponding state threshold in a certain state, the data acquisition and processing module automatically identifies that time point as an over-threshold event and records the timestamp, amplitude, and the state curve number to which the event belongs. Subsequently, based on the recorded over-threshold events, the data acquisition and processing module divides the analysis task into two types of analysis components: the first analysis unit and the second analysis unit. The first analysis unit is mainly used to process the unit's own behavioral characteristics. The system analyzes the parasitic energy consumption fluctuation pattern, performance degradation trend, and stability of the energy storage unit itself based on the interval time in the dynamic characteristic information. The second analysis unit is used for behavioral correlation analysis between energy storage units. When the first energy storage unit detects an over-threshold event in its parasitic energy consumption time curve during its charging state, the second analysis unit initiates cross-unit detection logic to compare the over-threshold events occurring in neighboring energy storage units within the same analysis window, thereby constructing a threshold over-threshold time difference matrix. This threshold over-threshold time difference matrix is ​​used to express the time difference structure of the occurrence times of over-threshold events between energy storage units. The standby parasitic energy consumption mainly consists of the basic maintenance power consumption of the temperature control system, the low-power monitoring module, and the safety protection unit, and its energy consumption fluctuation range is relatively limited. Therefore, in this embodiment, the original sequence of standby parasitic energy consumption of the energy storage unit is first collected in multiple monitoring cycles, and statistical modeling is performed on the continuous operating cycle. The specific steps are as follows: Steady-state extraction: High-value points caused by temporary state switching, such as instantaneous communication load increase and temperature control pulse action, are removed from the standby parasitic energy consumption sequence, and the stable segment is obtained by the sliding window method. Power consumption distribution modeling: A probability distribution model is constructed using the stable segment data, such as fitting a normal distribution or using a quantile-based statistical method. Threshold calculation: An upper confidence boundary is selected from the distribution model, such as the 95th percentile of the distribution or the mean plus two standard deviations, as the standby state threshold. The standby state threshold obtained in this way can effectively reflect the upper limit of parasitic energy consumption level of the energy storage unit under basic operating conditions, ensuring sufficient sensitivity to abnormal temperature control actions or hidden faults; Parasitic energy consumption during charging is significantly affected by the dynamic adjustment of the cooling system, the amplitude of the charging current, and the rate of battery temperature rise. Therefore, this embodiment obtains a more robust threshold through piecewise modeling. The steps include: Charging condition normalization: Parasitic energy consumption is normalized according to the real-time charging rate to make the energy consumption characteristics under different rates comparable. Temperature rise energy consumption coupling analysis: Based on the temperature change rate collected by the temperature sensor, the coupling characteristics of temperature control power consumption and temperature rise rate are estimated using linear regression or locally weighted regression. Normal operation bandwidth determination: After statistical analysis of multiple charging cycles, a normal range between the temperature control operation frequency and parasitic energy consumption is constructed, and the upper bound of the range is used as the charging state threshold. The above threshold can be used to identify the systemic risks of decreased cooling efficiency and abnormal cell temperature rise caused by battery aging during charging. Under discharge conditions, parasitic energy consumption is mainly affected by the discharge rate, internal heat generation in the battery, and the inverter unit's operating mode. This embodiment employs a joint feature modeling approach to determine the threshold, specifically including: Discharge rate stratification: dividing parasitic energy consumption data into multiple levels based on different discharge rates; Steady-state analysis within each level: extracting the steady-state segment of the parasitic energy consumption sequence for each rate level and calculating its average energy consumption range; Cross-level feature fusion: considering the energy consumption differences between levels, obtaining the overall discharge state threshold through weighted synthesis, where the weights are determined by the frequency of different discharge rates. The threshold constructed in this way can adapt to multiple discharge modes and effectively detect insufficient heat dissipation and abnormal inverter operation during the discharge process. Parasitic energy consumption is most stable during hibernation, primarily originating from minimal monitoring and protection links. This invention employs a more stringent threshold calculation method, focusing on identifying latent faults such as temperature control failure and incomplete hibernation. The steps are as follows: Deep stable segment extraction: Extracting long-term stable segments from the parasitic energy consumption time curve during hibernation, typically exceeding twice the task cycle. Noise baseline estimation: Calculating the minimum range of stable segment data and further eliminating occasional pulse points using a median filter. Threshold setting: Using the mean of the stable segment plus the minimum fluctuation amplitude as the hibernation state threshold. This threshold accurately reflects the ideal upper limit of energy consumption during hibernation and is suitable for detecting incomplete hibernation and residual temperature control actions.

[0027] The parasitic energy consumption fluctuation characteristics of each energy storage unit are analyzed based on the interval time. The analysis includes identifying periodic over-threshold events, shortening interval time, unstable interval time, and synchronization of over-threshold events among multiple energy storage units. When the interval time is stable and periodic, it is determined that the corresponding energy storage unit has decreased temperature control efficiency or auxiliary system performance degradation. When the interval time gradually shortens, it is determined that the corresponding energy storage unit has signs of internal degradation such as increased battery internal resistance and frequent equalization. When the interval time fluctuates irregularly, it is determined that there is external disturbance or abnormal state switching. When the over-threshold events of multiple energy storage units occur synchronously within a preset time window, it is determined to be a system-level anomaly. When the first energy storage unit detects an over-threshold event in the parasitic energy consumption time curve of the charging state, it is jointly detected with the time curves of the adjacent energy storage units.

[0028] When the system identifies a highly stable periodicity in the intervals between threshold events in a certain unit, for example, each abnormal fluctuation occurs approximately every 4 hours, the first analysis unit will activate the corresponding diagnostic rules. At this point, it is determined that the corresponding energy storage unit has experienced a decrease in temperature control efficiency or a degradation in the performance of its auxiliary system. The underlying logic is that periodicity often points to a controlled system, such as a temperature control system whose adjustment cycle shortens regularly due to performance degradation. When analysis reveals that the interval between anomalies in a certain unit is gradually shortening, for example, from once every 8 hours to once every 6 hours and then once every 4 hours, the first analysis unit determines that the corresponding energy storage unit shows signs of internal degradation, such as increased battery internal resistance and more frequent equalization. The reasoning is that the shortened interval reflects accelerated deterioration of the battery itself, leading to the need for more frequent activation of internal equalization or thermal management to cope with it, thereby increasing the frequency of parasitic energy consumption. If the time intervals fluctuate irregularly, sometimes long and sometimes short, without a clear pattern, it indicates the presence of external disturbances or abnormal state transitions. This usually means that the anomaly is not caused by slow, evolving degradation within the unit, but rather by irregular power grid shocks, random changes in ambient temperature, or non-periodic commands from the upper-level control system.

[0029] When monitoring detects that multiple energy storage units experience threshold-exceeding events simultaneously within a preset time window—for example, three adjacent units successively reporting anomalies within a 2-minute time window—the first analysis unit will escalate the alarm level and classify it as a system-level anomaly. This indicates the existence of a common external trigger affecting more than a single unit, such as a failure in the shared cooling system within the station, a sudden voltage drop on the grid side, or regional electromagnetic interference. When the first energy storage unit detects an over-threshold event in its parasitic energy consumption time curve during charging, the analysis work of the first analysis unit is not performed in isolation; it needs to be jointly detected with the time curves of neighboring energy storage units. In this embodiment, this means that while analyzing its own performance mode, the first analysis unit will simultaneously retrieve various state curves of other units identified by the system as adjacent within the same time period for rapid cross-validation. For example, if its own analysis points to internal degradation, but joint detection finds that neighboring units also exhibit abnormal discharge states within the same time window, this initial judgment will be questioned, and more weight will be allocated to the hypothesis of correlational perturbations caused by the discharge operations of neighboring units. This provides important preliminary clues for the second analysis unit to initiate more complex cross-unit correlation analysis.

[0030] Analyze the standby parasitic energy consumption time curves of neighboring energy storage units. In the standby parasitic energy consumption time curves, within the time window centered on the threshold event, when the threshold event occurs and the parasitic energy consumption time curve of the first energy storage unit rises abnormally in the charging state, the threshold of the first energy storage unit is adjusted, the threshold duration of the threshold event within the time window is obtained, and the first neighboring energy storage units are sorted from largest to smallest according to the threshold duration to obtain the sorting sequence of the first neighboring energy storage units. Analyze the parasitic energy consumption time curve of the charging state of neighboring energy storage units. In the time window centered on the threshold event, when both the neighboring energy storage unit and the first energy storage unit experience a threshold event, implement an overall coordinated charging strategy, obtain the number of times the threshold event occurs within the time window, and sort the neighboring energy storage units from most to least frequent to obtain the second neighboring energy storage unit sorting sequence. Analyze the parasitic energy consumption time curve of the discharge state of the neighboring energy storage unit. In the time window centered on the threshold event, when the neighboring energy storage unit experiences a threshold event, a delay filter interval is added to the first energy storage unit. According to the pre-constructed threshold exceedance time difference matrix, the propagation delay time between the occurrence time of the threshold event and the occurrence time of the threshold event of the first energy storage unit within the time window is obtained as the time difference value. The third neighboring energy storage unit sorting sequence is obtained by sorting the time difference values ​​from smallest to largest. Analyze the parasitic energy consumption time curves of the dormant state of neighboring energy storage units. In the time window centered on the over-threshold event in the dormant state parasitic energy consumption time curve, when the energy consumption of neighboring energy storage units fluctuates, the over-threshold standard of the charging state of the first energy storage unit is increased; and the fourth neighboring energy storage unit sorting sequence is obtained by sorting the energy consumption fluctuations from largest to smallest.

[0031] This embodiment combines the parasitic energy consumption time curves of four operating states to comprehensively analyze the over-threshold events of neighboring energy storage units and constructs multiple sorting sequences to provide data support for subsequent identification of major related units.

[0032] First, when the parasitic energy consumption time curve of the first energy storage unit in its charging state exceeds a threshold event, the second analysis unit sets a preset time window with the occurrence time of this threshold event as the center time, and analyzes the parasitic energy consumption time curves in the standby state of neighboring energy storage units. Within this time window, if a threshold event also occurs in the standby state parasitic energy consumption time curve of a neighboring energy storage unit, and the charging state parasitic energy consumption time curve of the first energy storage unit shows an abnormal upward trend, the system adjusts the threshold for the first energy storage unit and extracts the threshold-exceeding duration of all threshold-exceeding events within the time window. Subsequently, the second analysis unit sorts the neighboring energy storage units from largest to smallest according to their threshold-exceeding duration within the window, obtaining a sorting sequence of the first neighboring energy storage units. This sorting sequence reflects the correlation between the duration of threshold-exceeding events and the abnormal conditions of the first energy storage unit.

[0033] Next, the second analysis unit further analyzes the parasitic energy consumption time curves of neighboring energy storage units during the charging state. Within the same time window centered on the threshold event, if a neighboring energy storage unit experiences a threshold event in its charging state parasitic energy consumption time curve, and the first energy storage unit is simultaneously in a similar threshold state, then an overall coordinated charging strategy is executed. At this time, the second analysis unit counts the number of threshold events occurring within this time window and sorts them from most frequent to least frequent, forming a second neighboring energy storage unit ranking sequence. This sequence reflects the cooperative behavior pattern between the neighboring energy storage units and the first energy storage unit during the charging state.

[0034] Furthermore, the second analysis unit will analyze the parasitic energy consumption time curves of the discharge state of neighboring energy storage units. Under this state, within a time window centered on the threshold event, if a neighboring energy storage unit experiences a threshold event, a delay filter interval is added to the first energy storage unit, and the propagation delay time of the threshold event is calculated based on a pre-constructed threshold exceedance time difference matrix. The propagation delay time refers to the time difference between the occurrence of the threshold event in the neighboring energy storage unit and the occurrence of the threshold event in the first energy storage unit within the time window. The second analysis unit uses the obtained time difference as the key time difference value and sorts them from smallest to largest to obtain a third neighboring energy storage unit ranking sequence. The smaller the time difference, the closer the state fluctuation relationship between the neighboring energy storage unit and the first energy storage unit.

[0035] Finally, the second analysis unit analyzes the parasitic energy consumption time curves of neighboring energy storage units in their dormant state. Within a set time window, if energy consumption fluctuations occur in the parasitic energy consumption time curves of neighboring energy storage units in their dormant state, the second analysis unit increases the charging state exceeding threshold standard for the first energy storage unit to enhance the sensitivity of anomaly detection. During this process, the magnitude of energy consumption fluctuations for each neighboring energy storage unit within the time window is extracted and sorted from largest to smallest fluctuation amplitude, thus obtaining a fourth neighboring energy storage unit ranking sequence. This sequence is used to assess the impact of dormant state fluctuations on the first energy storage unit.

[0036] Based on the first, second, third, and fourth neighboring energy storage unit sorting sequences, the number of intersections of neighboring energy storage units in the four sequences is calculated. When the number of intersections of the target neighboring energy storage unit reaches a preset threshold, the target neighboring energy storage unit is identified as the main associated unit of the parasitic cause of the charging state of the first energy storage unit, and association prediction and active optimization are performed based on the main associated unit.

[0037] The primary associated unit determination module acquires the first, second, third, and fourth neighboring energy storage unit ranking sequences, and calculates the position of each neighboring energy storage unit in each of the four ranking sequences. When a neighboring energy storage unit appears consecutively in multiple ranking sequences, it indicates a strong correlation with the first energy storage unit under different state trends. The module then calculates the number of intersections for each neighboring energy storage unit in the four ranking sequences. For example, if a neighboring energy storage unit appears simultaneously in the first, second, and third ranking sequences, its intersection count is recorded as three. The module compares these intersection counts with a preset threshold. When the intersection count of a target neighboring energy storage unit reaches or exceeds this threshold, the neighboring energy storage unit is identified as the primary associated unit for the parasitic cause of the first energy storage unit's charging state. After identifying the primary associated unit, the module performs an association prediction and proactive optimization process based on the over-threshold event characteristics, ranking sequence performance, and over-threshold event propagation patterns of the primary associated unit. The correlation prediction process utilizes the fluctuation trends of key correlated units to infer potential parasitic energy consumption changes in the first energy storage unit. Proactive optimization adjusts the scheduling strategy, threshold settings, or operating status of the first energy storage unit in advance when early changes occur in key correlated units, reducing potential anomaly propagation and energy consumption fluctuations. Through this embodiment, the system can accurately identify key correlated units with critical influence on the first energy storage unit from multi-sequence, multi-state, and multi-dimensional data in complex energy storage cluster environments. This identification result enables more targeted intelligent collaborative control, improving the overall stability and energy efficiency of the energy storage system. The method for obtaining the preset quantity threshold involves collecting historical operating data over a period of time during the initial deployment phase or a specific training phase. Using this data, the system offline simulates the entire analysis process of the second analysis unit. For a large number of first energy storage unit charging status exceeding threshold events recorded in history, corresponding first, second, third, and fourth neighboring energy storage unit ranking sequences are generated in batches. For each historical event, the number of intersections of all neighboring units in the corresponding four sequences is calculated, thereby obtaining a statistical distribution sample of the intersection quantity. Subsequently, by combining known root cause analysis reports or using expert rules to label the correlation strength of historical events, and by analyzing the correspondence between the distribution of the number of intersections and the correlation strength of the labels, an optimal discrimination boundary point is determined as the preset quantity threshold.

[0038] Based on historical operating data, the parasitic energy consumption time-series curves generated by the main associated unit and the first energy storage unit under various operating state combinations are analyzed to identify statistically significant dynamic correlation feature patterns. Based on these dynamic correlation feature patterns, a prediction model is constructed and trained. The prediction model takes the curve shape features collected in real time as input to obtain the change path of the parasitic energy consumption of the first energy storage unit within a specified time range and simultaneously provides the confidence interval of the prediction path. Based on the prediction model, an optimization function is constructed to obtain the optimal coordinated control instruction set.

[0039] The correlation optimization module first extracts parasitic energy consumption time-series curves, corresponding operating status identifiers, and related environmental and control variables from the historical database, including the first energy storage unit and the main associated units under different operating state combinations. These variables include ambient temperature, cooling system status, charging and discharging power commands, SOC, and balancing action times. The raw time-series data undergoes unified time alignment, denoising, resampling, and normalization. Missing values ​​are handled using short-time interpolation or removal strategies, and data quality indicators are labeled. All data must meet the requirements of a unified sampling rate and time-scale synchronization before entering the feature identification module. Based on the historical data, the correlation optimization module uses the first energy storage unit's threshold-exceeding events as anchor points to calculate the time-series relationship feature set between the first energy storage unit and the main associated units. Typical dynamic correlation features include: threshold exceedance time difference vector, exceedance duration difference, event frequency difference, exceedance magnitude ratio, short-term energy consumption gradient before and after the event, and Granger causality measure based on multivariate lag sequences. Through statistical tests, such as significance tests, mutual information tests, or Granger causality tests, a subset of features statistically significantly correlated with the parasitic energy consumption changes of the first energy storage unit is selected from these candidate features and used as the input feature set for the prediction model. Based on the selected input features, the correlation optimization module selects an appropriate model structure for training. Optional models include: recurrent neural networks, time-series graphical neural networks, autoregressive moving average models, ensemble learning models, or Bayesian time series models. The model selection depends on engineering requirements. During training, rolling window cross-validation or time series segmentation validation is used to evaluate the model's generalization ability. Sample weights are introduced when necessary to balance rare exceedance event samples. To obtain a measure of the uncertainty in the prediction, one or more of the following schemes are constructed simultaneously during the training phase to obtain the confidence interval: estimating the confidence interval through the distribution of the output of each model after model integration; directly outputting the prediction distribution using a Bayesian neural network or a deep Gaussian process; estimating the prediction error distribution based on the residual bootstrap method. During online operation, the correlation optimization module uses the curve shape characteristics collected in real time, including the parasitic energy consumption values ​​at the current and several historical moments, the recent event characteristics of the main correlated units, environmental quantities, and control quantities as inputs to the prediction model to obtain the prediction path of the parasitic energy consumption of the first energy storage unit within a specified time range; and simultaneously provides the confidence interval of the prediction path so that the optimization module can formulate control strategies under risk constraints.

[0040] Provide a schematic diagram of the mathematical expression for the output of the prediction model, along with an explanation of the symbols: In the formula For the information set at time t0 The parasitic energy consumption prediction value of the first energy storage unit i at a future time t is given below; For a trained prediction model, the parameter set is: ; The input feature vector, based on time t0, contains dynamic correlation features, environmental variables, control variables, and historical parasitic energy consumption fragments from the main correlated units; T p To predict the time domain length, the unit is consistent with the sampling period.

[0041] Based on the prediction model, an optimization function is constructed with the objective of minimizing the future parasitic energy consumption peak of the first energy storage unit and the total system disturbance. Solving this function yields the optimal coordinated control command set for the first energy storage unit and the main associated units. Adjustments are made according to the optimal coordinated control command set, and the actual parasitic energy consumption curve is monitored in real time. The actual parasitic energy consumption curve is compared with the predicted trajectory, and the prediction model is adaptively calibrated online based on the comparison results. The decision variable of the optimization problem is the sequence of coordinated control commands for the first energy storage unit and the identified main associated units in the control time domain, denoted as... In the formula, To coordinate the sequence of control commands; For control commands to the first energy storage unit, such as charging power commands, threshold adjustment range, and equalization action trigger flags; For the j-th primary related unit k Control commands; T c To control the time domain length, the value is set to be equal to or less than the prediction time window. In this embodiment, to simultaneously minimize the parasitic energy consumption peak of the target unit and the total system disturbance, the following objective function is constructed: In the formula, is the objective function value, a scalar; α, β, γ are non-negative weighting coefficients used to weigh the maximum parasitic peak value, the total system disturbance, and the control effort. In order to implement a given control strategy u Under the influence of the control command, the predicted value of the parasitic energy consumption of the first energy storage unit is calculated by incorporating the influence of the control command on the parasitic energy consumption into the model or approximating it through a simulation function. Let be the disturbance of the parasitic energy consumption of the Kth unit relative to its reference trajectory; The L2 norm squared of the control energy or control amplitude is used as the control cost term. The first term in the objective function is the peak minimization term, which highlights the suppression of peak events. The second term is the time integral term of the total system disturbance, which is used to suppress the cumulative disturbance of multiple units. The third term is the control regularization term, which prevents excessively frequent or large control actions. Among them, α, β, and γ are obtained by offline optimization algorithm calibration based on historical operating data with the goal of minimizing the long-term comprehensive performance index; or by the system administrator according to different operating priority strategies; or by online adaptive fine-tuning based on the initial calibration value and real-time control effect feedback.

[0042] The constraint optimization problem should satisfy several physical and safety constraints, including: upper and lower bound constraints on element power, SOC range constraints, temperature safety constraints, control rate limits, and communication and execution delay constraints.

[0043] If the objective function and constraints are transformed into a convex quadratic programming problem in the control space after appropriate linearization or convex approximation, an efficient solver, such as a QP solver based on the interior-point method, should be used for real-time solving. If the problem contains significant nonlinearity or nonconvexity, such as a peak minimization term, a rolling optimization strategy based on model predictive control is recommended: at each control cycle t0, the optimal control sequence in the time domain is solved using the predictive model output as input, and the calculation is repeated in the next cycle after the first control action is executed. For situations where online latency and computational resources are limited, complex nonlinear optimizations should be transformed into an approximately solvable form, such as using sequential quadratic programming, mixed-integer linear programming, or heuristic algorithms to obtain an approximate optimal solution. The optimal coordinated control instruction set output by the solver should be decoded into specific execution commands at the control level, such as charging power setting, threshold increase / decrease commands, and equalization trigger commands, and issued by the control execution unit to the corresponding energy storage unit controller or energy management system. During execution, the correlation optimization module should continuously monitor the actual parasitic energy consumption curve and the response of the main correlation units for subsequent error comparison and model calibration.

[0044] Example 2: Based on Example 1, such as Figure 2 As shown, an intelligent energy coordination method for large-scale energy storage systems also includes: S1: Obtain relevant energy consumption data of shared energy storage power stations and multiple energy storage containers arranged adjacent to each other; S2: Obtain dynamic characteristic information based on parasitic energy consumption data; S3: Obtain the four parasitic energy consumption time curves of the energy storage unit, and construct the first analysis unit and the second analysis unit; S4: Identify the main associated units based on the four sorting sequences of adjacent energy storage units; S5: Perform association prediction and active optimization based on the main association units.

[0045] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. An intelligent energy coordination system for large-scale energy storage systems, characterized in that, include: The data acquisition module is used to acquire relevant energy consumption data of the shared energy storage power station and multiple energy storage containers arranged adjacent to each other; The dynamic feature extraction module is used to obtain dynamic feature information based on parasitic energy consumption data; The data acquisition and processing module is used to acquire four parasitic energy consumption time curves of the energy storage unit and to construct the first analysis unit and the second analysis unit. The main associated unit determination module is used to identify the main associated units based on four sorting sequences of adjacent energy storage units; The association optimization module is used to perform association prediction and proactive optimization based on the main association units.

2. The energy intelligent coordination system for large-scale energy storage systems according to claim 1, characterized in that, The data acquisition module is used to acquire relevant energy consumption data of the shared energy storage power station and multiple energy storage containers arranged adjacent to each other, including: the relevant energy consumption data includes parasitic energy consumption data of the shared energy storage power station and parasitic energy consumption data of the energy storage containers; the parasitic energy consumption data is the total electrical energy consumed by the internal auxiliary equipment necessary for the energy storage system to maintain its own safety, controllability and availability in standby, charging, discharging or hibernation states.

3. The energy intelligent coordination system for large-scale energy storage systems according to claim 2, characterized in that, The dynamic feature extraction module is used to obtain dynamic feature information based on parasitic energy consumption data, including: generating parasitic energy consumption time curves for each energy storage unit based on the acquired parasitic energy consumption data; detecting non-stationary data segments in the parasitic energy consumption time curves that exceed a preset steady-state threshold in real time; and extracting dynamic feature information of the non-stationary data segments, wherein the dynamic feature information includes the interval time and the duration of exceeding the threshold between adjacent non-stationary data segments within the same energy storage unit.

4. The energy intelligent coordination system for large-scale energy storage systems according to claim 1, characterized in that, The data acquisition and processing module is used to acquire four parasitic energy consumption time curves of the energy storage unit and construct a first analysis unit and a second analysis unit, including: establishing a standby state parasitic energy consumption time curve, a charging state parasitic energy consumption time curve, a discharging state parasitic energy consumption time curve, and a dormant state parasitic energy consumption time curve for each energy storage unit; detecting the time curves of each energy storage unit according to preset state thresholds, and marking the time points when the parasitic energy consumption exceeds the corresponding state threshold as threshold-exceeding events; the first analysis unit analyzes the performance degradation and fluctuation patterns of each energy storage unit based on the interval time in the dynamic feature information; the second analysis unit is used to perform cross-unit correlation analysis on the threshold-exceeding events of different energy storage units when the first energy storage unit detects a threshold-exceeding event in the charging state parasitic energy consumption time curve, and constructing a threshold-exceeding time difference matrix between each energy storage unit; the first energy storage unit is an energy storage unit that has an anomaly and needs to be diagnosed and protected.

5. The energy intelligent coordination system for large-scale energy storage systems according to claim 4, characterized in that, The first analysis unit analyzes the performance degradation and fluctuation patterns of the unit itself based on the interval time in the dynamic feature information, including: analyzing the parasitic energy consumption fluctuation characteristics of each energy storage unit based on the interval time, the analysis including identifying periodic over-threshold events, shortening interval time, unstable interval time, and synchronization of over-threshold events among multiple energy storage units; when the interval time is stable and periodic, it is determined that the corresponding energy storage unit has a decrease in temperature control efficiency or a degradation in auxiliary system performance; when the interval time gradually shortens, it is determined that the corresponding energy storage unit has signs of internal degradation such as increased battery internal resistance and frequent equalization; when the interval time fluctuates irregularly, it is determined that there is an external disturbance or an abnormal state switching; when the over-threshold events of multiple energy storage units occur synchronously within a preset time window, it is determined to be a system-level anomaly; when the first energy storage unit detects an over-threshold event in the parasitic energy consumption time curve of the charging state, it performs joint detection with the time curves of neighboring energy storage units.

6. The energy intelligent coordination system for large-scale energy storage systems according to claim 4, characterized in that, The second analysis unit is used to perform cross-unit correlation analysis on the threshold-exceeding events of different energy storage units when the first energy storage unit detects an over-threshold event in the parasitic energy consumption time curve of the charging state, and to construct a threshold-exceeding time difference matrix among the energy storage units, including: Analyze the standby parasitic energy consumption time curves of neighboring energy storage units. In the standby parasitic energy consumption time curves, within the time window centered on the threshold event, when the threshold event occurs and the parasitic energy consumption time curve of the first energy storage unit rises abnormally in the charging state, the threshold of the first energy storage unit is adjusted, the threshold duration of the threshold event within the time window is obtained, and the first neighboring energy storage units are sorted from largest to smallest according to the threshold duration to obtain the sorting sequence of the first neighboring energy storage units. Analyze the parasitic energy consumption time curve of the charging state of neighboring energy storage units. In the time window centered on the threshold event, when both the neighboring energy storage unit and the first energy storage unit experience a threshold event, implement an overall coordinated charging strategy, obtain the number of times the threshold event occurs within the time window, and sort the neighboring energy storage units from most to least frequent to obtain the second neighboring energy storage unit sorting sequence. Analyze the parasitic energy consumption time curve of the discharge state of the neighboring energy storage unit. In the time window centered on the threshold event, when the neighboring energy storage unit experiences a threshold event, a delay filter interval is added to the first energy storage unit. According to the pre-constructed threshold exceedance time difference matrix, the propagation delay time between the occurrence time of the threshold event and the occurrence time of the threshold event of the first energy storage unit within the time window is obtained as the time difference value. The third neighboring energy storage unit sorting sequence is obtained by sorting the time difference values ​​from smallest to largest. Analyze the parasitic energy consumption time curves of the dormant state of neighboring energy storage units. In the time window centered on the over-threshold event in the dormant state parasitic energy consumption time curve, when the energy consumption of neighboring energy storage units fluctuates, the over-threshold standard of the charging state of the first energy storage unit is increased; and the fourth neighboring energy storage unit sorting sequence is obtained by sorting the energy consumption fluctuations from largest to smallest.

7. An intelligent energy coordination system for large-scale energy storage systems according to claim 1, characterized in that, The main associated unit determination module is used to identify the main associated units based on four sorting sequences of neighboring energy storage units, including: calculating the number of intersections of neighboring energy storage units in the four sequences based on the first, second, third, and fourth neighboring energy storage unit sorting sequences; when the number of intersections of the target neighboring energy storage unit reaches a preset threshold, the target neighboring energy storage unit is determined as the main associated unit of the parasitic cause of the charging state of the first energy storage unit, and association prediction and active optimization are performed based on the main associated unit.

8. The energy intelligent coordination system for large-scale energy storage systems according to claim 1, characterized in that, The correlation optimization module is used to perform correlation prediction and active optimization based on the main correlation unit, including: analyzing the parasitic energy consumption time series curves generated by the main correlation unit and the first energy storage unit under various operating state combinations according to historical operating data, identifying statistically significant dynamic correlation feature patterns, constructing and training a prediction model based on the dynamic correlation feature patterns, the prediction model taking the curve shape features collected in real time as input to obtain the change path of parasitic energy consumption of the first energy storage unit within a specified time range, and simultaneously providing the confidence interval of the prediction path; and constructing an optimization function based on the prediction model to obtain the optimal coordinated control instruction set.

9. The energy intelligent coordination system for large-scale energy storage systems according to claim 8, characterized in that, The step of constructing an optimization function to obtain the optimal coordinated control instruction set based on the prediction model includes: constructing an optimization function based on the prediction model with the objective of minimizing the future parasitic energy consumption peak of the first energy storage unit and the total system disturbance; solving for the optimal coordinated control instruction set for the first energy storage unit and the main associated units; adjusting according to the optimal coordinated control instruction set; monitoring the actual parasitic energy consumption curve in real time; comparing the actual parasitic energy consumption curve with the predicted trajectory; and performing online adaptive calibration of the prediction model based on the comparison results.

10. A method for intelligent energy coordination for large-scale energy storage systems, used to implement the intelligent energy coordination system for large-scale energy storage systems as described in any one of claims 1-9, characterized in that, Also includes: S1: Obtain relevant energy consumption data of shared energy storage power stations and multiple energy storage containers arranged adjacent to each other; S2: Obtain dynamic characteristic information based on parasitic energy consumption data; S3: Obtain the four parasitic energy consumption time curves of the energy storage unit, and construct the first analysis unit and the second analysis unit; S4: Identify the main associated units based on the four sorting sequences of adjacent energy storage units; S5: Perform association prediction and active optimization based on the main association units.

Images