Energy storage residual capacity allocation method suitable for multi-scene application

Abstract

The application discloses a kind of energy storage residual capacity allocation methods suitable for multi-scene application, it is related to multi-scene optimization control technical field, including: based on the multi-dimensional characteristics of scene demand, dynamic scene cluster is generated by improved clustering algorithm, and pre-matching relationship library is formed;According to the construction of pre-matching relationship library space joint cost model, the dynamic constraints of power deviation, capacity difference and geographical distance are comprehensively considered, and initial allocation scheme is generated using hybrid optimization algorithm;Through real-time monitoring the health degree of energy storage unit, residual capacity change rate and communication state, trigger flexible redistribution mechanism;When energy storage capacity approaches critical value, allocation proportion is dynamically reduced according to priority;When detecting sudden power fluctuation, enable adjacent energy storage collaborative compensation;Using reinforcement learning model to predict capacity change trend, adjust allocation strategy in advance, output dynamic optimization scheme and adjustment log;Multi-dimensional verification is carried out to dynamic optimization scheme.

Description

Technical Field

[0001] This invention relates to the field of multi-scenario optimization control technology, and in particular to a method for allocating remaining energy storage capacity that is adaptable to multi-scenario applications. Background Technology

[0002] With the increasing penetration of new energy sources and the construction of new power systems, energy storage capacity optimization and configuration technology has become a key to supporting the stable operation of the power grid and the consumption of renewable energy. Existing research focuses on hybrid energy storage structure design, multi-objective collaborative optimization and scenario adaptability improvement. It has formed capacity combination optimization based on cooperative game theory, multi-objective models that consider low carbon emissions and load fluctuation balance, and distributed energy storage configuration methods based on greedy algorithms. For multi-scenario needs, scholars have proposed to build objective function libraries and dynamic weight allocation mechanisms for different scenarios, and introduce reinforcement learning to predict capacity change trends. However, existing technologies still have problems such as static scenario division and rigid response strategies, which make it difficult to meet the dynamic allocation needs of multi-entity collaborative scenarios of source-grid-load-storage.

[0003] Current typical methods, such as grid-side energy storage planning based on a fixed ratio, battery selection models with a single economic objective, and capacity configuration schemes driven by traditional genetic algorithms, have limitations, mainly in that the scenario feature library uses an offline clustering algorithm, which cannot perceive power fluctuation characteristics and changes in geographical constraints in real time. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method for allocating remaining energy storage capacity that is adaptable to multiple application scenarios, solving the problem that the scenario feature library uses an offline clustering algorithm, which cannot perceive power fluctuation characteristics and changes in geographical constraints in real time.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for allocating surplus energy storage capacity adaptable to multiple application scenarios, comprising:

[0008] Based on the multidimensional characteristics of scenario requirements, dynamic scenario clusters are generated through an improved clustering algorithm. Each dynamic scenario cluster is automatically associated with energy storage units that meet the capacity and response conditions, forming a pre-matching relationship library.

[0009] A spatiotemporal joint cost model is constructed based on a pre-matching relational database. Taking into account the dynamic constraints of power deviation, capacity difference and geographical distance, a hybrid optimization algorithm is used to generate an initial allocation scheme.

[0010] By monitoring the health status, remaining capacity change rate, and communication status of energy storage units in real time, an elastic redistribution mechanism is triggered.

[0011] When the energy storage capacity approaches the critical value, the allocation ratio is dynamically reduced according to priority.

[0012] When a sudden power fluctuation is detected, nearby energy storage is activated for collaborative compensation.

[0013] A reinforcement learning model is used to predict capacity change trends, adjust allocation strategies in advance, and output dynamic optimization schemes and adjustment logs.

[0014] The dynamic optimization scheme was validated from multiple dimensions.

[0015] As a preferred embodiment of the energy storage surplus capacity allocation method adapted to multiple application scenarios described in this invention, wherein: the multidimensional characteristics of the scenario requirements refer to:

[0016] Five types of data are collected for the target scenario: baseline power demand, baseline capacity demand, demand duration distribution parameters, historical power fluctuation characteristics, and scenario priority marking parameters. A multidimensional feature dataset is constructed. The baseline power demand is obtained through real-time power sensors. The baseline capacity demand is generated by fitting historical operating data using a linear regression algorithm. The demand duration distribution parameters are obtained by calculating the mean and standard deviation of the scenario start and end timestamps in the historical logs and then normalizing them periodically. The historical power fluctuation characteristics are the maximum fluctuation coefficient of a specific time window in the power data sequence within a preset time period. The scenario priority marking parameters are dynamically adjusted according to the real-time load level.

[0017] As a preferred embodiment of the energy storage surplus capacity allocation method adaptable to multiple application scenarios described in this invention, the specific steps of generating dynamic scenario clusters through an improved clustering algorithm are as follows:

[0018] The multidimensional feature dataset is input into the improved clustering algorithm. The weight ratios of the power demand baseline value, the capacity demand baseline value, and the spatiotemporal attribute are set. The spatiotemporal attribute only includes the demand duration distribution parameter. A Laplacian matrix is ​​generated based on the similarity matrix. Gaussian filtering is applied to the smoothing of scenarios with a duration standard deviation greater than the statistical threshold. The feature vectors of scenarios with a priority lower than the business rule setting value are reduced in dimensionality. The top K feature vectors are selected to perform K-means clustering and output dynamic scene clusters.

[0019] As a preferred embodiment of the energy storage surplus capacity allocation method adapted to multiple application scenarios described in this invention, the specific steps for forming the pre-matching relational database are as follows:

[0020] The maximum power demand peak and capacity demand extreme value are extracted from the dynamic scene cluster. Candidate energy storage units that meet the requirements of rated power not less than the peak preset multiple, remaining capacity not less than the extreme value preset multiple, and geographical distance not exceeding the set range are selected. The power matching degree, capacity matching degree and comprehensive matching degree are calculated to generate a pre-matching relation library entry containing dynamic scene cluster ID, energy storage unit ID, comprehensive matching degree and geographical location.

[0021] As a preferred embodiment of the energy storage surplus capacity allocation method adapted to multiple application scenarios described in this invention, the specific steps of constructing a spatiotemporal joint cost model based on a pre-matching relational database are as follows:

[0022] The peak power demand, extreme capacity demand, and energy storage unit parameters are extracted from the pre-matching relation library. Power deviation cost, capacity difference cost, and geographical distance cost are set, and a spatiotemporal joint cost model is constructed based on the three types of cost functions.

[0023] As a preferred embodiment of the energy storage surplus capacity allocation method adaptable to multiple application scenarios described in this invention, wherein: the triggering elastic reallocation mechanism specifically comprises:

[0024] The system sets a monitoring cycle to monitor the health of energy storage units, the rate of change of remaining capacity, and communication latency. When the health is lower than the set threshold and the rate of change of capacity exceeds the set rate, the communication latency exceeds the set threshold multiple times in a row, or the health of some units in the cluster is abnormal, the elastic reallocation mechanism is triggered. The allocation ratio is dynamically adjusted according to the priority of comprehensive matching degree, and low-health units are removed. When the real-time power demand exceeds the preset threshold, the system searches for candidate energy storage units within the geographical range and selects the energy storage unit with the lowest comprehensive cost to add preset power compensation.

[0025] As a preferred embodiment of the energy storage surplus capacity allocation method adapted to multiple application scenarios described in this invention, the specific steps of using a reinforcement learning model to predict capacity change trends are as follows:

[0026] Construct a reinforcement learning state space, including historical health curves, current power demand baseline values, capacity demand baseline values, reallocation records, and real-time geographical constraint parameters; the action space is defined as a set of actions for adjusting capacity allocation ratios, including expansion, reduction, and compensation actions;

[0027] A reinforcement learning model is built based on a deep Q-network (DQN), and a strategy for predicting capacity change trends is learned through offline training.

[0028] During online deployment, real-time status parameters are input into the trained reinforcement learning model, which outputs a predicted value of the content change trend within a preset time window and corresponding adjustment strategies.

[0029] As a preferred embodiment of the energy storage surplus capacity allocation method adaptable to multiple application scenarios described in this invention, the multi-dimensional verification of the dynamic optimization scheme specifically includes:

[0030] Verify whether the energy storage health, communication latency, and geographical distance meet the set thresholds, generate full-cycle traceability records, update the abnormal unit status in the pre-matching relational database and trigger alarm logs, and output the association index between the dynamic optimization scheme and the traceability records.

[0031] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the energy storage remaining capacity allocation method adapted to multiple application scenarios as described in the first aspect of the present invention.

[0032] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the energy storage remaining capacity allocation method adapted to multiple application scenarios as described in the first aspect of the present invention.

[0033] The beneficial effects of this invention are as follows: This invention achieves accurate allocation of energy storage capacity across multiple scenarios through multi-dimensional feature clustering and dynamic optimization mechanisms. It constructs a multi-dimensional feature set based on five types of data, including power demand baseline and capacity benchmark values. An improved clustering algorithm is used to generate dynamic scenario clusters, improving scenario segmentation accuracy. A spatiotemporal joint cost model integrates power deviation, capacity difference, and geographical distance constraints, combined with a hybrid optimization algorithm to generate an initial allocation scheme, improving overall matching. Real-time monitoring of health and capacity change rate triggers an elastic reallocation mechanism, prioritizing the removal of low-health units and enabling neighbor-to-neighbor collaborative compensation to shorten the response time to sudden power fluctuations. A deep Q-network is used to construct a reinforcement learning model to predict capacity trends, adjusting strategies in advance to reduce prediction deviation rates. Through health threshold verification and full-cycle traceability record generation, real-time updates of energy storage units and association with scheme traceability indexes are achieved. Attached Figure Description

[0034] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 A flowchart of a method for allocating remaining energy storage capacity to adapt to various application scenarios.

[0036] Figure 2 A flowchart for building a pre-matching relational database.

[0037] Figure 3 This is a schematic diagram of the spatiotemporal joint cost model.

[0038] Figure 4 This is a schematic diagram of the flexible redistribution mechanism. Detailed Implementation

[0039] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0040] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0041] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0042] Example 1, referring to Figures 1 to 4 This is the first embodiment of the present invention, which provides a method for allocating surplus energy storage capacity adaptable to multiple application scenarios, including the following steps:

[0043] S1. Based on the multi-dimensional features of scenario requirements, dynamic scenario clusters are generated through an improved clustering algorithm. Each dynamic scenario cluster is automatically associated with energy storage units that meet the capacity and response conditions, forming a pre-matching relationship library.

[0044] Five types of data are collected for the target scenario: baseline power demand, baseline capacity demand, demand duration distribution parameters, historical power fluctuation characteristics, and scenario priority marking parameters. A multidimensional feature dataset containing complete feature vectors for each scenario is constructed. The baseline power demand is obtained through real-time power sensors. The baseline capacity demand is generated by fitting the scenario's capacity consumption records in historical operation data using a linear regression algorithm. The demand duration distribution parameters are obtained by extracting the scenario's start and end timestamps from historical operation logs, calculating the mean and standard deviation of the duration, and normalizing it according to a 24-hour cycle. The historical power fluctuation characteristics are obtained by collecting the power data sequence of the target scenario over the past 30 days, calculating the fluctuation coefficient for each 5-minute time window. The fluctuation coefficient refers to the mean of the time window, and the maximum fluctuation coefficient is taken as the feature value. The scenario priority marking parameters are preset with initial values ​​by operators based on the importance of the scenario's business and are dynamically adjusted according to the real-time load level. The priority increases by 0.1 for each increase in load level.

[0045] The multidimensional feature dataset is input into the improved clustering algorithm to calculate the similarity weight between scenarios and obtain a similarity matrix. The power demand baseline value accounts for 30% of the weight, the capacity demand baseline value accounts for 30% of the weight, and the spatiotemporal attribute accounts for 40%. The spatiotemporal attribute refers to the distribution parameter of demand duration. Historical power fluctuation feature values ​​and scenario priority label parameters are not included in the similarity calculation.

[0046] The Laplacian matrix is ​​generated based on the similarity matrix, and feature decomposition is performed. Specifically, Gaussian filtering is applied to smooth scenarios with a duration standard deviation greater than a standard deviation threshold, and feature vector dimensionality reduction is performed on scenarios with a priority lower than a priority setting value. The standard deviation threshold is based on historical data statistics, and the priority setting value is based on business rules and real-time status. The top K feature vectors are selected for K-means clustering, outputting a dynamic scenario cluster set. The peak power demand and extreme capacity demand of the scenarios within each dynamic scenario cluster are extracted. The set of energy storage units is traversed, and candidate energy storage units that meet the adaptation conditions are selected. The specific adaptation conditions are as follows:

[0047] The rated power of the energy storage unit is ≥ 1.2 times the peak power demand within the cluster;

[0048] The current remaining capacity of the energy storage unit is greater than or equal to 1.15 times the extreme value of the cluster's capacity requirement;

[0049] The average geographical distance between the energy storage unit and the cluster environment is ≤50km;

[0050] The set of energy storage units that pass the adaptation condition evaluation is associated with each dynamic scenario cluster, and the matching degree score between each energy storage unit and the dynamic scenario cluster is recorded. The matching degree score includes power matching degree, capacity matching degree, and overall matching degree. The matching degree score is calculated as follows:

[0051] Power matching degree = (rated power of energy storage unit - peak power demand within cluster) / rated power of energy storage unit; Capacity matching degree = (current remaining capacity of energy storage unit - extreme value of capacity demand within cluster) / rated capacity of energy storage unit; Overall matching degree = 0.6 × power matching degree + 0.4 × capacity matching degree;

[0052] The pre-matching relational database entries are generated based on the dynamic scene cluster ID, energy storage unit ID, comprehensive matching degree, and geographical location. When generating the pre-matching relational database entries, the extended fields include energy storage unit attributes. Geographical location refers to the actual location of the energy storage unit. Energy storage unit attributes include the rated power, current remaining capacity, and rated capacity of the energy storage unit.

[0053] S2. Construct a spatiotemporal joint cost model based on the pre-matching relation library, comprehensively consider the dynamic constraints of power deviation, capacity difference and geographical distance, and use a hybrid optimization algorithm to generate an initial allocation scheme;

[0054] Extract the peak power demand, extreme capacity demand, rated power, current remaining capacity, rated capacity, geographical coordinates, and comprehensive matching degree of the dynamic scene cluster from the pre-matching relation library to generate an initial allocation candidate set. Multiply the peak power demand and extreme capacity demand by a safety factor of 1.2 and 1.15, respectively, as the input constraint boundary of the spatiotemporal joint cost model.

[0055] The three cost components of the spatiotemporal joint cost model are defined as power bias cost C. p Cost of capacity difference C q and geographical distance cost C d ;

[0056] The relative deviation between the rated power of the candidate energy storage unit and the peak demand of the dynamic scenario cluster is calculated using the following formula:

[0057]

[0058] in, P represents the rated power of the j-th energy storage unit. peak This represents the peak power demand of the dynamic scene cluster;

[0059] The capacity difference cost is calculated based on the available margin between the remaining capacity of the energy storage unit and the extreme value of the cluster capacity demand in dynamic scenarios. The formula is as follows:

[0060]

[0061] Among them, Q max This represents the extreme value of capacity requirement for dynamic scene clusters. This represents the current remaining available capacity of the j-th energy storage unit. This represents the rated capacity of the j-th energy storage unit;

[0062] Using the geographic center of the dynamic scene cluster as a reference, the Euclidean distance cost between the energy storage unit and the cluster center is calculated using the following formula:

[0063]

[0064] Among them, Lat j Lon represents the latitude coordinates of the j-th energy storage unit. j Represents the longitude coordinates of the j-th energy storage unit, Lat c Lon represents the latitude coordinates of the geographic center of a dynamic scene cluster. c Represents the longitude coordinates of the geographic center of the dynamic scene cluster;

[0065] The comprehensive cost function C is calculated based on the power deviation cost, capacity difference cost, and geographical distance cost. total The expression is:

[0066] C total =0.5C p +0.3C q +0.2C d ;

[0067] Set hard constraints on power, capacity and geographical distance. The hard constraints include power constraints, capacity constraints and communication delay constraints. The power constraint means that the rated power of the energy storage unit is greater than or equal to the peak demand of the cluster in the dynamic scenario. The capacity constraint means that the current remaining capacity of the energy storage unit is greater than or equal to the extreme value of the cluster capacity demand. The communication delay constraint means that the signal transmission delay is calculated based on the geographical distance and the delay is required to be ≤10ms.

[0068] A hybrid algorithm strategy combining an improved genetic algorithm and a variable neighborhood search is adopted. In the global search phase, the improved genetic algorithm generates the initial solution set. Chromosomes encode the allocation relationship between energy storage units and dynamic scene clusters. The chromosome encoding rule uses permutation encoding to represent the allocation relationship between energy storage units and dynamic scene clusters. Each chromosome represents an allocation scheme, with gene positions corresponding to dynamic scene cluster IDs and gene values ​​representing a list of energy storage unit IDs assigned to the dynamic scene cluster. Each energy storage unit can only be assigned to one dynamic scene cluster and must meet the fitness condition. Based on the comprehensive matching degree, a roulette wheel selection method is used, with individuals with high matching degrees participating in crossover first. Partial matching crossover (PMX) is performed on the selected parent chromosomes, randomly selecting crossover points to exchange gene fragments, and correcting duplicate allocations. In the case of energy storage unit conflicts, after crossover, gene segments with a power matching degree ≥ 0.8 from the parent generation are retained to ensure that highly fit units are preferentially transmitted. The mutation operation is to randomly replace low-matching units. Further, a comprehensive matching degree threshold (e.g., < 0.5) is set, and energy storage units with a matching degree lower than the comprehensive matching degree threshold in the pre-matching relation library are selected as mutation candidates. Low-matching energy storage units are randomly selected and replaced with energy storage units with a higher comprehensive matching degree within the same dynamic scene cluster. Energy storage units with a geographical distance ≤ 15km are preferentially selected. The mutation rate is set to 10% in the initial stage and gradually reduced to 5% as the number of iterations increases to balance search speed and stability. In the local optimization stage, the initial solution set obtained by global search is finely adjusted using the Variable Neighborhood Search (VNS) algorithm.

[0069] Fine-tuning refers to the exchange of allocations of different energy storage units within the same scenario cluster. The top 20% of energy storage units by comprehensive cost function value are designated as high-cost units, while the bottom 20% that meet the requirements of power matching degree ≥ 0.7 and geographical distance cost ≤ 0.3 (i.e., distance ≤ 15km) are designated as low-cost units. High-cost units are replaced with low-cost units in the same region, and the allocation is redistributed according to geographical distance cost. The output is an allocation scheme that satisfies both the comprehensive cost function and hard constraints. Specifically:

[0070] The comprehensive cost function value is minimized while all hard constraints are satisfied, and the energy storage unit health is ≥0.8. The energy storage unit health is the ratio of the current available capacity to the rated capacity of the energy storage unit.

[0071] The initial allocation scheme entries include dynamic scenario cluster ID, energy storage unit ID list, allocated power value, allocated capacity value, and overall cost.

[0072] S3. By monitoring the health status, remaining capacity change rate, and communication status of the energy storage units in real time, an elastic reallocation mechanism is triggered, specifically as follows:

[0073] The system reads the energy storage unit health status, remaining capacity change rate (obtained by dividing the difference between the current remaining capacity and the previous period by the time interval), and communication status (including signal strength and latency) every minute. A real-time monitoring dataset is generated, containing the dynamic scene cluster ID, energy storage unit ID, health status value, capacity change rate value, and communication latency value. Based on this dataset, an elastic reallocation mechanism is triggered with the following conditions: an energy storage unit health status below 0.85 and an absolute remaining capacity change rate exceeding 0.5% per minute of the rated capacity indicate a health anomaly; a communication latency exceeding 8ms for three consecutive samples indicates a communication anomaly; and a cluster health anomaly occurs when more than 30% of the associated energy storage units within the same dynamic scene cluster have a health status below 0.8. If any condition is met, an elastic reallocation trigger command is sent to the associated dynamic scene cluster. This command includes a list of energy storage unit IDs associated with the affected dynamic scene cluster and the anomaly parameter types, including health anomaly, communication anomaly, and cluster health anomaly.

[0074] When the energy storage capacity approaches a critical value, the allocation ratio is dynamically reduced according to priority. The specific steps are as follows: After receiving the elastic reallocation trigger command, the list of energy storage unit IDs associated with the affected dynamic scene cluster is traversed. The allocation ratio is adjusted by extracting the comprehensive matching degree value of the energy storage units in the pre-matching relation library and generating a priority queue in descending order; for energy storage units with a health level between 0.75 and 0.8, the corresponding energy storage unit allocation capacity value is reduced to 80% of the original value; for energy storage units with a health level below 0.75, they are removed from the energy storage unit ID list of the current dynamic scene cluster, and the pre-matching relation library entries are updated; an allocation ratio update record containing the dynamic scene cluster ID, the adjusted list of energy storage unit IDs, and the reduction ratio value is generated according to the adjusted allocation ratio.

[0075] When a sudden power fluctuation is detected, nearby energy storage collaborative compensation is activated. Specifically, when the real-time power demand peak of a dynamic scenario cluster exceeds 15% of the value recorded in the pre-matching relational database, energy storage units associated with other dynamic scenario clusters are searched within a 50km radius, with the geographical center of the affected dynamic scenario cluster as the origin. A candidate compensation unit list is generated based on the screening conditions of comprehensive matching degree ≥ 0.7 and geographical distance cost ≤ 0.3 (i.e., ≤ 15km). Real-time calculation of power deviation cost and capacity difference cost is performed on the candidate compensation units, and C is selected. total The five units with the smallest values ​​are designated as collaborative compensation units; 20% of the rated power of the collaborative compensation units is temporarily added to the original allocation scheme, where the original allocation scheme refers to the allocation scheme currently being implemented, and a record of the compensation power increment is generated.

[0076] The reinforcement learning model is used to predict capacity change trends, adjust allocation strategies in advance, and output dynamic optimization schemes and adjustment logs. Specifically, the energy storage unit health change curves, power demand baseline values, capacity demand baseline values, elastic reallocation mechanism trigger frequency and adjustment records are input into the reinforcement learning model to output the capacity change prediction values ​​and suggested adjustment strategies for a period of time in the future.

[0077] Recommended adjustment strategies include reducing the allocation ratio of energy storage units with a predicted health rate decline exceeding 0.5% / hour to 90% of the predicted health value in advance; reserving 10%-30% of the compensation unit capacity for dynamic scene clusters predicted to experience sudden power fluctuations, based on the fluctuation amplitude; and generating a dynamic optimization scheme document that includes timestamps, predicted values, adjustment action types, and a list of affected dynamic scene cluster IDs.

[0078] The allocation ratio update records, compensation power increment records, and dynamic optimization scheme documents are merged in chronological order to generate an adjustment log that includes trigger timestamps, operation types (ratio reduction / compensation addition / predictive adjustment), a list of affected dynamic scenario cluster IDs and energy storage unit IDs, a comparison table of power allocation values ​​and capacity allocation values ​​before and after adjustment, and a snapshot of the predicted data output by the reinforcement learning model.

[0079] S4. Verify the dynamic optimization scheme in multiple dimensions, including verifying whether the energy storage health meets the long-term operation requirements, checking whether the communication delay is within the tolerance range, assessing whether the geographical distribution meets the emergency response requirements, and generating a full-cycle traceability record through the verified scheme. The full-cycle traceability record includes the decision basis, adjustment process and anomaly handling log, providing a data closed loop for strategy optimization.

[0080] Extract the health data of energy storage units in the dynamic optimization scheme and verify the long-term operation requirements. Read the current health values ​​(current available capacity / rated capacity) of all energy storage units from the real-time monitoring dataset, and filter out the list of energy storage unit IDs with a health value ≥ 0.8. Count the number of energy storage units with a health value lower than 0.8 and the dynamic scene cluster IDs, and generate a health verification result that includes a list of dynamic scene cluster IDs with abnormal health values, a list of abnormal energy storage unit IDs, and health value values.

[0081] Based on the health verification results, check the communication latency tolerance. Input the list of dynamic scene cluster IDs in the health verification results and extract the communication latency values ​​of the corresponding dynamic scene clusters from the adjustment log. Determine whether the communication latency of all associated energy storage units is ≤10ms. If there are units with a latency >10ms, record the dynamic scene cluster ID, energy storage unit ID, and actual latency value, and generate a communication latency check result containing a list of dynamic scene cluster IDs with abnormal communication latency, a list of abnormal energy storage unit IDs, and latency values.

[0082] Spatial coverage analysis is performed by combining geographical distribution and emergency response requirements. The coordinates of the geographic center of the dynamic scene cluster in the dynamic optimization scheme and the geographic location data of the associated energy storage units are input to calculate the Euclidean distance between the energy storage units and the geographic center of the dynamic scene cluster. A list of energy storage unit IDs with a distance > 50km is selected and associated with the dynamic scene cluster IDs to generate a geographical distribution assessment result that includes a list of dynamic scene cluster IDs with abnormal geographical distribution, a list of abnormal energy storage unit IDs, and the actual distance values.

[0083] Integrate multi-dimensional verification data to generate a full-cycle traceability record. Merge health verification results, communication delay check results, and geographic distribution assessment results by timestamp, and associate them with decision-making basis (such as elastic reallocation trigger command type), adjustment process (proportional reduction / compensation addition / predictive adjustment operation records), and anomaly handling log (handling actions such as health anomalies and communication anomalies) in the adjustment log to generate a full-cycle traceability record containing the following fields:

[0084] Trigger timestamp; list of dynamic scene cluster IDs and energy storage unit IDs; health verification status (pass / fail) and specific value; communication delay status (pass / fail) and actual delay value; geographical distribution status (pass / fail) and actual distance value; index of associated adjustment log entries.

[0085] Based on the full-cycle traceability record, the data closed-loop processing is performed. The full-cycle traceability record is input into the pre-matching relation database. Energy storage units that fail the health test are marked as "to be maintained" and restricted from participating in subsequent allocation. For dynamic scene clusters with abnormal communication delays, their priority marking parameter value is reduced (priority is reduced by 0.1). Energy storage units with geographical distribution exceeding the limit are removed from the pre-matching relation database and alarm logs are triggered.

[0086] Output the updated pre-matched relational database entries and alarm logs as input for the next round of allocation strategy optimization.

[0087] This embodiment also provides a computer device applicable to the energy storage surplus capacity allocation method adapted to multiple application scenarios, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the energy storage surplus capacity allocation method adapted to multiple application scenarios proposed in the above embodiment.

[0088] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0089] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program implements the energy storage remaining capacity allocation method adapted to multiple application scenarios proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0090] In summary, this invention achieves precise allocation of energy storage capacity across multiple scenarios through multi-dimensional feature clustering and dynamic optimization mechanisms. It constructs a multi-dimensional feature set based on five types of data, including power demand baseline and capacity benchmark values. An improved clustering algorithm generates dynamic scenario clusters, enhancing scenario segmentation accuracy. A spatiotemporal joint cost model integrates power deviation, capacity difference, and geographical distance constraints, combined with a hybrid optimization algorithm to generate an initial allocation scheme, improving overall matching. Real-time monitoring of health and capacity change rate triggers an elastic reallocation mechanism, prioritizing the removal of low-health units and enabling neighbor-to-neighbor collaborative compensation to shorten response time to sudden power fluctuations. A deep Q-network constructs a reinforcement learning model to predict capacity trends, adjusting strategies in advance to reduce prediction bias. Through health threshold verification and full-cycle traceability record generation, real-time updates of energy storage units and association with scheme traceability indexes are achieved.

[0091] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for allocating surplus energy storage capacity adaptable to multiple application scenarios, characterized in that: include: Based on the multidimensional characteristics of scenario requirements, dynamic scenario clusters are generated through an improved clustering algorithm. Each dynamic scenario cluster is automatically associated with energy storage units that meet the capacity and response conditions, forming a pre-matching relationship library. A spatiotemporal joint cost model is constructed based on a pre-matching relational database. Taking into account the dynamic constraints of power deviation, capacity difference and geographical distance, a hybrid optimization algorithm is used to generate an initial allocation scheme. The specific steps for constructing the spatiotemporal joint cost model based on the pre-matched relational database are as follows: The peak power demand, extreme capacity demand, and energy storage unit parameters are extracted from the pre-matching relation library. Power deviation cost, capacity difference cost, and geographical distance cost are set, and a spatiotemporal joint cost model is constructed based on the three types of cost functions. By monitoring the health status, remaining capacity change rate, and communication status of energy storage units in real time, an elastic redistribution mechanism is triggered. The triggering elastic reallocation mechanism is specifically as follows: Set a monitoring cycle to monitor the health of energy storage units, the rate of change of remaining capacity, and communication latency. When the health is lower than the set threshold and the rate of change of capacity exceeds the set rate, the communication latency exceeds the set threshold multiple times in a row, or the health of some units in the cluster is abnormal, trigger the elastic reallocation mechanism. Dynamically adjust the allocation ratio according to the priority of comprehensive matching degree and remove low health units. When the real-time power demand exceeds the preset threshold, search for candidate energy storage units within the geographical range and select the energy storage unit with the lowest comprehensive cost to add preset power compensation. When the energy storage capacity approaches the critical value, the allocation ratio is dynamically reduced according to priority. When a sudden power fluctuation is detected, nearby energy storage is activated for collaborative compensation. A reinforcement learning model is used to predict capacity change trends, adjust allocation strategies in advance, and output dynamic optimization schemes and adjustment logs. The dynamic optimization scheme was validated from multiple dimensions.

2. The energy storage surplus capacity allocation method adaptable to multiple application scenarios as described in claim 1, characterized in that: The multidimensional characteristics of the scenario requirements refer to: Five types of data are collected for the target scenario: baseline power demand, baseline capacity demand, demand duration distribution parameters, historical power fluctuation characteristics, and scenario priority marking parameters. A multidimensional feature dataset is constructed. The baseline power demand is obtained through real-time power sensors. The baseline capacity demand is generated by fitting historical operating data using a linear regression algorithm. The demand duration distribution parameters are obtained by calculating the mean and standard deviation of the scenario start and end timestamps in the historical logs and then normalizing them periodically. The historical power fluctuation characteristics are the maximum fluctuation coefficient of a specific time window in the power data sequence within a preset time period. The scenario priority marking parameters are dynamically adjusted according to the real-time load level.

3. The energy storage surplus capacity allocation method adaptable to multiple application scenarios as described in claim 2, characterized in that: The process of generating dynamic scene clusters using an improved clustering algorithm involves the following steps: The multidimensional feature dataset is input into the improved clustering algorithm. The weight ratios of the power demand baseline value, the capacity demand baseline value, and the spatiotemporal attribute are set. The spatiotemporal attribute only includes the demand duration distribution parameter. A Laplacian matrix is ​​generated based on the similarity matrix. Gaussian filtering is applied to the smoothing of scenarios with a duration standard deviation greater than the statistical threshold. The feature vectors of scenarios with a priority lower than the business rule setting value are reduced in dimensionality. The top K feature vectors are selected to perform K-means clustering and output dynamic scene clusters.

4. The energy storage surplus capacity allocation method adaptable to multiple application scenarios as described in claim 3, characterized in that: The specific steps for forming the pre-matching relation database are as follows: The maximum power demand peak and capacity demand extreme value are extracted from the dynamic scene cluster. Candidate energy storage units that meet the requirements of rated power not less than the peak preset multiple, remaining capacity not less than the extreme value preset multiple, and geographical distance not exceeding the set range are selected. The power matching degree, capacity matching degree and comprehensive matching degree are calculated to generate a pre-matching relation library entry containing dynamic scene cluster ID, energy storage unit ID, comprehensive matching degree and geographical location.

5. The energy storage surplus capacity allocation method adaptable to multiple application scenarios as described in claim 1, characterized in that: The specific steps for using a reinforcement learning model to predict capacity change trends are as follows: Construct a reinforcement learning state space, including historical health curves, current power demand baseline values, capacity demand baseline values, reallocation records, and real-time geographical constraint parameters; the action space is defined as a set of actions for adjusting capacity allocation ratios, including expansion, reduction, and compensation actions; A reinforcement learning model is built based on a deep Q-network (DQN), and a strategy for predicting capacity change trends is learned through offline training. During online deployment, real-time status parameters are input into the trained reinforcement learning model, which outputs a predicted value of the content change trend within a preset time window and corresponding adjustment strategies.

6. The energy storage surplus capacity allocation method adaptable to multiple application scenarios as described in claim 5, characterized in that: The multi-dimensional verification of the dynamic optimization scheme specifically includes: Verify whether the energy storage health, communication latency, and geographical distance meet the set thresholds, generate full-cycle traceability records, update the abnormal unit status in the pre-matching relational database and trigger alarm logs, and output the association index between the dynamic optimization scheme and the traceability records.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the energy storage remaining capacity allocation method adapted to multiple application scenarios as described in any one of claims 1 to 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the energy storage surplus capacity allocation method adapted to multiple application scenarios as described in any one of claims 1 to 6.

Images