Intelligent collaborative configuration and high-precision layout method and system for multiple energy storage systems

Abstract

The invention discloses an intelligent cooperative configuration and high-precision layout method and system for multiple energy storage systems, and relates to the technical field of energy storage, and the method comprises the steps: collecting the installed capacity and patent data of multiple energy storage modes, and building a dynamic technical cost prediction model through a two-factor learning curve model; sTL decomposition is adopted to analyze energy storage demand characteristics, demands are predicted based on a three-parameter secondary exponential smoothing model, the lowest coverage demand is determined, and the stability of the model is evaluated in combination with a rolling window and noise injection; constructing a structured input and multi-level constraint system based on the operation data set, and realizing multi-target collaborative prediction by combining a multi-scale convolutional neural network with a dynamic weight loss function; and fusing a simulated annealing algorithm and a neural network feedback mechanism to optimize the layout model, and solving an optimal layout scheme through multi-scene sensitivity analysis. According to the method, system optimization configuration and operation efficiency can be improved, system economy, flexibility and environmental benefits are comprehensively improved, and support is provided for safe and efficient operation of a novel power system.

Description

Technical Field

[0001] This invention relates to the field of energy storage technology, and more specifically to a method and system for intelligent collaborative configuration and high-precision layout of multiple energy storage systems. Background Technology

[0002] In recent years, under the global strategy of "carbon peaking and carbon neutrality," the energy system is undergoing a profound transformation. my country's energy structure has long been dominated by coal, and there is an urgent need to achieve a green and low-carbon transformation through large-scale development of new energy sources and energy storage. Energy storage, as a key bridge connecting power sources and loads, and power grids and users, plays a vital role in peak shaving and valley filling, system frequency regulation, and the integration of renewable energy.

[0003] Currently, the proportion of installed capacity of new energy sources is rapidly increasing, and the complexity of energy system operation is constantly growing. Single energy storage technologies are no longer sufficient to meet the comprehensive demands of systems for capacity, response speed, and economic efficiency. However, in terms of energy storage demand forecasting and system deployment, traditional models mostly rely on static assumptions, making it difficult to reflect the dynamic cost reductions brought about by technological advancements and the competitive relationship between different energy storage methods. Furthermore, demand forecasting is often based on an annual scale, ignoring seasonal fluctuations and random disturbances, resulting in insufficient adaptability between planning results and actual systems.

[0004] Intelligent collaborative configuration and high-precision layout of multiple energy storage systems have become an important direction for future energy storage development. By integrating the characteristics of different energy storage technologies, flexible scheduling and maximization of system benefits can be achieved across multiple time scales and multi-functional scenarios. However, existing technical solutions are mostly limited to capacity configuration of a single energy storage method or optimization analysis under static scenarios, lacking comprehensive consideration of dynamic technological progress and demand fluctuations, and are difficult to adapt to the actual needs of the rapid evolution of energy storage technologies.

[0005] Therefore, how to organically combine technology cost evolution prediction, energy storage demand dynamic analysis and multi-objective optimization model to achieve a comprehensive improvement in system economy, flexibility and environmental benefits, and provide support for the safe and efficient operation of new power systems, is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, the present invention provides a method and system for intelligent collaborative configuration and high-precision layout of multiple energy storage systems, which solves the problems existing in the background technology.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A method for intelligent collaborative configuration and high-precision layout of multiple energy storage systems includes the following steps: S1. Collect the cumulative installed capacity and patent authorization of different energy storage methods, establish a two-factor learning curve model, and then construct a dynamic technology cost prediction model; S2. The STL decomposition method is used to analyze the characteristics of energy storage demand data. Energy storage demand is predicted based on a three-parameter quadratic exponential smoothing model to clarify the minimum demand that the energy storage system needs to cover. At the same time, the stability and robustness of the prediction model are evaluated through rolling window and noise injection methods. S3. Based on the multi-energy storage system operation dataset, construct a structured input and establish a multi-level operation constraint system; use a multi-scale convolutional neural network combined with a dynamic weighted loss function to construct a multi-objective collaborative prediction model for energy storage systems under multiple constraints; S4. By integrating the simulated annealing algorithm with the feedback mechanism of multi-scale convolutional neural networks, a multi-energy storage intelligent collaborative layout optimization model is constructed. Based on multi-scenario sensitivity analysis, the cost reduction trend under different technology learning speeds is predicted, and finally, the optimal layout structure scheme of multi-energy storage technologies that adapts to the needs and balances performance and economy is obtained.

[0008] Optionally, in S1, the two-factor learning curve model uses cumulative installed capacity and patent licensing volume as key parameters for quantifying energy storage costs. A nonlinear regression fitting algorithm is used to solve for the key parameters of energy storage devices, obtaining the technology evolution s-curves for different energy storage devices, i.e.:

[0009] In the formula: For the first t The cumulative installed capacity or cumulative patent grants per year This represents the maximum cumulative installed capacity or the cumulative number of patents granted. A fixed growth rate reflects the speed of technology diffusion; The expected inflection point time; This is an adjustment coefficient used to control the shape of the curve.

[0010] Optionally, in S1, the expression for the technology cost prediction model is:

[0011] In the formula: For the first t Annual unit cost of energy storage Represents the unit energy storage cost in the base period. Representing the t Cumulative installed capacity in [year] This represents the cumulative installed capacity in the base period. Representing the t Cumulative number of patents granted in the year This represents the cumulative number of patents granted in the base period; b An empirical parameter reflecting the unit cost reduction effect of energy storage technology as the cumulative installed capacity expands, used to characterize the learning-by-doing (LBD) effect of energy storage technology.c An empirical parameter reflecting the effect of R&D investment and knowledge accumulation on the unit cost of energy storage technology is used to characterize the learning-by-R&D (LRD) effect.

[0012] Optionally, in S2, the specific steps for energy storage demand forecasting are as follows: Based on Loess's seasonality and trend analysis of energy storage demand data:

[0013] In the formula: For the first Actual energy storage demand in the period It is a seasonal ingredient. As a trend component, For residual terms; The three-parameter quadratic exponential smoothing demand forecasting model uses energy storage demand data as input variables and predicts future demand for multiple types of energy storage through layer-by-layer balancing. The basic changes are as follows:

[0014] The coefficients are calculated using the following formula:

[0015] In the formula: To predict the step size, the coefficients , , Reflecting the first t The basic demand level, the trend of demand changes, and the acceleration of demand changes during the period; , , The first t The first, second, and third smoothing values ​​for each period; a The smoothing coefficient is 0 < a <1.

[0016] Optionally, in S2, the stability and robustness of the prediction model are evaluated, specifically as follows: The initial training and testing windows are as follows:

[0017] After fitting the model to the training data, predict the test period data and calculate the error. for:

[0018] In the formula: For training window, For testing windows, The sliding step size predicted by the model; , The first t Actual and predicted values ​​of energy storage demand during the period; For the number of slides, ; Inject Gaussian noise:

[0019] Calculate the reference error Noise data error and relative error increase :

[0020] In the formula: This represents the proportion of noise intensity. The mean absolute error, The root mean square error, It represents the percentage of mean absolute error; , These are the data after noise injection and the predicted value corresponding to the noise data, respectively. For actual time series data, To follow a normal distribution Gaussian noise, The standard deviation of the noise. These are the predicted values ​​corresponding to the clean data.

[0021] Optionally, in S3, the established multi-level operational constraint system includes: basic constraints on energy storage system operation, energy storage demand constraints, and technology cost constraints; the basic constraints on energy storage system operation include: node construction constraints, system capacity constraints, power transmission and flow constraints, conventional power generation constraints, and power balance constraints.

[0022] Optionally, in S3, the multi-objective collaborative prediction model for energy storage systems under multiple constraints is as follows:

[0023] In the formula: n Types of energy storage Total power loss, For the first t Total annual electricity production The energy storage ratio of total electricity generation. For the first n Energy storage efficiency of different types; The total economic cost, For the first t Annual increase n Installed capacity of energy storage systems For unit investment cost, For unit operation and maintenance costs, For the technology life cycle, r is the discount rate.

[0024] Optionally, in S4, the specific steps of the simulated annealing algorithm are as follows: initialize the solution space and temperature parameters; randomly generate candidate solutions and calculate the objective function value; accept or reject new solutions according to the Metropolis criterion; gradually decrease the temperature until the convergence condition is met; output the Pareto optimal solution set.

[0025] Optionally, in S4, the optimal layout structure scheme for multiple energy storage technologies includes: the capacity size, operation strategy, and system layout location of different energy storage units.

[0026] A multi-energy storage system intelligent collaborative configuration and high-precision layout system, which executes the multi-energy storage system intelligent collaborative configuration and high-precision layout method described above, includes: a technology cost analysis module, an energy storage demand prediction module, a deep learning optimization module, and a multi-objective optimization configuration module; The technology cost analysis module is used to collect the cumulative installed capacity and patent authorization of different energy storage methods, and to establish the cost evolution prediction results of each energy storage technology based on a two-factor learning curve model. The energy storage demand forecasting module is used to collect power load and renewable energy output data, and obtain the dynamic change characteristics of energy storage demand through trend decomposition and seasonal decomposition methods. The deep learning optimization module is used to process temporal feature data in dynamically changing features based on multi-scale convolutional neural networks to form an optimized input vector; The multi-objective optimization configuration module is used to output configuration schemes and layout parameters for multiple energy storage systems, including pumped hydro storage, electrochemical energy storage and hydrogen energy storage, based on the simulated annealing algorithm with the objectives of minimizing cost and power loss.

[0027] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a method and system for intelligent collaborative configuration and high-precision layout of multiple energy storage systems, which has the following beneficial effects: (1) This invention introduces a two-factor learning curve model of cumulative installed capacity and patent authorization, and comprehensively considers the learning-by-doing effect and the R&D-driven learning effect to achieve dynamic prediction of the unit cost of different energy storage technologies. It can accurately depict the long-term impact of technological progress on the economics of energy storage and provide a scientific basis for energy storage investment planning.

[0028] (2) This invention extracts multi-scale features from time-series data of energy storage demand through convolutional neural networks and combines them with dynamic weighted loss functions to achieve multi-objective collaborative prediction. This not only preserves the temporal correlation but also takes into account energy storage performance and economy, providing high-quality input for subsequent optimization.

[0029] (3) This invention establishes a multi-type energy storage joint optimization model covering pumped storage, electrochemical energy storage and hydrogen energy storage. Through multi-scenario sensitivity analysis and simulated annealing iterative solution, the complementarity and synergy between different energy storage types are realized.

[0030] (4) The method of the present invention can adjust the parameters according to the proportion of new energy installed capacity, the rate of technological progress and policy constraints in different regions. It is applicable to multiple scenarios such as parks, regional power grids and national energy storage planning. It has important engineering application value and promotion significance for realizing high proportion of new energy grid connection and low-carbon energy system construction. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0032] Figure 1 Flowchart of the intelligent collaborative configuration and high-precision layout method for multiple energy storage systems provided by the present invention; Figure 2 A flowchart of the simulated annealing method provided by the present invention; Figure 3 The system architecture diagram of intelligent collaborative configuration and high-precision layout of multi-energy storage systems provided by the present invention. Detailed Implementation

[0033] 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.

[0034] To address the lack of dynamic technological advancements, economic efficiency, and comprehensive consideration of environmental benefits in the configuration of multi-energy storage systems, this invention discloses a method for intelligent collaborative configuration and high-precision layout of multi-energy storage systems, such as... Figure 1 As shown, it includes the following steps: S1. Collect the cumulative installed capacity and patent authorization of different energy storage methods, introduce the learning-by-doing effect and R&D-driven learning effect, establish a two-factor learning curve model, and then construct a dynamic technology cost prediction model. S2. The STL decomposition method is used to analyze the characteristics of energy storage demand data. Energy storage demand is predicted based on a three-parameter quadratic exponential smoothing model to clarify the minimum demand that the energy storage system needs to cover. At the same time, the stability and robustness of the prediction model are evaluated through rolling window and noise injection methods. S3. Based on the multi-energy storage system operation dataset, construct structured input and establish a multi-level operation constraint system; use a multi-scale convolutional neural network combined with a dynamic weighted loss function to construct a multi-objective collaborative prediction model for energy storage systems under multiple constraints, so as to achieve collaborative optimization of operation performance and economy; S4. By integrating the simulated annealing algorithm with the feedback mechanism of multi-scale convolutional neural networks, a multi-energy storage intelligent collaborative layout optimization model is constructed. Based on multi-scenario sensitivity analysis, the cost reduction trend under different technology learning speeds is predicted, and finally, the optimal layout structure scheme of multi-energy storage technologies that adapts to the needs and balances performance and economy is obtained.

[0035] Next, for Figure 1 The steps shown are described in detail to further understand the technical solution to be protected by this invention.

[0036] I. Technical Cost Analysis

[0037] Relevant basic data were obtained from power system planning and management departments and new energy dispatch centers. The data obtained includes the following categories: (1) Multi-type energy storage planning data: covering the unit power construction cost, unit capacity construction cost, fixed and variable operation and maintenance costs, charging and discharging efficiency and operation loss coefficient of pumped hydro storage, hydrogen energy storage, lithium battery energy storage, etc. (2) System technical parameters: Based on data on the cost levels of energy storage technologies such as pumped hydro storage, lithium battery energy storage, and hydrogen energy storage collected from the China Carbon Neutrality and Technology Database and the State Intellectual Property Office of China; (3) Operational constraint parameters: covering key data such as node construction constraints, system capacity constraints, power transmission flow constraints, conventional power generation constraints, and power balance constraints; (4) Predicting operational scenario data: Using ArcGIS to overlay regional electricity consumption data with geographic information, presenting the distribution patterns of electricity hotspots and the evolution trend of electricity consumption.

[0038] In S1 of this embodiment, the two-factor learning curve model uses the cumulative installed capacity and patent grants as key parameters for quantifying energy storage costs. A nonlinear regression fitting algorithm is used to solve for the key parameters of the energy storage equipment, obtaining the S-curve of the technological evolution of different energy storage devices. Taking pumped hydro storage, hydrogen energy storage, and lithium battery energy storage as examples, their cumulative installed capacity and cumulative patent grants are expressed as follows:

[0039] In the formula: Q Represents the cumulative installed capacity. PT Represents the cumulative number of patents granted. These correspond to pumped hydro storage, lithium battery energy storage, and hydrogen energy storage, respectively. For the first t The cumulative installed capacity or cumulative patent grants per year This represents the maximum cumulative installed capacity or the cumulative number of patents granted. A fixed growth rate reflects the speed of technology diffusion; The expected inflection point time; This is an adjustment coefficient used to control the shape of the curve.

[0040] Modeling is performed separately for three types of technologies: pumped hydro storage, lithium battery energy storage, and hydrogen storage, to construct a high-precision technology cost prediction model:

[0041] In the formula: For the first t Annual unit cost of energy storage Represents the unit energy storage cost in the base period. Representing the t Cumulative installed capacity in [year] This represents the cumulative installed capacity in the base period. Representing the t Annual cumulative number of patents granted This represents the cumulative number of patents granted in the base period; These correspond to pumped hydro storage, lithium battery energy storage, and hydrogen energy storage, respectively. These are empirical parameters representing the learning-by-doing effect of three energy storage technologies; These represent empirical parameters that illustrate the learning effect driven by the research and development of three energy storage technologies. In practical implementation, these parameters... b , c It can be determined based on historical energy storage project costs, installed capacity, and R&D activity data through regression analysis or empirical fitting methods, and its value can be reasonably set according to different types of energy storage technologies.

[0042] II. Energy Storage Demand Forecast

[0043] STL decomposes and analyzes the characteristic model of energy storage demand data, and uses Loess's seasonality and trend analysis of energy storage demand data:

[0044] In the formula: For the first Actual energy storage demand in the period It is a seasonal ingredient. As a trend component, The residual term is calculated as follows: trend components are extracted using local weighted regression with a sliding window, seasonal components are estimated by smoothing the periodic subsequences, and the residual term is calculated as: original data / (season × trend).

[0045] The three-parameter quadratic exponential smoothing demand forecasting model uses energy storage demand data as input variables and predicts future demand for multiple types of energy storage through layer-by-layer balancing. The basic changes are as follows:

[0046] The coefficients are calculated using the following formula:

[0047] In the formula: To predict the step size, the coefficients , , Reflecting the first t The basic demand level, the trend of demand changes, and the acceleration of demand changes during the period; , , The first t The first, second, and third smoothing values ​​for each period; a The smoothing coefficient is 0 < a <1.

[0048] Furthermore, in S2 of this embodiment, the stability and robustness of the prediction model are evaluated, specifically as follows: (1) Stability model analysis of sliding window The initial training and testing windows are as follows:

[0049] After fitting the model to the training data, predict the test period data and calculate the error. for:

[0050] In the formula: For training window, For testing windows, The sliding step size predicted by the model. , The first t Actual observed and predicted values ​​of energy storage demand in the near term. Number of slides .

[0051] (2) Robustness model analysis of noise injection

[0052] Inject Gaussian noise:

[0053] Calculate the reference error Noise data error and relative error increase :

[0054] In the formula: This represents the proportion of noise intensity. The mean absolute error, The root mean square error, It represents the percentage of mean absolute error; , These are the data after noise injection and the predicted value corresponding to the noise data, respectively. For actual time series data, To follow a normal distribution Gaussian noise, The standard deviation of the noise. These are the predicted values ​​corresponding to the clean data. The relative error increase and error stability are robustness indicators; the smaller the relative error increase and error variance, the better the model's robustness.

[0055] III. Deep Learning Optimization

[0056] Based on the input data from the above steps, a multi-objective collaborative prediction model is established with the goal of optimizing the overall economic efficiency and technological adaptability of the system.

[0057] The model's dual-objective loss function includes: (i) economic costs such as the construction cost of energy storage equipment, fixed and variable operating costs, and conventional generator fuel costs; and (ii) energy loss costs. By weighted summing of the above costs, a comprehensive optimization objective is constructed, with minimizing the total system cost as the core objective function.

[0058] In S3 of this embodiment, to ensure the physical feasibility of the model and the safety of system operation, a multi-level operational constraint system is established, including: basic constraints on the operation of the energy storage system, constraints on energy storage demand, and constraints on technical costs. (i) Energy storage demand constraint, that is, the energy storage capacity of multiple energy storage systems in a given year should meet the energy storage demand.

[0059] (ii) Technical cost constraints, that is, the cost of multi-energy storage systems during the construction and operation process in the current year should be controlled within the range predicted by the model.

[0060] (iii) The basic constraints on the operation of energy storage systems include: node construction constraints, system capacity constraints, power transmission flow constraints, conventional power generation constraints, power balance constraints, etc., to determine the stability and safety of the operation of energy storage systems.

[0061] Node deployment constraints refer to the number and capacity limits of various energy storage devices at each node, which are used to ensure that the equipment configuration does not exceed the node's buildable scale and resource conditions.

[0062] System capacity constraints, namely the requirement that the installed capacity of each type of energy storage equipment cannot exceed the total installed capacity limit, are used to ensure that the overall installed capacity meets the planning requirements.

[0063] Power flow constraints mean that the power flow of each transmission line must not exceed the line's transmission capacity limit and must satisfy the power flow balance relationship to ensure the safe operation of the network.

[0064] Conventional power generation constraints refer to the limitations on the output range and operating characteristics of various conventional generators, including minimum output, maximum output, and ramp-up capability constraints, in order to ensure the safe and stable operation of thermal power units and other similar units.

[0065] Power balance constraint means that the total amount of power generation, energy storage discharge and power purchase in the system is equal to the total amount of load demand, energy storage charging and losses, so as to ensure the balance of power supply and demand in the power system.

[0066] Furthermore, in S3 of this embodiment, the multi-objective collaborative prediction model for the energy storage system under multiple constraints is specifically as follows:

[0067] In the formula: n Types of energy storage Total power loss, For the first t Total annual electricity production The energy storage ratio of total electricity generation. For the first n Energy storage efficiency of different types; The total economic cost, For the first t Annual increase n Installed capacity of energy storage systems For unit investment cost, For unit operation and maintenance costs, For the technology life cycle, r is the discount rate.

[0068] IV. Multi-objective optimization configuration

[0069] In S4, the specific steps of the simulated annealing algorithm are as follows: initialize the solution space and temperature parameters; randomly generate candidate solutions and calculate the objective function value; accept or reject new solutions according to the Metropolis criterion; gradually decrease the temperature until the convergence condition is met; output the Pareto optimal solution set. The optimal layout structure scheme for multi-energy storage technologies includes: the capacity size of different energy storage units, operating strategies, and system layout locations.

[0070] Specifically, such as Figure 2 As shown, the optimization process of the simulated annealing algorithm is as follows: (1) The generation mechanism of new solutions and the determination of parameters of simulated annealing algorithm First, a state generation function is used to generate a new solution. The state generation function serves as the mechanism for generating new solutions in the algorithm. Then, a multi-objective evaluation index is introduced as a quality assessment standard for the solution, as shown below:

[0071] For multi-objective problems, this can be extended to a weighted average. F Value (in) (For example)

[0072] In the formula, For correctly predicted positive class, For the incorrectly predicted positive class, This is a positive class that was missed in the evaluation; F The value is the harmonic mean of precision and recall, used to comprehensively evaluate model performance. To adjust the recall weight, when The era represents a greater emphasis on recall rates.

[0073] (2) Key steps of the algorithm

[0074] By employing simulated annealing algorithm for temperature control and probabilistic acceptance strategy, a robust solution to multi-objective optimization problems is effectively balanced between global exploration and local exploitation capabilities.

[0075] (3) Parameter sensitivity analysis

[0076] Key parameter settings for the simulated annealing algorithm include: initial temperature T0 Set to 15000 to enhance global search capabilities; cooling coefficient A value of 0.9 is used to control the cooling rate and prevent premature convergence; the maximum number of iterations. The algorithm is set to run 150 times to strike a balance between computational efficiency and solution quality. These parameters collectively optimize algorithm performance, ensuring efficient exploration of the solution space and finding high-quality solutions in multi-storage layout problems.

[0077] Furthermore, in S4 of this embodiment, the parameter optimization feedback mechanism based on CNN is as follows: First, a trained CNN model is used to perform a dual-objective evaluation on the candidate solutions generated by the simulated annealing algorithm, outputting predicted energy loss and economic cost levels. Then, the prediction results are used as feedback signals to input into the objective function calculation module of simulated annealing, and the current solution is compared with historical solutions. F The algorithm parameters are dynamically adjusted based on the differences in values. For solutions with significantly worsened power loss, the temperature is increased to enhance global exploration; for solutions with excellent economic performance, the neighborhood is narrowed for a refined local search. Simultaneously, the feature importance analysis results output by the CNN are used to guide the reconstruction of the solution space.

[0078] Furthermore, in step S4 of this embodiment, the sensitivity analysis model based on different progress scenarios is as follows: Technology learning rates are influenced by many factors. Factors affecting the learning-by-doing effect mainly include improvements in employee work skills, management improvements, and equipment quality improvements; while factors affecting the learning rate driven by R&D mainly include knowledge reserves and breakthroughs in key theoretical or technological research. To account for the volatility of technology learning rates, this embodiment sets up scenarios with three technology learning speeds: fast, medium, and slow. Key parameter settings are as follows: In slow learning scenarios, a minimum technology learning rate is set to keep the rate of technological progress at a low level. In a medium-speed learning scenario, a baseline learning rate obtained by fitting historical technology development data is used to reflect the evolution trend of conventional technologies. In a rapid learning scenario, a maximum learning rate is set to simulate the rapid breakthrough and accelerated diffusion of technology.

[0079] For the above dual-objective optimization model, after optimization and solution by simulated annealing algorithm, the cumulative installed capacity of each energy storage technology under three different technological progress scenarios of fast, medium and slow can be obtained, so as to realize the layout optimization of multiple energy storage systems.

[0080] Therefore, after solving the above technical solutions, the output results are: energy storage demand prediction and trend of various types of energy storage equipment under different scenarios, technical cost prediction and trend, multi-energy storage technology installation ratio and high-precision layout.

[0081] Simultaneously calculate the following two key evaluation indicators: 1) Energy utilization rate (ECUL), which reflects the degree to which the system utilizes the time-series transfer of energy storage; 2) Capacity utilization rate (PCUL) is used to reflect the system's utilization of energy storage peak-shaving capacity.

[0082] Both are calculated from the energy storage state of charge and the charging and discharging power timing, respectively, and can be used to measure the comprehensive utilization level of energy storage resources.

[0083] and Figure 1 Corresponding to the method described above, embodiments of the present invention also provide an intelligent collaborative configuration and high-precision layout system for multiple energy storage systems, used for... Figure 1 The specific implementation of the method, the intelligent collaborative configuration and high-precision layout system for multiple energy storage systems provided in this embodiment of the invention, can be applied to computer terminals or various mobile devices, such as... Figure 3 As shown, it specifically includes: a technology cost analysis module, an energy storage demand forecasting module, a deep learning optimization module, and a multi-objective optimization configuration module; The technology cost analysis module is used to collect the cumulative installed capacity and patent authorization of different energy storage methods, and to establish the cost evolution prediction results of each energy storage technology based on a two-factor learning curve model. The energy storage demand forecasting module is used to collect power load and renewable energy output data, and obtain the dynamic change characteristics of energy storage demand through trend decomposition and seasonal decomposition methods. The deep learning optimization module is used to process temporal feature data in dynamically changing features based on multi-scale convolutional neural networks to form an optimized input vector; The multi-objective optimization configuration module is used to output configuration schemes and layout parameters for multiple energy storage systems, including pumped hydro storage, electrochemical energy storage and hydrogen energy storage, based on the simulated annealing algorithm with the objectives of minimizing cost and power loss.

[0084] Furthermore, the technology cost analysis module establishes a two-factor learning curve using historical installed capacity and technological progress factors, and calculates the rate of decrease in unit cost of future energy storage technologies based on cumulative experience coefficients.

[0085] Furthermore, the energy storage demand forecasting module employs seasonal decomposition and exponential smoothing methods to decompose electricity load data and renewable energy output data into trend terms, periodic terms, and stochastic terms, which are used to characterize the long-term changes, seasonal fluctuations, and short-term disturbances in energy storage demand.

[0086] Furthermore, the deep learning optimization module includes a data preprocessing unit and a convolutional feature extraction unit; wherein, the data preprocessing unit is used to normalize and slide window segment the input power load and renewable energy output data, and the convolutional feature extraction unit is used to extract time-series features and output them to the multi-objective optimization configuration module.

[0087] Furthermore, the multi-objective optimization configuration module includes an objective function construction unit and an iterative optimization unit; wherein, the objective function construction unit is used to establish a multi-objective function for the total system cost and energy efficiency level, and the iterative optimization unit uses the simulated annealing algorithm to perform iterative search until the Pareto optimal solution set is obtained.

[0088] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0089] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for intelligent collaborative configuration and high-precision layout of multiple energy storage systems, characterized in that, Includes the following steps: S1. Collect the cumulative installed capacity and patent authorization of different energy storage methods, establish a two-factor learning curve model, and then construct a dynamic technology cost prediction model; S2. The STL decomposition method is used to analyze the characteristics of energy storage demand data. Energy storage demand is predicted based on a three-parameter quadratic exponential smoothing model to clarify the minimum demand that the energy storage system needs to cover. At the same time, the stability and robustness of the prediction model are evaluated through rolling window and noise injection methods. S3. Based on the multi-energy storage system operation dataset, construct a structured input and establish a multi-level operation constraint system; use a multi-scale convolutional neural network combined with a dynamic weighted loss function to construct a multi-objective collaborative prediction model for energy storage systems under multiple constraints; S4. By integrating the simulated annealing algorithm with the feedback mechanism of multi-scale convolutional neural networks, a multi-energy storage intelligent collaborative layout optimization model is constructed. Based on multi-scenario sensitivity analysis, the cost reduction trend under different technology learning speeds is predicted, and finally the optimal layout structure scheme of multiple energy storage technologies that adapts to the needs and balances performance and economy is obtained.

2. The intelligent collaborative configuration and high-precision layout method for multiple energy storage systems according to claim 1, characterized in that, In S1, the two-factor learning curve model uses cumulative installed capacity and patent licensing volume as key parameters for quantifying energy storage costs. A nonlinear regression fitting algorithm is employed to solve for the key parameters of energy storage devices, obtaining the S-curve of technological evolution for different energy storage devices. In the formula: For the first t The cumulative installed capacity or cumulative patent grants per year This represents the maximum cumulative installed capacity or the cumulative number of patents granted. A fixed growth rate reflects the speed of technology diffusion; This is the estimated inflection point time; This is an adjustment coefficient used to control the shape of the curve.

3. The intelligent collaborative configuration and high-precision layout method for multiple energy storage systems according to claim 1, characterized in that, In S1, the expression for the technology cost prediction model is: In the formula: For the first t Annual unit cost of energy storage Represents the unit energy storage cost in the base period. Representing the t Cumulative installed capacity in [year] This represents the cumulative installed capacity in the base period. Representing the t Cumulative number of patents granted in the year This represents the cumulative number of patents granted in the base period; b An empirical parameter reflecting the unit cost reduction effect of energy storage technology as the cumulative installed capacity expands, used to characterize the learning-by-doing effect of energy storage technology; c An empirical parameter reflecting the effect of R&D investment and knowledge accumulation on the unit cost of energy storage technology is used to characterize the learning effect driven by R&D.

4. The intelligent collaborative configuration and high-precision layout method for multiple energy storage systems according to claim 1, characterized in that, In S2, the specific steps for energy storage demand forecasting are as follows: Based on Loess's seasonality and trend analysis of energy storage demand data: In the formula: For the first Actual energy storage demand in the period It is a seasonal ingredient. As a trend component, For residual terms; The three-parameter quadratic exponential smoothing demand forecasting model uses energy storage demand data as input variables and predicts future demand for multiple types of energy storage through layer-by-layer balancing. The basic changes are as follows: The coefficients are calculated using the following formula: In the formula: To predict the step size, the coefficients , , Reflecting the first t The basic demand level, the trend of demand changes, and the acceleration of demand changes during the period; , , The first t The first, second, and third smoothed values ​​for each period; a The smoothing coefficient is 0 < a <1.

5. The intelligent collaborative configuration and high-precision layout method for multiple energy storage systems according to claim 1, characterized in that, In S2, the stability and robustness of the prediction model are evaluated, specifically as follows: The initial training and testing windows are as follows: After fitting the model to the training data, predict the test period data and calculate the error. for: In the formula: For training window, For testing windows, The sliding step size predicted by the model; , The first t Actual and predicted values ​​of energy storage demand during the period; For the number of slides, ; Inject Gaussian noise: Calculate the reference error Noise data error and relative error increase : In the formula: This represents the proportion of noise intensity. The mean absolute error, The root mean square error, It represents the percentage of mean absolute error; , These are the data after noise injection and the predicted value corresponding to the noise data, respectively. For actual time series data, To follow a normal distribution Gaussian noise, The standard deviation of the noise. These are the predicted values ​​corresponding to the clean data.

6. The intelligent collaborative configuration and high-precision layout method for multiple energy storage systems according to claim 1, characterized in that, In S3, the established multi-level operational constraint system includes: basic constraints on energy storage system operation, energy storage demand constraints, and technology cost constraints; the basic constraints on energy storage system operation include: node construction constraints, system capacity constraints, power transmission and flow constraints, conventional power generation constraints, and power balance constraints.

7. The intelligent collaborative configuration and high-precision layout method for multiple energy storage systems according to claim 1, characterized in that, In S3, the multi-objective collaborative prediction model for energy storage systems under multiple constraints is as follows: In the formula: n Types of energy storage Total power loss, For the first t Total annual electricity production The energy storage ratio of total electricity generation. For the first n Energy storage efficiency of different types; The total economic cost, For the first t Annual increase n Installed capacity of energy storage systems For unit investment cost, For unit operation and maintenance costs, For the technology life cycle, r is the discount rate.

8. The intelligent collaborative configuration and high-precision layout method for multiple energy storage systems according to claim 1, characterized in that, In S4, the specific steps of the simulated annealing algorithm are as follows: initialize the solution space and temperature parameters; randomly generate candidate solutions and calculate the objective function value; accept or reject new solutions according to the Metropolis criterion; gradually reduce the temperature until the convergence condition is met; and output the Pareto optimal solution set.

9. The intelligent collaborative configuration and high-precision layout method for multiple energy storage systems according to claim 1, characterized in that, In S4, the optimal layout structure scheme for multiple energy storage technologies includes: the capacity size, operation strategy, and system layout location of different energy storage units.

10. A multi-energy storage system intelligent collaborative configuration and high-precision layout system, characterized in that, The method for intelligent collaborative configuration and high-precision layout of multiple energy storage systems as described in any one of claims 1-9 includes: a technical cost analysis module, an energy storage demand prediction module, a deep learning optimization module, and a multi-objective optimization configuration module; The technology cost analysis module is used to collect the cumulative installed capacity and patent authorization of different energy storage methods, and to establish the cost evolution prediction results of each energy storage technology based on a two-factor learning curve model. The energy storage demand forecasting module is used to collect power load and renewable energy output data, and obtain the dynamic change characteristics of energy storage demand through trend decomposition and seasonal decomposition methods. The deep learning optimization module is used to process temporal feature data in dynamically changing features based on multi-scale convolutional neural networks to form an optimized input vector; The multi-objective optimization configuration module is used to output configuration schemes and layout parameters for multiple energy storage systems, including pumped hydro storage, electrochemical energy storage and hydrogen energy storage, based on the simulated annealing algorithm with the objectives of minimizing cost and power loss.

Images