A method and system for planning and arranging energy storage scale and structure

Abstract

The present invention relates to the field of electric energy storage technology, and in particular to a method and system for planning and layout of energy storage scale and structure. The method comprises the following steps: obtaining original historical electricity load data; performing multi-scale load characteristic analysis based on the original historical electricity load data to generate load characteristic vector data; performing enterprise load pattern matching on the load characteristic vector data to generate enterprise load fluctuation pattern data; performing load demand expected value calculation on the enterprise load fluctuation pattern data to generate multi-scenario load demand forecast data; performing optimal energy storage scale adaptation interval processing based on the multi-scenario load demand forecast data to obtain refined energy storage scale interval data; performing energy storage unit planning and layout optimization based on the refined energy storage scale interval data to obtain optimal energy storage structure layout plan data. The present invention determines the optimal energy storage scale configuration through dynamic analysis of multi-scale electricity load characteristics, thereby realizing a refined layout of the energy storage structure.

Description

Technical Field

[0001] The present invention relates to the technical field of electric energy storage, and in particular to a method and system for planning and laying out energy storage scale and structure. Background Art

[0002] As the scale of industrial parks continues to expand, their electricity load has also increased, placing higher demands on the reliability and stability of the power grid. At the same time, the rapid development of renewable energy in recent years has led to a large number of intermittent power sources in the power system, bringing new challenges to the stable operation of the power grid. On the one hand, the production power load of enterprises in the industrial park fluctuates greatly, with obvious peak-to-valley differences, which has an impact on the stable operation of the power grid and increased the power supply cost; on the other hand, with the promotion and application of renewable energy, its intermittent and volatile characteristics have also placed higher demands on the reliability and stability of the power grid. However, the traditional energy storage structure layout method usually only considers the energy storage capacity and only uses a single type of energy storage equipment, which cannot meet the needs of different load scenarios, resulting in excessive or insufficient energy storage scale configuration, and cannot effectively improve the overall performance of the power grid. Summary of the invention

[0003] Based on this, the present invention provides a method and system for planning and laying out energy storage scale and structure to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a method for planning and layout of energy storage scale and structure includes the following steps:

[0005] Step S1: Obtain original historical power load data; draw a power load curve based on the original historical power load data to generate time series power load data; perform multi-scale load characteristic analysis based on the time series power load data to generate load characteristic vector data;

[0006] Step S2: performing enterprise load pattern matching on load characteristic vector data based on time-series power load data to obtain enterprise load prototype pattern data; performing adaptive hierarchical pattern cutting processing on the enterprise load prototype pattern data to generate enterprise load fluctuation pattern data;

[0007] Step S3: Perform load scenario probability processing on the enterprise load fluctuation pattern data to generate load scenario multi-combination data; calculate the load demand expectation value based on the load scenario multi-combination data to generate multi-scenario load demand forecast data;

[0008] Step S4: perform preliminary energy storage demand estimation based on multi-scenario load demand forecast data to generate preliminary energy storage capacity demand data; perform optimal energy storage scale adaptation interval processing based on the preliminary energy storage capacity demand data to obtain refined energy storage scale interval data;

[0009] Step S5: Perform energy storage unit configuration processing according to the refined energy storage scale interval data to generate energy storage unit configuration plan data; optimize the energy storage unit planning layout of the energy storage unit configuration plan data to obtain optimal energy storage structure layout plan data.

[0010] The present invention can accurately reflect the actual situation of enterprise electricity consumption by acquiring the original historical electricity load data. The drawing of the electricity load curve provides a basis for analyzing the electricity consumption trend and reveals the time distribution of peak and trough electricity consumption. The generated time series electricity load data makes the subsequent analysis more accurate. Through multi-scale load characteristic analysis, the load characteristics of different time scales (such as day, week, month) can be extracted to form load characteristic vector data. By matching the time series electricity load data with the load characteristic vector data, the typical load patterns of different enterprises, that is, the enterprise load prototype pattern data, are identified, which is crucial to understanding the energy consumption laws of different enterprises in the park. The subsequent adaptive hierarchical pattern cutting processing further refines these prototype patterns, and the generated enterprise load fluctuation pattern data can more finely reflect the dynamic characteristics of enterprise load changes. The load scenario probability processing is carried out through the enterprise load fluctuation pattern data, covering a variety of load change scenarios, increasing the comprehensiveness of planning and the ability to cope with uncertainty. The load demand expectation value calculation performed on this basis forms the multi-scenario load demand forecast data, which not only takes into account the average level of load, but also fully estimates the fluctuation range of future load demand. Based on the preliminary energy storage demand estimation based on multi-scenario load demand forecast data, the refined energy storage scale interval data obtained through the optimal energy storage scale adaptation interval processing further considers factors such as economy, technical feasibility and future scalability, ensuring the flexibility and adaptability of the energy storage system design and avoiding waste or shortage of resources. The energy storage unit configuration processing through the refined energy storage scale interval data provides a specific configuration plan for the actual deployment of the energy storage system, considering multiple dimensions such as capacity, type, and layout, ensuring the functionality and economy of the energy storage system. On this basis, the planning and layout optimization of the energy storage unit configuration plan data not only considers factors such as space, cost, and operation and maintenance, but also realizes the optimal layout of the energy storage unit in the industrial park through the generation of the optimal energy storage structure layout plan data, thereby improving the overall efficiency and stability of the entire energy system. Therefore, a planning and layout method for energy storage scale and structure of the present invention performs multi-scale load characteristic analysis through original historical electricity load data, fully considers the dynamic changes of electricity load characteristics, performs load enterprise pattern matching, and performs multi-scenario load demand forecasting according to the matched load pattern. In combination with energy storage targets and different fluctuation levels, the optimal energy storage scale adaptation interval is determined to solve the problem of excessive or insufficient energy storage scale configuration, and realize the refined planning and layout of the energy storage structure of the industrial park.

[0011] Preferably, step S1 comprises the following steps:

[0012] Step S11: obtaining original historical power load data; performing electrical parameter cleaning on the original historical power load data to generate standard historical power load data;

[0013] Step S12: performing time stamp alignment processing on the standard historical power load data, and drawing a power load curve to generate time series power load data;

[0014] Step S13: performing multi-scale load feature decomposition processing according to the time series power load data to generate multi-scale load feature point data;

[0015] Step S14: constructing a characteristic vector based on the multi-scale load characteristic point data to generate load characteristic vector data.

[0016] By acquiring the original historical power load data, the present invention can ensure that the basic data for analysis is derived from the actual power consumption situation. Electrical parameter cleaning can remove noise, outliers and missing values ​​in the data to ensure the accuracy and reliability of the data. Through timestamp alignment processing, the synchronization of power consumption data in different time periods can be ensured, thereby truly reflecting the dynamic process of power consumption changes. Through multi-scale load feature decomposition processing, power consumption characteristics on different time scales can be extracted, revealing the relationship between short-term and long-term power consumption patterns, reflecting the changing laws and periodic characteristics of power load in different time periods. By selecting or constructing appropriate feature indicators (such as maximum load, load change rate, load fluctuation frequency, etc.), the complex and changeable load behavior is converted into a numerical vector form that is easy to process and analyze, which can comprehensively reflect the various characteristics of the power load.

[0017] Preferably, step S13 comprises the following steps:

[0018] Step S131: using a preset time sliding window to divide the time series power load data into time windows to generate power load window data;

[0019] Step S132: Calculate the peak and valley values ​​of the power load window data to generate load peak and valley value data;

[0020] Step S133: classifying the load peak-valley value data into peak-valley feature points by using a preset peak-valley classification threshold to obtain classified peak-valley feature point data, wherein the classified peak-valley feature point data includes peak load feature point data, high-valley load feature point data, and low-valley load feature point data;

[0021] Step S134: Perform moving standard deviation processing on the power load window data to obtain time series average volatility data;

[0022] Step S135: using the time series average fluctuation rate data to analyze the long-term and short-term fluctuation trends of the time series power load data, and identifying fluctuation feature points to generate fluctuation feature point data;

[0023] Step S136: using the fluctuation characteristic point data to perform event load trend analysis on the time series power load data to obtain special event characteristic point data;

[0024] Step S137: Perform Fourier spectrum characteristic analysis on the time series power load data to generate load spectrum characteristic data; perform seasonal cycle identification based on the load spectrum characteristic data, and extract cycle segment feature points to obtain seasonal cycle feature point data;

[0025] Step S138: Perform multi-scale feature point integration on the classified peak and valley feature point data, fluctuation feature point data, seasonal cycle feature point data, and special event feature point data to generate multi-scale load feature point data.

[0026] The present invention processes the continuous time series power load data in segments by dividing the time sliding window, so that the data in each window can be analyzed independently, and this segmentation processing can more clearly capture the changing characteristics of the power load in different time periods. By calculating the peak and valley values ​​of the power load data in each time window, the highest and lowest points of the power load can be identified, which can help identify the specific time of the peak and valley of power consumption and its corresponding power consumption level. By classifying the load peak and valley values, the power load characteristics can be subdivided into different types of characteristic points, such as peaks, high valleys and low valleys. This classification helps to more accurately identify different states of power consumption behavior. Through moving standard deviation processing, the volatility of power load data can be quantified, and the stability and volatility of power load can be better identified. Through long-term and short-term fluctuation trend analysis, the short-term and long-term change trends of power load can be deeply understood, which helps to accurately capture the changing laws of power load on different time scales. Through event load trend analysis, power fluctuation patterns related to specific events (such as equipment failures, production scheduling changes, etc.) can be identified. Fourier spectrum characteristic analysis and seasonal cycle identification can reveal the periodic law of load changes from the frequency domain perspective. The extraction of seasonal cycle feature point data helps to understand the pattern of load changes with seasons. By integrating different feature point data, a comprehensive and detailed description of load characteristics is provided to ensure the comprehensiveness and accuracy of planning.

[0027] Preferably, step S136 includes the following steps:

[0028] Step S1361: Acquire power consumption history event data; perform event type semantic analysis based on the power consumption history event data to generate load event type data;

[0029] Step S1362: extracting the event time sequence of the electricity consumption historical event data to obtain the historical event time axis data;

[0030] Step S1363: Processing the time series power load data according to the historical load fluctuation trajectory through the fluctuation characteristic point data to generate historical load fluctuation trajectory data;

[0031] Step S1364: performing time window positioning processing on the historical load fluctuation trajectory data through the historical event time axis data to generate load event window locking data;

[0032] Step S1365: extracting event load characteristics according to the load event window locking data, and obtaining load characteristic data before the event, load characteristic data during the event, and load characteristic data after the event;

[0033] Step S1366: Perform load baseline statistical analysis on the load characteristic data before the event to generate the load baseline data before the event; perform load fluctuation analysis on the load characteristic data during the event to generate the load fluctuation data during the event; perform load event difference analysis on the load characteristic data after the event through the load baseline data before the event to obtain the event load recovery index;

[0034] Step S1367: performing a weighted calculation of the impact degree of the load fluctuation data in the event and the load recovery index after the event based on the load event type data to generate load event impact degree data;

[0035] Step S1368: Based on the load event impact degree data, the special event feature point division is performed through the load event window locking data to generate special event feature point data.

[0036] The present invention performs semantic analysis of event types through historical power consumption event data to generate load event type data, which is helpful for classifying and identifying different event categories that affect the load (such as holidays, extreme weather, maintenance downtime, etc.). Through event time series extraction, power consumption events can be sorted in chronological order to form historical event timeline data. By processing the historical load fluctuation trajectory of the time series power load data, the fluctuation pattern of the power load in different time periods can be revealed, and a detailed trajectory of the power load change is provided, which is helpful for analyzing the response of the power load before and after a specific event occurs. By positioning the historical load fluctuation trajectory data in a time window, the power fluctuation interval related to a specific event can be accurately locked, which is convenient for evaluating the specific impact of the event on the power load, thereby improving the accuracy and pertinence of the event analysis. By analyzing the load event window locking data, the power consumption characteristic data before and after the event can be extracted respectively, which can reflect the power consumption level under normal circumstances. The load characteristic data in the event reveals the power consumption state when the event occurs, and the load characteristic data after the event helps to analyze the impact of the event on power recovery. By performing statistical analysis and weighted calculation on the characteristic data at different stages, we generated the load baseline before the event, the load fluctuation data during the event, and the load event recovery index and impact data, which not only revealed the direct impact of the event on the load, but also evaluated the efficiency of the system returning to normal. By analyzing the load event impact data, we can identify the characteristic regional points of special events within a specific time window. These characteristic regional point data can indicate the change pattern of the power load during a specific event.

[0037] Preferably, step S2 comprises the following steps:

[0038] Step S21: performing enterprise load aggregation processing on the load characteristic vector data based on the time series power load data to generate enterprise load characteristic vector cluster data;

[0039] Step S22: performing similarity measurement calculation based on the enterprise load characteristic vector cluster data, and performing typical load pattern matching through a preset load pattern library to obtain enterprise load prototype pattern data;

[0040] Step S23: constructing a load fuzzy membership model according to the enterprise load prototype pattern data;

[0041] Step S24: Calculate the uncertainty index of the enterprise load characteristic vector cluster data through the load fuzzy membership model, and perform load pattern soft classification processing to generate soft classification load pattern data;

[0042] Step S25: performing hierarchical clustering optimization on the soft-classified load pattern data to generate load pattern hierarchical data;

[0043] Step S26: Adaptively perform hierarchical pattern cutting processing on the load pattern hierarchical data using preset load clustering evaluation indicators to generate enterprise load fluctuation pattern data.

[0044] The present invention can aggregate the load characteristic vectors of different time periods by aggregating the time series power load data, effectively integrate the power consumption characteristics of the enterprise in a specific time period, and reflect the overall power consumption behavior and pattern of the enterprise. By performing similarity measurement calculation on the enterprise load characteristic vector cluster data, the similarity with the typical load pattern can be identified. This process reflects the characteristics of the enterprise's power consumption behavior by matching with the preset load pattern library. By constructing a load fuzzy membership model, the enterprise load prototype pattern data can be analyzed more flexibly and finely, so that the complexity and diversity of power consumption behavior can be taken into account when classifying load characteristics. By performing uncertainty index calculation on the enterprise load characteristic vector cluster data, the fuzziness and uncertainty in the load characteristics can be quantified, and the classification of load patterns can be further refined. By performing hierarchical clustering optimization on the soft classification load pattern data, different load patterns can be hierarchically classified and integrated. By using the preset load clustering evaluation index to perform adaptive hierarchical pattern cutting, the load pattern hierarchical data can be effectively refined and adjusted, which can reflect the power consumption fluctuation characteristics of the enterprise under different scenarios and help identify the peak and valley periods of power consumption.

[0045] Preferably, step S3 comprises the following steps:

[0046] Step S31: performing empirical mode decomposition on the enterprise load fluctuation mode data and performing independent component analysis to construct load mode feature space data;

[0047] Step S32: performing load scenario probability processing according to the load pattern characteristic space data to generate load scenario multi-combination data;

[0048] Step S33: Acquire enterprise production and operation data, wherein the enterprise production and operation data includes enterprise production equipment data, production energy consumption type data, production process data and enterprise production plan data;

[0049] Step S34: extracting operating load influencing factors according to the enterprise production and operation data to obtain key operating influencing factor data; processing load demand constraint conditions based on the key operating influencing factor data to obtain load demand constraint condition data;

[0050] Step S35: performing load scenario association processing on the load demand constraint condition data and the load scenario multi-combination data, and performing load demand expected value calculation to generate multi-scenario load demand forecast data.

[0051] The present invention can decompose the complex load fluctuation signal into several simple intrinsic mode functions by performing empirical mode decomposition on the enterprise load fluctuation mode data, so that each component can more clearly reflect the fluctuation characteristics on different frequencies and time scales, not only separating the load fluctuation characteristics on different time scales, but also effectively removing noise interference, making the characteristic expression of the load mode purer and clearer. Through load scenario probability processing, the load pattern characteristic space data can be analyzed, and the changes in electricity demand under different circumstances can be simulated and predicted. The key operation influencing factors are extracted and converted into load demand constraint condition data, and the actual operation status of the enterprise is directly closely linked to the load demand forecast, ensuring that the prediction model can accurately reflect the direct impact of production activities on electricity demand, and the load demand constraint condition data is applied to the multi-combination data of the load scenario, and the load scenario association processing is performed to realize the refined constraint of the prediction model. The subsequent load demand expected value calculation not only considers the natural characteristics of load fluctuations, but also fully incorporates the specific impact of the enterprise's production and operation, making the prediction results more accurate and comprehensive, providing a more accurate basis for the planning of energy storage scale and structure, ensuring that the energy storage system can effectively respond to various load demand changes, and improving the stability and economy of the entire park energy system.

[0052] Preferably, step S32 includes the following steps:

[0053] Step S321: performing load scenario encoding processing on the load pattern feature space data through a pre-trained variational autoencoder, and using a generative adversarial network to perform load scenario deduction to generate load scenario sequence sample data;

[0054] Step S322: using a preset recurrent neural network model to perform transfer learning on the load scenario sequence sample data to obtain an initial load scenario evolution model; identifying the scenario transition points according to the initial load scenario evolution model, and performing attention mechanism embedding processing to generate a load scenario evolution model;

[0055] Step S323: extracting dynamic scene evolution rules from load scene sequence sample data using the load scene evolution model to generate dynamic scene evolution rule data;

[0056] Step S324: performing Markov decision state transformation processing according to the dynamic scene evolution rule data to generate dynamic scene transition probability matrix data;

[0057] Step S325: performing Monte Carlo tree search processing on the dynamic scenario transition probability matrix data to obtain a multi-path load scenario tree;

[0058] Step S326: perform Simpson probability integral calculation based on the multi-path load scenario tree, and perform load scenario combination processing to generate load scenario multi-combination data.

[0059] The present invention uses a pre-trained variational autoencoder to encode the load mode feature space data, which can effectively extract the potential characteristics of the load mode. By generating an adversarial network to deduce the load scenario, a variety of load scenario sequence sample data can be generated, providing rich training data. Through the preset recurrent neural network (RNN) model for transfer learning, the initial load scenario evolution model is further refined and optimized, and the model's recognition ability of key scenario transition points is enhanced through the attention mechanism embedding processing. This not only improves the learning efficiency of the model, but also ensures that the model can better focus on the key moments that affect the load evolution, and improves the accuracy and interpretability of the prediction. Through the analysis of the load scenario sequence sample data by the load scenario evolution model, the rules of dynamic scenario evolution can be extracted, and the evolution law of power load in different scenarios can be revealed. By analyzing the dynamic scenario evolution rule data, a state transition model in the Markov decision process can be constructed, and the transition probability between different load scenarios can be quantified. By performing Monte Carlo tree search on the dynamic scenario transition probability matrix, multiple combination paths of load scenarios can be effectively explored, the influence of different decision paths can be simulated, and multiple possible analyses of future power consumption scenarios can be provided. Through the Simpson probability integral calculation, the various scenarios in the multi-path load scenario tree can be effectively weighted and combined, which can comprehensively consider various power consumption scenarios and their probability of occurrence, and reflect the changes in power demand under different scenarios.

[0060] Preferably, step S4 comprises the following steps:

[0061] Step S41: performing multi-scenario fractional calculation based on multi-scenario load demand forecast data to generate load demand fractional data; performing generalized extreme load processing based on multi-scenario load demand forecast data to generate extreme load demand data;

[0062] Step S42: fitting the load demand quantile data with extreme value distribution through the extreme load demand data, and performing multi-level fluctuation interval processing to generate multi-level load demand fluctuation data;

[0063] Step S43: performing preliminary energy storage demand estimation on the multi-level load demand fluctuation data to generate preliminary energy storage capacity demand data;

[0064] Step S44: performing energy storage scale configuration on the preliminary energy storage capacity demand data through a preset energy storage technical parameter library to obtain energy storage scale candidate data;

[0065] Step S45: Performing an energy storage investment economic evaluation on the energy storage scale candidate data to obtain energy storage investment economic scoring data; performing an energy storage technology feasibility evaluation on the energy storage scale candidate data to obtain energy storage scale feasibility scoring data;

[0066] Step S46: constructing a multi-objective energy storage structure optimization model through energy storage scale candidate data based on energy storage scale feasibility scoring data, energy storage investment economic scoring data, and preset energy storage scale constraint rules;

[0067] Step S47: Use a non-dominated sorting genetic algorithm to solve the Pareto energy storage structure of the multi-objective energy storage structure optimization model, and perform optimal energy storage scale adaptation interval processing to obtain refined energy storage scale interval data.

[0068] The present invention generates load demand quantile data through multi-scenario fractional calculation, provides quantitative indicators for understanding the distribution characteristics of load demand, ensures that various extreme situations can be fully considered in the planning process, and improves the risk resistance of the system. Through extreme value distribution fitting and multi-level fluctuation interval processing, it not only reveals the complexity of load demand changes, but also provides a more accurate fluctuation range for subsequent energy storage demand estimation, and improves the adaptability and flexibility of energy storage planning. Through the analysis of multi-level load demand fluctuation data, a preliminary energy storage demand estimation can be carried out to ensure that the electricity demand of the enterprise under different load demand conditions can be met. The energy storage technology parameter library is used to formulate an energy storage scale configuration plan suitable for the actual situation of the enterprise, ensuring the comprehensive consideration of different energy storage technologies and capacity options, making the design of the energy storage system more scientific and reasonable. Economic evaluation and feasibility evaluation provide economic and technical feasibility considerations for decision-making, ensuring that the selection of energy storage scale is both economically reasonable and technically feasible. The constructed multi-objective energy storage structure optimization model realizes the comprehensive optimization of energy storage scale and structure by integrating energy storage scale feasibility scoring data, energy storage investment economic scoring data, and considering actual constraints, reflecting the systematic and comprehensive nature of planning. The multi-objective energy storage structure optimization model is efficiently solved to find the best balance point between multiple objectives, and the optimal energy storage structure selection under different objectives can be identified to ensure the best comprehensive effect between energy storage scale, investment cost and technical feasibility.

[0069] Preferably, step S5 comprises the following steps:

[0070] Step S51: performing energy storage unit configuration processing according to the refined energy storage scale interval data to generate energy storage unit configuration plan data;

[0071] Step S52: acquiring power network structure diagram data; performing energy storage node mapping processing on the power network structure diagram data using the energy storage unit configuration scheme data to obtain power energy storage layout structure data;

[0072] Step S53: extracting the power network topology structure according to the power energy storage layout structure data to obtain the power network topology structure data;

[0073] Step S54: performing a network centrality analysis of energy storage nodes according to the power network topology data, and performing AC power flow calculation of energy storage nodes, to obtain node voltage sensitivity data and node line power flow distribution data respectively;

[0074] Step S55: Based on the preset energy storage unit layout target, the node voltage sensitivity data and the node line power flow distribution data are used to optimize the energy storage unit planning layout, and the data envelopment efficiency is scored to obtain the optimal energy storage structure layout solution data.

[0075] By analyzing the refined energy storage scale interval data, the present invention can formulate a reasonable energy storage unit configuration plan to ensure that the configuration of the energy storage unit matches the enterprise's electricity demand and optimize the utilization efficiency of energy storage resources. By obtaining the power network structure diagram data, the overall architecture of the power system can be fully understood. By using the energy storage unit configuration plan data to perform energy storage node mapping processing, the energy storage unit can be effectively integrated into the power network, ensuring that the energy storage facilities can be reasonably integrated into the power network and optimizing the operating efficiency of the power system. By analyzing the power storage layout structure data, the topological structure of the power network can be extracted, and each node in the power system and its connection relationship can be identified. Through the centrality analysis of the energy storage node network, the importance and influence of each energy storage node in the power network can be evaluated, and the node voltage sensitivity data can reveal the node's responsiveness to voltage changes. The planning and layout optimization based on the energy storage unit layout target not only takes into account the actual operating conditions of the power grid, but also quantifies the efficiency of different layout schemes through data envelopment efficiency scoring. The optimal energy storage structure layout scheme data is obtained to ensure that the energy storage system can be efficiently integrated into the existing power grid and improve the overall performance of the park energy system, including but not limited to improving power supply reliability, optimizing power quality, enhancing system flexibility and promoting the efficient use of renewable energy.

[0076] The present invention also provides a system for planning and layout of energy storage scale and structure, which executes the above-mentioned method for planning and layout of energy storage scale and structure, and the system for planning and layout of energy storage scale and structure comprises:

[0077] The load characteristic analysis module is used to obtain the original historical power load data; draw the power load curve based on the original historical power load data to generate time series power load data; perform multi-scale load characteristic analysis based on the time series power load data to generate load characteristic vector data;

[0078] The load pattern recognition module is used to match the enterprise load pattern with the load characteristic vector data based on the time-series power load data to obtain the enterprise load prototype pattern data; perform adaptive hierarchical pattern cutting processing according to the enterprise load prototype pattern data to generate the enterprise load fluctuation pattern data;

[0079] The multi-scenario load demand forecasting module is used to perform load scenario probability processing on the enterprise load fluctuation pattern data to generate multi-combination data of load scenarios; calculate the expected value of load demand based on the multi-combination data of load scenarios to generate multi-scenario load demand forecasting data;

[0080] The energy storage scale assessment module is used to make a preliminary energy storage demand estimate based on the multi-scenario load demand forecast data and generate preliminary energy storage capacity demand data; based on the preliminary energy storage capacity demand data, the module performs optimal energy storage scale adaptation interval processing to obtain refined energy storage scale interval data;

[0081] The energy storage structure layout optimization module is used to perform energy storage unit configuration processing according to the refined energy storage scale interval data to generate energy storage unit configuration plan data; optimize the energy storage unit planning layout of the energy storage unit configuration plan data to obtain the optimal energy storage structure layout plan data. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 A schematic diagram of the steps of a method for planning and laying out energy storage scale and structure according to the present invention;

[0083] Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart;

[0084] Figure 3 for Figure 1 Detailed implementation steps of step S5;

[0085] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0086] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0087] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different network and / or processor methods and / or microcontroller methods.

[0088] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0089] To achieve this, please refer to Figures 1 to 3 The present invention provides a method for planning and arranging energy storage scale and structure, comprising the following steps:

[0090] Step S1: Obtain original historical power load data; draw a power load curve based on the original historical power load data to generate time series power load data; perform multi-scale load characteristic analysis based on the time series power load data to generate load characteristic vector data;

[0091] Step S2: performing enterprise load pattern matching on load characteristic vector data based on time-series power load data to obtain enterprise load prototype pattern data; performing adaptive hierarchical pattern cutting processing on the enterprise load prototype pattern data to generate enterprise load fluctuation pattern data;

[0092] Step S3: Perform load scenario probability processing on the enterprise load fluctuation pattern data to generate load scenario multi-combination data; calculate the load demand expectation value based on the load scenario multi-combination data to generate multi-scenario load demand forecast data;

[0093] Step S4: perform preliminary energy storage demand estimation based on multi-scenario load demand forecast data to generate preliminary energy storage capacity demand data; perform optimal energy storage scale adaptation interval processing based on the preliminary energy storage capacity demand data to obtain refined energy storage scale interval data;

[0094] Step S5: Perform energy storage unit configuration processing according to the refined energy storage scale interval data to generate energy storage unit configuration plan data; optimize the energy storage unit planning layout of the energy storage unit configuration plan data to obtain optimal energy storage structure layout plan data.

[0095] In the embodiment of the present invention, reference Figure 1 The above is a schematic flow chart of the steps of a method for planning and laying out energy storage scale and structure of the present invention. In this embodiment, the method for planning and laying out energy storage scale and structure includes the following steps:

[0096] Step S1: Obtain original historical power load data; draw a power load curve based on the original historical power load data to generate time series power load data; perform multi-scale load characteristic analysis based on the time series power load data to generate load characteristic vector data;

[0097] In an embodiment of the present invention, for example, the historical power load data of each enterprise in the industrial park is obtained. The data source can be the power consumption data provided by the power company, or the historical power consumption record of the enterprise itself. The time granularity of the data can be 15 minutes, 30 minutes or 1 hour, and the obtained data includes the power load value at each time point. The Pandas library of Python is used to read and process these data, convert them into a time series data format, and use the Matplotlib library to draw a power load curve chart, with time as the horizontal axis and power consumption as the vertical axis. Signal processing methods such as wavelet transform and empirical mode decomposition (EMD) can be used to extract different time scale characteristics of load data, such as daily periodicity, weekly periodicity, monthly periodicity, and some mutation and fluctuation information. Finally, for each load feature point, statistical features such as peak value, valley value, mean value, standard deviation, etc. are extracted, and combined with indicators such as daily peak-to-valley difference, load rate, and fluctuation coefficient, a load characteristic vector is constructed.

[0098] Step S2: performing enterprise load pattern matching on load characteristic vector data based on time-series power load data to obtain enterprise load prototype pattern data; performing adaptive hierarchical pattern cutting processing on the enterprise load prototype pattern data to generate enterprise load fluctuation pattern data;

[0099] In an embodiment of the present invention, several typical enterprise load prototype modes are predefined according to the production characteristics and electricity consumption patterns of different types of enterprises in the industrial park, and the load characteristic vector data of each enterprise are clustered, and enterprises with similar load characteristics are classified into one category to obtain enterprise load prototype mode data. For example, the load characteristic vector data of 10 enterprises can be clustered to finally obtain 3 enterprise load prototype modes, representing continuous type, intermittent type, and peak-valley type respectively. The fuzzy C-means clustering (FCM) algorithm is used to calculate the membership of each enterprise load characteristic vector belonging to different load prototype modes, and soft classification is performed according to the membership value to divide some enterprises into mixed load modes. Hierarchical clustering is performed on the load mode data of all enterprises, and adaptive hierarchical mode cutting is performed according to indicators such as the silhouette coefficient to obtain the load fluctuation mode data of each enterprise, and each enterprise corresponds to multiple load fluctuation modes.

[0100] Step S3: Perform load scenario probability processing on the enterprise load fluctuation pattern data to generate load scenario multi-combination data; calculate the load demand expectation value based on the load scenario multi-combination data to generate multi-scenario load demand forecast data;

[0101] In an embodiment of the present invention, based on the statistical analysis of historical data, the probability of each enterprise load fluctuation mode occurring in different time periods (e.g., different seasons, different months, working days / non-working days, etc.) is calculated to construct a load scenario probability model. Then, a Monte Carlo simulation method is used to randomly generate multiple load scenarios according to the load scenario probability model, each scenario containing a combination of load fluctuation modes of different enterprises and duration, and multiple combination data of load scenarios are generated. Based on the multiple combination data of load scenarios and the power consumption characteristics of each mode, the expected value of load demand under each scenario is calculated. For example, the total power consumption, peak load and other information under each scenario can be calculated. For each load scenario, the load fluctuation modes of each enterprise are superimposed, and the total load demand curve of the industrial park under the scenario is calculated, thereby generating multi-scenario load demand forecast data.

[0102] Step S4: perform preliminary energy storage demand estimation based on multi-scenario load demand forecast data to generate preliminary energy storage capacity demand data; perform optimal energy storage scale adaptation interval processing based on the preliminary energy storage capacity demand data to obtain refined energy storage scale interval data;

[0103] In an embodiment of the present invention, quantile calculation is performed on multi-scenario load demand forecast data to obtain load demand conditions under different probability levels. At the same time, the extreme value distribution of load demand is fitted using a generalized extreme value analysis method to infer extreme load demand values ​​that will occur in the future. Secondly, based on the load demand quantile data and extreme load demand data, the load demand fluctuation range is divided into multiple levels, such as extreme fluctuation, high fluctuation, medium fluctuation, low fluctuation, etc. For each fluctuation level, the required energy storage capacity is calculated according to the set energy storage target (such as the peak shaving and valley filling ratio) to generate preliminary energy storage capacity demand data. Then, based on the preliminary energy storage capacity demand data, combined with the parameters and cost information of energy storage technologies such as lithium-ion batteries and flow batteries, multiple groups of different energy storage scale configuration schemes are formulated. The economic and feasibility evaluation of each group of energy storage scale configuration schemes is performed, and the multi-objective optimization algorithm is used to solve the optimal energy storage scale adaptation interval, such as 8-12MWh, as the refined energy storage scale interval data.

[0104] Step S5: Perform energy storage unit configuration processing according to the refined energy storage scale interval data to generate energy storage unit configuration plan data; optimize the energy storage unit planning layout of the energy storage unit configuration plan data to obtain optimal energy storage structure layout plan data.

[0105] In an embodiment of the present invention, for example, according to the refined energy storage scale interval data (8-12MWh) and the project budget, it is determined to use 80 150kWh lithium-ion battery units for configuration, with a total capacity of 12MWh. The power network structure diagram data of the industrial park is obtained, and the 80 energy storage unit nodes are added to the power network structure diagram, and according to the main functions of the energy storage system, they are preliminarily selected to be connected to each load node of the 10kV distribution network. The power network topology data containing the energy storage unit nodes is extracted, and the network centrality analysis and flow calculation are performed on each energy storage node to obtain the node voltage sensitivity data and the node line power flow distribution data. Combined with the goal of improving power supply reliability, the access location and capacity of the energy storage unit are optimized using a multi-objective optimization algorithm, and it is finally determined that 48 energy storage units are connected to nodes with higher voltage sensitivity, and 32 energy storage units are connected to areas with heavier loads, and the DEA score of the scheme is calculated to be 0.85, indicating that the scheme is reasonable and effective.

[0106] Preferably, step S1 comprises the following steps:

[0107] Step S11: obtaining original historical power load data; performing electrical parameter cleaning on the original historical power load data to generate standard historical power load data;

[0108] Step S12: performing time stamp alignment processing on the standard historical power load data, and drawing a power load curve to generate time series power load data;

[0109] Step S13: performing multi-scale load feature decomposition processing according to the time series power load data to generate multi-scale load feature point data;

[0110] Step S14: constructing a characteristic vector based on the multi-scale load characteristic point data to generate load characteristic vector data.

[0111] In an embodiment of the present invention, original historical power load data is collected from an industrial park power management system, a smart meter or a SCADA system. The collected data contains various time granularities, such as 1 minute, 5 minutes, 15 minutes, 30 minutes or 1 hour. Missing values ​​are detected for the original data, and the missing data can be filled using linear interpolation, spline interpolation and other methods. Outliers are identified, for example, using the 3σ principle, data beyond three standard deviations of the mean value is regarded as outliers, and the outliers are replaced or removed using adjacent values. The target time granularity is determined, for example, 15 minutes. If the time granularity of the original data is inconsistent with the target time granularity, data resampling is required. For example, the resample function in the Pandas library can be used to convert the original data into time series data of the target time granularity. If the data comes from multiple data sources, timestamp alignment is required to ensure that the data from different data sources are synchronized in time. Use Python's Matplotlib library to draw a power load curve. The horizontal axis is time and the vertical axis is power load. Load curves of different time scales can be drawn. Select a suitable signal decomposition method to perform multi-scale decomposition on the time series power load data, and extract features from the decomposed multi-scale sub-signals as multi-scale load feature point data. According to the characteristics of the power load in the industrial park and the analysis objectives, select suitable load characteristic indicators for the load characteristic points. For example, you can select indicators such as daily peak-to-valley difference, load rate, fluctuation coefficient, peak occurrence time, valley occurrence time, etc., calculate the specific values ​​of the selected characteristic indicators, and combine the calculated multiple characteristic indicators into a vector, which is the load characteristic vector data.

[0112] Preferably, step S13 comprises the following steps:

[0113] Step S131: using a preset time sliding window to divide the time series power load data into time windows to generate power load window data;

[0114] Step S132: Calculate the peak and valley values ​​of the power load window data to generate load peak and valley value data;

[0115] Step S133: classifying the load peak-valley value data into peak-valley feature points by using a preset peak-valley classification threshold to obtain classified peak-valley feature point data, wherein the classified peak-valley feature point data includes peak load feature point data, high-valley load feature point data, and low-valley load feature point data;

[0116] Step S134: Perform moving standard deviation processing on the power load window data to obtain time series average volatility data;

[0117] Step S135: using the time series average fluctuation rate data to analyze the long-term and short-term fluctuation trends of the time series power load data, and identifying fluctuation feature points to generate fluctuation feature point data;

[0118] Step S136: using the fluctuation characteristic point data to perform event load trend analysis on the time series power load data to obtain special event characteristic point data;

[0119] Step S137: Perform Fourier spectrum characteristic analysis on the time series power load data to generate load spectrum characteristic data; perform seasonal cycle identification based on the load spectrum characteristic data, and extract cycle segment feature points to obtain seasonal cycle feature point data;

[0120] Step S138: Perform multi-scale feature point integration on the classified peak and valley feature point data, fluctuation feature point data, seasonal cycle feature point data, and special event feature point data to generate multi-scale load feature point data.

[0121] In an embodiment of the present invention, the size of the time sliding window is preset according to the target of data analysis and the change law of the load curve. For example, if the daily load characteristics are to be analyzed, the time sliding window can be set to 24 hours and the step size can be 1 hour. Secondly, the sliding window function in Python (such as the rolling function in the Pandas library) is used to divide the time series power load data into windows. The sliding window moves along the time axis, and the load data in each window is extracted as a fragment to form the power load window data. For each power load window data, the maximum and minimum values ​​in the window are calculated to form the load peak and valley value data. According to historical load data analysis or expert experience, the peak and valley classification thresholds are preset. For example, a value higher than a certain proportion of the average load can be set as the peak load threshold, and a value lower than a certain proportion of the average load can be set as the valley load threshold, and the value between the two is the high valley load threshold. The load peak and valley value data is compared with the preset threshold, and the data points are classified into peak load feature point data, high valley load feature point data, and low valley load feature point data according to the size of the peak and valley values. The sliding window function and the standard deviation function in Python (such as the std function in the Numpy library) are used to calculate the moving standard deviation of the time series power load data. The size of the sliding window can be the same as the window size set in step S131, or it can be adjusted according to the actual situation. For example, the standard deviation of each 24-hour power load window data in step S131 can be calculated to obtain 49 standard deviation values. These standard deviation values ​​are arranged in chronological order to obtain the time series average volatility data, and the time series average volatility data is analyzed, for example, using methods such as wavelet transform or moving average to identify long-term and short-term fluctuation trends. For example, the volatility data can be decomposed into fluctuation components of different time scales, representing short-term fluctuations, medium-term fluctuations, and long-term fluctuations, respectively. According to the results of the long-term and short-term fluctuation trend analysis, the fluctuation feature points are identified, for example, the points where the volatility obviously rises or falls can be marked as fluctuation feature points. According to the results of the long-term and short-term fluctuation trend analysis, the fluctuation feature points are identified, for example, the points where the volatility obviously rises or falls can be marked as fluctuation feature points. According to the results of the event load trend analysis, the special event feature points are identified. The fast Fourier transform (FFT) is used to perform spectrum analysis on the time-series power load data to obtain the amplitude of the load signal at different frequencies, that is, the load spectrum characteristic data. Then, the load spectrum characteristic data is analyzed to identify the main periodic components. For the power load of industrial parks, seasonal cycles such as annual cycles and weekly cycles are usually concerned. According to the identified cycle length, the time-series power load data is divided into multiple cycle segments, and the starting point of each cycle segment is extracted as the seasonal cycle feature point data. All the feature point data obtained are integrated in chronological order to form a data set containing multiple feature point types, that is, multi-scale load feature point data.

[0122] Preferably, step S136 includes the following steps:

[0123] Step S1361: Acquire power consumption history event data; perform event type semantic analysis based on the power consumption history event data to generate load event type data;

[0124] Step S1362: extracting the event time sequence of the electricity consumption historical event data to obtain the historical event time axis data;

[0125] Step S1363: Processing the time series power load data according to the historical load fluctuation trajectory through the fluctuation characteristic point data to generate historical load fluctuation trajectory data;

[0126] Step S1364: performing time window positioning processing on the historical load fluctuation trajectory data through the historical event time axis data to generate load event window locking data;

[0127] Step S1365: extracting event load characteristics according to the load event window locking data, and obtaining load characteristic data before the event, load characteristic data during the event, and load characteristic data after the event;

[0128] Step S1366: Perform load baseline statistical analysis on the load characteristic data before the event to generate the load baseline data before the event; perform load fluctuation analysis on the load characteristic data during the event to generate the load fluctuation data during the event; perform load event difference analysis on the load characteristic data after the event through the load baseline data before the event to obtain the event load recovery index;

[0129] Step S1367: performing a weighted calculation of the impact degree of the load fluctuation data in the event and the load recovery index after the event based on the load event type data to generate load event impact degree data;

[0130] Step S1368: Based on the load event impact degree data, the special event feature point division is performed through the load event window locking data to generate special event feature point data.

[0131] In an embodiment of the present invention, historical event data related to electricity consumption are collected from data sources such as industrial park management systems, power operation logs, and meteorological records. These data include equipment maintenance plans, holiday arrangements, large-scale event notifications, and temperature sudden change records. Natural language processing (NLP) technology is used to perform semantic analysis on the historical electricity consumption event data to extract event type information. The event occurrence time information is extracted from the historical electricity consumption event data, and arranged in chronological order to form historical event timeline data. The time series power load data is divided into multiple time segments according to the fluctuation feature point data. Each segment represents a relatively stable load fluctuation cycle. For each time segment, the load value, duration, average load, fluctuation amplitude and other characteristic indicators of its starting point and end point are extracted to form historical load fluctuation trajectory data. Each historical event time period obtained in step S1362 is matched with the historical load fluctuation trajectory data in step S1363, and the load fluctuation segments that coincide with or are adjacent to the event time period are found, and these load fluctuation segments are marked as "event windows". For each load event window, the load feature data before, during, and after the event are extracted. For example, the average value and standard deviation of electricity consumption in a period before the event can be calculated to generate the load baseline data before the event. The change amplitude and fluctuation frequency of electricity consumption during the event can be calculated to generate the load fluctuation data during the event. The time and degree of recovery of electricity consumption after the event can be calculated to obtain the load recovery index of the event. Different types of events are given different impact weights according to the load event type data and expert experience. For example, the weight of planned events can be lower than that of sudden events, and the weight of short-term events can be lower than that of long-term events. The load fluctuation data during the event and the load recovery index after the event are multiplied by the weight of the corresponding event type to obtain the weighted load fluctuation data and load recovery index. The weighted data are integrated to generate the load event impact data, and the threshold for dividing the characteristic points of special events is set according to the load event impact data. For example, events with load fluctuations exceeding a certain amplitude or load recovery time exceeding a certain length of time can be marked as special events. According to the set threshold, the events that meet the conditions in the load event window lock data are screened out, and the time of event occurrence, load value, event type, impact degree and other information are extracted to generate special event feature point data.

[0132] Preferably, step S2 comprises the following steps:

[0133] Step S21: performing enterprise load aggregation processing on the load characteristic vector data based on the time series power load data to generate enterprise load characteristic vector cluster data;

[0134] Step S22: performing similarity measurement calculation based on the enterprise load characteristic vector cluster data, and performing typical load pattern matching through a preset load pattern library to obtain enterprise load prototype pattern data;

[0135] Step S23: constructing a load fuzzy membership model according to the enterprise load prototype pattern data;

[0136] Step S24: Calculate the uncertainty index of the enterprise load characteristic vector cluster data through the load fuzzy membership model, and perform load pattern soft classification processing to generate soft classification load pattern data;

[0137] Step S25: performing hierarchical clustering optimization on the soft-classified load pattern data to generate load pattern hierarchical data;

[0138] Step S26: Adaptively perform hierarchical pattern cutting processing on the load pattern hierarchical data using preset load clustering evaluation indicators to generate enterprise load fluctuation pattern data.

[0139] As an example of the present invention, refer to Figure 2 As shown, Figure 1 Detailed implementation steps of step S2 in the embodiment are shown in the flowchart. In this embodiment, step S2 includes:

[0140] Step S21: performing enterprise load aggregation processing on the load characteristic vector data based on the time series power load data to generate enterprise load characteristic vector cluster data;

[0141] In an embodiment of the present invention, the load characteristic vector data needs to be grouped according to the enterprise to which they belong. For example, assuming that the load characteristic vector data contains the power load characteristics of three enterprises A, B, and C in the industrial park, the load characteristic vector data belonging to enterprise A needs to be grouped into one group, the load characteristic vector data belonging to enterprise B into one group, and the load characteristic vector data belonging to enterprise C into one group. Cluster analysis is performed on the load characteristic vector data within each enterprise. Machine learning algorithms such as K-means clustering algorithm and density clustering algorithm can be used to divide load characteristic vectors with similar characteristics into the same cluster. Finally, the clustering results of the load characteristic vector data of each enterprise are integrated to form enterprise load characteristic vector cluster data.

[0142] Step S22: performing similarity measurement calculation based on the enterprise load characteristic vector cluster data, and performing typical load pattern matching through a preset load pattern library to obtain enterprise load prototype pattern data;

[0143] In an embodiment of the present invention, a load pattern library including typical load patterns is preset. For example, typical load patterns including continuous type, intermittent type, peak-valley type, etc. can be preset. Typical load patterns are matched through the preset load pattern library to obtain enterprise load prototype pattern data. For example, a nearest neighbor algorithm can be used to match each enterprise load characteristic vector cluster with a typical load pattern in the load pattern library, and similarity metrics such as Euclidean distance, Manhattan distance, and cosine similarity can be used to calculate the similarity between each enterprise load characteristic vector cluster and each typical load pattern in the load pattern library, and the most similar typical load pattern is found as the load prototype pattern of the enterprise.

[0144] Step S23: constructing a load fuzzy membership model according to the enterprise load prototype pattern data;

[0145] In an embodiment of the present invention, a load fuzzy membership model is used to describe the degree to which each enterprise's load characteristic vector cluster belongs to different load prototype modes. The load fuzzy membership model can be constructed using methods such as the fuzzy C-means clustering algorithm (FCM). For example, assume that a load characteristic vector cluster of enterprise A is matched to a "peak-valley type" load mode according to step S22, but some load characteristic vectors in the cluster are also similar to a "continuous type" load mode. At this time, the FCM algorithm can be used to calculate the membership of the cluster to the two load modes of "peak-valley type" and "continuous type", for example, the membership to the "peak-valley type" is 0.7, and the membership to the "continuous type" is 0.3.

[0146] Step S24: Calculate the uncertainty index of the enterprise load characteristic vector cluster data through the load fuzzy membership model, and perform load pattern soft classification processing to generate soft classification load pattern data;

[0147] In an embodiment of the present invention, the uncertainty index of each enterprise load characteristic vector cluster is calculated according to the load fuzzy membership model. Common uncertainty indicators include entropy value and fuzzy membership mean. For example, assuming that the membership of a certain enterprise load characteristic vector cluster to the "peak and valley type" load mode is 0.6, and the membership to the "continuous type" load mode is 0.4, then the entropy value of the cluster can be calculated as -0.6log2(0.6)-0.4log2(0.4)=0.97. Secondly, an uncertainty threshold is preset, for example 0.5. The enterprise load characteristic vector cluster with uncertainty index higher than the threshold is marked as a "mixed type" load mode, and the enterprise load characteristic vector cluster with uncertainty index lower than the threshold is marked as a "single type" load mode.

[0148] Step S25: performing hierarchical clustering optimization on the soft-classified load pattern data to generate load pattern hierarchical data;

[0149] In an embodiment of the present invention, a suitable hierarchical clustering algorithm is selected, such as Ward's method or average connection method, to perform hierarchical clustering on the soft classified load pattern data. The clustering distance can be calculated based on the similarity between the enterprise load characteristic vectors or the similarity between the load patterns. A tree structure, i.e., load pattern hierarchical data, is constructed based on the clustering results. The root node of the tree structure represents the set of all enterprise load patterns, each leaf node represents a cluster of enterprise load characteristic vectors, and the intermediate nodes represent the load pattern clustering results at different levels.

[0150] Step S26: Adaptively perform hierarchical pattern cutting processing on the load pattern hierarchical data using preset load clustering evaluation indicators to generate enterprise load fluctuation pattern data.

[0151] In an embodiment of the present invention, a preset load clustering evaluation index, such as a silhouette coefficient (Silhouette Coefficient) or a Calinski-Harabasz index, is used to evaluate the pros and cons of clustering results at different levels. Starting from the root node of the load pattern hierarchy data, traverse downward layer by layer to calculate the load clustering evaluation index of each node. Then, find the node with the highest evaluation index, and cut off the subtree of the node to form an independent load pattern cluster. Repeat the above steps until all nodes are divided into different load pattern clusters. The load data of all enterprise load characteristic vector clusters contained in each load pattern cluster are integrated to form enterprise load fluctuation pattern data. Each enterprise load fluctuation pattern data contains a set of enterprise load curves with similar fluctuation characteristics, as well as corresponding load pattern type, prototype pattern, membership degree and other information.

[0152] Preferably, step S3 comprises the following steps:

[0153] Step S31: performing empirical mode decomposition on the enterprise load fluctuation mode data and performing independent component analysis to construct load mode feature space data;

[0154] Step S32: performing load scenario probability processing according to the load pattern characteristic space data to generate load scenario multi-combination data;

[0155] Step S33: Acquire enterprise production and operation data, wherein the enterprise production and operation data includes enterprise production equipment data, production energy consumption type data, production process data and enterprise production plan data;

[0156] Step S34: extracting operating load influencing factors according to the enterprise production and operation data to obtain key operating influencing factor data; processing load demand constraint conditions based on the key operating influencing factor data to obtain load demand constraint condition data;

[0157] Step S35: performing load scenario association processing on the load demand constraint condition data and the load scenario multi-combination data, and performing load demand expected value calculation to generate multi-scenario load demand forecast data.

[0158] In an embodiment of the present invention, the empirical mode decomposition (EMD) method is used to decompose the load fluctuation pattern data of each enterprise obtained in step S26 into several intrinsic mode functions (IMFs) and a residual term. Each IMF represents the fluctuation components of different time scales in the enterprise load fluctuation pattern data, and the residual term represents the long-term trend of the load fluctuation pattern data. All IMFs are subjected to independent component analysis (ICA) processing, and each independent component represents a potential driving factor in the enterprise load fluctuation pattern data. All independent components are used as features to construct load pattern feature space data. Each enterprise load fluctuation pattern data corresponds to a point in the feature space, and its coordinates are composed of the values ​​of each independent component. The load pattern feature space is divided into multiple regions, each of which represents a typical load scenario. For example, the load pattern feature space can be divided into high load, low load, stable load, fluctuating load and other scenarios according to the characteristics of load level, fluctuation amplitude, periodicity, etc. Using the Monte Carlo simulation method, multiple load scenario combinations are randomly generated according to the load scenario probability. Each load scenario combination contains multiple enterprise load fluctuation pattern data, representing a future load demand scenario. Obtain enterprise production equipment data, production energy consumption type data, production process data, and enterprise production plan data from data sources such as the enterprise's production management system (MES) and enterprise resource planning system (ERP). Analyze the enterprise's production and operation data to identify key operating factors that have a significant impact on load demand. For example, correlation analysis, principal component analysis, and other methods can be used to identify factors such as production equipment, production processes, and production plans that are highly correlated with load demand. Secondly, a load demand constraint model is established based on the historical data and future plans of key operating factors. For example, the maximum load demand of the equipment can be calculated based on the rated power and planned operating time of the equipment, and the total energy consumption demand of the product can be calculated based on the unit energy consumption and planned output of the product. According to the enterprise load fluctuation pattern data and load demand constraints contained in the load scenario combination, the expected value of load demand for each time period is calculated. For example, the load demand of each enterprise's load fluctuation pattern data in this scenario can be superimposed, and modified according to the load demand constraints to obtain the final expected value of load demand. The expected values ​​of load demand for all load scenario combinations are integrated to generate multi-scenario load demand forecast data. The data contains multiple future load demand scenarios and the expected values ​​of load demand for each time period under each scenario.

[0159] Preferably, step S32 includes the following steps:

[0160] Step S321: performing load scenario encoding processing on the load pattern feature space data through a pre-trained variational autoencoder, and using a generative adversarial network to perform load scenario deduction to generate load scenario sequence sample data;

[0161] Step S322: using a preset recurrent neural network model to perform transfer learning on the load scenario sequence sample data to obtain an initial load scenario evolution model; identifying the scenario transition points according to the initial load scenario evolution model, and performing attention mechanism embedding processing to generate a load scenario evolution model;

[0162] Step S323: extracting dynamic scene evolution rules from load scene sequence sample data using the load scene evolution model to generate dynamic scene evolution rule data;

[0163] Step S324: performing Markov decision state transformation processing according to the dynamic scene evolution rule data to generate dynamic scene transition probability matrix data;

[0164] Step S325: performing Monte Carlo tree search processing on the dynamic scenario transition probability matrix data to obtain a multi-path load scenario tree;

[0165] Step S326: perform Simpson probability integral calculation based on the multi-path load scenario tree, and perform load scenario combination processing to generate load scenario multi-combination data.

[0166] In an embodiment of the present invention, a pre-trained variational autoencoder (VAE) is used to encode the load pattern feature space data obtained in the step, and the high-dimensional load pattern features are compressed into a low-dimensional latent space representation. The latent space representation after VAE encoding is deduced using a generative adversarial network (GAN). GAN includes a generator and a discriminator, the generator attempts to generate realistic load scenario samples, and the discriminator attempts to distinguish between real load scenario samples and samples generated by the generator. Through adversarial training, the generator can learn the evolution law of the load scenario and generate more reasonable load scenario sequence sample data. Select a suitable recurrent neural network (RNN) model, such as a long short-term memory network (LSTM) or a gated recurrent unit (GRU), as the initial load scenario evolution model. The RNN model is transferred using the load scenario sequence sample data, and the training model learns the evolution law of the load scenario. The trained RNN model is analyzed to identify the critical point where the load scenario changes, such as the time point when the load level changes suddenly or the fluctuation increases. The attention mechanism is introduced into the RNN model, so that the model can pay more attention to the key information near the load scenario transition point and improve the accuracy of scenario evolution prediction. The trained load scenario evolution model is used to predict the load scenario sequence sample data to obtain the evolution trend of each scenario in the future. The transfer relationship between different load scenarios is analyzed and the corresponding evolution rules are extracted. For example, the load scenarios can be divided into different types according to the characteristics of load level, fluctuation amplitude, etc., and the frequency of mutual transfer between different types of scenarios is counted to extract evolution rules such as "high load scenarios have a certain probability of transferring to low load scenarios after a period of time". Each load scenario is regarded as a state in the MDP, and the dynamic scenario evolution rules are converted into the transition probability between states. For example, assuming that "high load scenarios have a 20% probability of transferring to low load scenarios after 1 hour", "high load scenarios" and "low load scenarios" can be set as two states in the MDP respectively, and the transition probability between the two states is set to 0.2. The dynamic scenario transition probability matrix is ​​constructed according to all dynamic scenario evolution rules, and each element in the matrix represents the transition probability between corresponding states. Based on the dynamic scenario transition probability matrix, the MCTS method is used to start from the initial load scenario, randomly select subsequent scenarios according to the state transition probability, and simulate the evolution process of the load scenario. Repeat MCTS sampling multiple times to generate multiple paths containing different load scenario sequences to form a multi-path load scenario tree. Each path represents a load scenario evolution process. The Simpson probability integral method is used to calculate the probability of each load scenario path. The Simpson probability integral method can calculate the probability of the entire path based on the transition probability of each state on the path. Secondly, all paths are grouped according to the load scenario combinations they contain, and the total probability of each load scenario combination is calculated.For example, the probabilities of all paths containing the "high load-low load-stable load" scenario sequence are accumulated to obtain the total probability of the scenario combination. All load scenario combinations and their corresponding probabilities form load scenario multi-combination data.

[0167] Preferably, step S4 comprises the following steps:

[0168] Step S41: performing multi-scenario fractional calculation based on multi-scenario load demand forecast data to generate load demand fractional data; performing generalized extreme load processing based on multi-scenario load demand forecast data to generate extreme load demand data;

[0169] Step S42: fitting the load demand quantile data with extreme value distribution through the extreme load demand data, and performing multi-level fluctuation interval processing to generate multi-level load demand fluctuation data;

[0170] Step S43: performing preliminary energy storage demand estimation on the multi-level load demand fluctuation data to generate preliminary energy storage capacity demand data;

[0171] Step S44: performing energy storage scale configuration on the preliminary energy storage capacity demand data through a preset energy storage technical parameter library to obtain energy storage scale candidate data;

[0172] Step S45: Performing an energy storage investment economic evaluation on the energy storage scale candidate data to obtain energy storage investment economic scoring data; performing an energy storage technology feasibility evaluation on the energy storage scale candidate data to obtain energy storage scale feasibility scoring data;

[0173] Step S46: constructing a multi-objective energy storage structure optimization model through energy storage scale candidate data based on energy storage scale feasibility scoring data, energy storage investment economic scoring data, and preset energy storage scale constraint rules;

[0174] Step S47: Use a non-dominated sorting genetic algorithm to solve the Pareto energy storage structure of the multi-objective energy storage structure optimization model, and perform optimal energy storage scale adaptation interval processing to obtain refined energy storage scale interval data.

[0175] In an embodiment of the present invention, quantile calculation is performed on multi-scenario load demand forecast data. For example, the 5%, 25%, 50%, 75% and 95% quantiles of the load demand in each time period can be calculated to obtain load demand quantile data. Generalized extreme value analysis (GEV) or other extreme value analysis methods are performed on the multi-scenario load demand forecast data to fit the extreme value distribution of the load demand, and infer the extreme load demand value that will appear in the future to generate extreme load demand data. According to the load demand quantile data and the extreme load demand data, the load demand fluctuation range is divided into multiple levels, such as extreme fluctuation, high fluctuation, medium fluctuation, low fluctuation, etc. Each level corresponds to a load demand interval. Finally, the load demand data in each time period is compared with each fluctuation interval to determine the fluctuation level to which it belongs, and generate multi-level load demand fluctuation data. For each fluctuation level and energy storage target, the required energy storage capacity is calculated. For example, the energy storage capacity required to meet different energy storage targets can be calculated based on indicators such as peak-to-valley load difference, fluctuation amplitude, and allowable fluctuation range, combined with parameters such as the charging and discharging efficiency and response time of the energy storage system. According to the preliminary energy storage capacity demand data, combined with the energy storage technology parameter library, select the appropriate energy storage technology combination, and determine the installation scale of each technology, such as 10MWh for lithium-ion batteries and 5MWh for flow batteries, to generate multiple sets of different energy storage scale configuration schemes. For each set of energy storage scale candidate data, perform energy storage investment economic evaluation, such as calculating indicators such as payback period, net present value, internal rate of return, and obtain energy storage investment economic score data. At the same time, perform energy storage technology feasibility evaluation on each set of energy storage scale candidate data, such as evaluating whether the energy storage system meets the grid access requirements and safe operation requirements, and obtain energy storage scale feasibility score data. Add energy storage scale constraint rules to the model, such as the maximum allowable installation capacity and the maximum allowable investment cost, to ensure the actual feasibility of the optimization results. Use multi-objective optimization algorithms such as non-dominated sorting genetic algorithm (NSGA) to solve the constructed model and obtain a set of Pareto optimal solution sets. Each solution in the Pareto optimal solution set represents an energy storage scale configuration scheme, which cannot be further optimized in terms of economy, feasibility, etc., and needs to be weighed according to actual conditions. According to the Pareto optimal solution set, the optimal energy storage scale adaptation range is determined, for example, the minimum energy storage scale is 5MWh and the maximum energy storage scale is 15MWh. Different adaptation range processing methods can be set according to actual conditions, for example, weighted average can be performed based on the weights of economic benefits and technical feasibility, and then the energy storage scale with the largest weight can be selected as the optimal energy storage scale.

[0176] Preferably, step S5 comprises the following steps:

[0177] Step S51: performing energy storage unit configuration processing according to the refined energy storage scale interval data to generate energy storage unit configuration plan data;

[0178] Step S52: acquiring power network structure diagram data; performing energy storage node mapping processing on the power network structure diagram data using the energy storage unit configuration scheme data to obtain power energy storage layout structure data;

[0179] Step S53: extracting the power network topology structure according to the power energy storage layout structure data to obtain the power network topology structure data;

[0180] Step S54: performing a network centrality analysis of energy storage nodes according to the power network topology data, and performing AC power flow calculation of energy storage nodes, to obtain node voltage sensitivity data and node line power flow distribution data respectively;

[0181] Step S55: Based on the preset energy storage unit layout target, the node voltage sensitivity data and the node line power flow distribution data are used to optimize the energy storage unit planning layout, and the data envelopment efficiency is scored to obtain the optimal energy storage structure layout solution data.

[0182] As an example of the present invention, refer to Figure 3 As shown, Figure 1 Detailed implementation steps of step S5 in the flowchart, in this example, step S5 includes:

[0183] Step S51: performing energy storage unit configuration processing according to the refined energy storage scale interval data to generate energy storage unit configuration plan data;

[0184] In an embodiment of the present invention, selectable energy storage technology types are determined based on the refined energy storage scale interval data and the project budget, such as lithium-ion batteries, flow batteries, supercapacitors, etc. The optimal configuration ratio of each energy storage technology is determined based on the performance parameters of each energy storage technology, such as single cell capacity, charge and discharge power, response time, etc., and the specific functional requirements of the project for the energy storage system, such as peak shaving and valley filling, smoothing output fluctuations, and improving power supply reliability. Based on the refined energy storage scale interval data and the configuration ratio, the number of units and total capacity required for each energy storage technology are calculated to generate energy storage unit configuration plan data.

[0185] Step S52: acquiring power network structure diagram data; performing energy storage node mapping processing on the power network structure diagram data using the energy storage unit configuration scheme data to obtain power energy storage layout structure data;

[0186] In an embodiment of the present invention, the power network structure diagram data of the industrial park is obtained, including information of nodes such as substations, lines, loads, and the connection relationship between them. According to the energy storage unit configuration plan data, the virtual energy storage unit node is added to the power network structure diagram. Finally, according to the main functions and application scenarios of the energy storage system, the access location of each energy storage unit node is determined. For example, if the energy storage system is mainly used to participate in the peak and frequency regulation of the power grid, the energy storage unit node can be connected to the substation or near the important load; if the energy storage system is mainly used to smooth the output fluctuations of distributed new energy, the energy storage unit node can be connected to the vicinity of the distributed new energy grid connection point.

[0187] Step S53: extracting the power network topology structure according to the power energy storage layout structure data to obtain the power network topology structure data;

[0188] In the embodiment of the present invention, the graph theory method is used to analyze the power storage layout structure data to extract the topological structure data of the power network. The data can be represented in the form of an adjacency matrix, an association matrix, etc., which describes the connection relationship between each node in the network.

[0189] Step S54: performing a network centrality analysis of energy storage nodes according to the power network topology data, and performing AC power flow calculation of energy storage nodes, to obtain node voltage sensitivity data and node line power flow distribution data respectively;

[0190] In an embodiment of the present invention, the network centrality index of each energy storage node, such as degree centrality, betweenness centrality, closeness centrality, etc., is calculated based on the power network topology data. The network centrality index can reflect the importance and influence range of the node in the network. The AC power flow simulation calculation is performed on the power network containing the energy storage unit to obtain information such as the voltage amplitude and phase angle of each node and the power flow of the line. For example, the AC power flow calculation can be performed using power system simulation software (such as PSCAD) to obtain the line power flow distribution data of different energy storage nodes. Finally, the impact of different energy storage unit access positions on the node voltage is analyzed, and the node voltage sensitivity data, such as voltage deviation, voltage fluctuation and other indicators, is calculated. At the same time, the impact of the charging and discharging behavior of the energy storage unit on the line power flow is analyzed to obtain the node line power flow distribution data.

[0191] Step S55: Based on the preset energy storage unit layout target, the node voltage sensitivity data and the node line power flow distribution data are used to optimize the energy storage unit planning layout, and the data envelopment efficiency is scored to obtain the optimal energy storage structure layout solution data.

[0192] In an embodiment of the present invention, the energy storage unit planning layout is optimized based on the preset energy storage unit layout objectives (such as maximizing energy storage efficiency, minimizing network losses, improving voltage stability, etc.) using node voltage sensitivity data and node line power flow distribution data. For example, optimization algorithms such as genetic algorithms and particle swarm optimization algorithms can be used to find the optimal energy storage unit layout scheme based on the objective function and constraints. Different constraints can be set according to actual conditions, such as energy storage capacity restrictions, line capacity restrictions, voltage stability restrictions, etc. The optimized energy storage structure layout scheme is evaluated, such as calculating the data envelopment efficiency (DEA) score, to evaluate the rationality and effectiveness of the scheme.

[0193] The present invention also provides a system for planning and layout of energy storage scale and structure, which executes the above-mentioned method for planning and layout of energy storage scale and structure, and the system for planning and layout of energy storage scale and structure comprises:

[0194] The load characteristic analysis module is used to obtain the original historical power load data; draw the power load curve based on the original historical power load data to generate time series power load data; perform multi-scale load characteristic analysis based on the time series power load data to generate load characteristic vector data;

[0195] The load pattern recognition module is used to match the enterprise load pattern with the load characteristic vector data based on the time-series power load data to obtain the enterprise load prototype pattern data; perform adaptive hierarchical pattern cutting processing according to the enterprise load prototype pattern data to generate the enterprise load fluctuation pattern data;

[0196] The multi-scenario load demand forecasting module is used to perform load scenario probability processing on the enterprise load fluctuation pattern data to generate multi-combination data of load scenarios; calculate the expected value of load demand based on the multi-combination data of load scenarios to generate multi-scenario load demand forecasting data;

[0197] The energy storage scale assessment module is used to make a preliminary energy storage demand estimate based on the multi-scenario load demand forecast data and generate preliminary energy storage capacity demand data; based on the preliminary energy storage capacity demand data, the module performs optimal energy storage scale adaptation interval processing to obtain refined energy storage scale interval data;

[0198] The energy storage structure layout optimization module is used to perform energy storage unit configuration processing according to the refined energy storage scale interval data to generate energy storage unit configuration plan data; optimize the energy storage unit planning layout of the energy storage unit configuration plan data to obtain the optimal energy storage structure layout plan data.

[0199] The beneficial effect of this application is that, starting from historical electricity consumption data, signal processing techniques such as wavelet transform are used to analyze load characteristics at different time scales, and combined with cluster analysis and fuzzy membership calculation, enterprises are divided into different load mode categories, and load fluctuation patterns are refined, overcoming the limitation of focusing only on the overall load and ignoring individual differences. Based on load fluctuation patterns and historical data, multiple future load demand scenarios are simulated, and combined with information such as the company's future production plan, multi-scenario load demand forecast data is generated to avoid the deviation of energy storage scale configuration in a single scenario. Considering the peak-to-valley load difference under different scenarios, the probability distribution and extreme cases of load demand are also analyzed, and quantile calculation and extreme value analysis methods are used. Combined with energy storage targets and different fluctuation levels, the required energy storage capacity is preliminarily estimated, and a multi-objective optimization algorithm is used to comprehensively consider economic performance and feasibility, determine the optimal energy storage scale adaptation interval, and solve the problem of excessive or insufficient energy storage scale configuration. According to the refined energy storage scale interval data and project requirements, a variety of energy storage technologies are selected for combined configuration. At the same time, the plan fully considers the impact of energy storage system access on the power grid. It uses network centrality analysis and power flow calculation to evaluate the impact of different layout schemes on node voltage and line power flow, and determines the optimal energy storage unit access location and capacity, thereby maximizing the effectiveness of the energy storage system and improving the overall performance of the power grid.

[0200] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0201] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be 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 present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for planning and layout of energy storage scale and structure, characterized in that: The following steps are involved: Step S1: Obtain original historical power load data; Draw the power load curve based on the original historical power load data to generate time series power load data; Perform multi-scale load characteristic analysis based on time-series power load data to generate load characteristic vector data; Step S2: performing enterprise load pattern matching on load characteristic vector data based on time-series power load data to obtain enterprise load prototype pattern data; Perform adaptive hierarchical pattern cutting processing based on enterprise load prototype pattern data to generate enterprise load fluctuation pattern data; Step S3: Perform load scenario probability processing on the enterprise load fluctuation pattern data to generate load scenario multi-combination data; calculate the load demand expectation value based on the load scenario multi-combination data to generate multi-scenario load demand forecast data; wherein step S3 is specifically as follows: Step S31: performing empirical mode decomposition on the enterprise load fluctuation mode data and performing independent component analysis to construct load mode characteristic space data; Step S32: performing load scenario probability processing according to the load pattern characteristic space data to generate load scenario multi-combination data; Step S33: Acquire enterprise production and operation data, wherein the enterprise production and operation data includes enterprise production equipment data, production energy consumption type data, production process data and enterprise production plan data; Step S34: extracting operating load influencing factors according to the enterprise production and operation data to obtain key operating influencing factor data; processing load demand constraint conditions based on the key operating influencing factor data to obtain load demand constraint condition data; Step S35: performing load scenario association processing on the load scenario multi-combination data using the load demand constraint condition data, and performing load demand expected value calculation to generate multi-scenario load demand forecast data; Step S4: perform preliminary energy storage demand estimation based on multi-scenario load demand forecast data to generate preliminary energy storage capacity demand data; perform optimal energy storage scale adaptation interval processing based on the preliminary energy storage capacity demand data to obtain refined energy storage scale interval data; Step S5: Perform energy storage unit configuration processing according to the refined energy storage scale interval data to generate energy storage unit configuration plan data; optimize the energy storage unit planning layout of the energy storage unit configuration plan data to obtain optimal energy storage structure layout plan data.

2. The method for planning and layout of energy storage scale and structure according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: obtaining original historical power load data; performing electrical parameter cleaning on the original historical power load data to generate standard historical power load data; Step S12: performing time stamp alignment processing on the standard historical power load data, and drawing a power load curve to generate time series power load data; Step S13: performing multi-scale load feature decomposition processing according to the time series power load data to generate multi-scale load feature point data; Step S14: constructing a characteristic vector based on the multi-scale load characteristic point data to generate load characteristic vector data.

3. The method for planning and layout of energy storage scale and structure according to claim 2, characterized in that: Step S13 includes the following steps: Step S131: using a preset time sliding window to divide the time series power load data into time windows to generate power load window data; Step S132: Calculate the peak and valley values ​​of the power load window data to generate load peak and valley value data; Step S133: classifying the load peak-valley value data into peak-valley feature points by using a preset peak-valley classification threshold to obtain classified peak-valley feature point data, wherein the classified peak-valley feature point data includes peak load feature point data, high-valley load feature point data, and low-valley load feature point data; Step S134: Perform moving standard deviation processing on the power load window data to obtain time series average volatility data; Step S135: using the time series average fluctuation rate data to analyze the long-term and short-term fluctuation trends of the time series power load data, and identifying fluctuation feature points to generate fluctuation feature point data; Step S136: using the fluctuation characteristic point data to perform event load trend analysis on the time series power load data to obtain special event characteristic point data; Step S137: Perform Fourier spectrum characteristic analysis on the time series power load data to generate load spectrum characteristic data; perform seasonal cycle identification based on the load spectrum characteristic data, and extract cycle segment feature points to obtain seasonal cycle feature point data; Step S138: Perform multi-scale feature point integration on the classified peak and valley feature point data, fluctuation feature point data, seasonal cycle feature point data, and special event feature point data to generate multi-scale load feature point data.

4. The method for planning and layout of energy storage scale and structure according to claim 3, characterized in that: Step S136 includes the following steps: Step S1361: Acquire power consumption history event data; perform event type semantic analysis based on the power consumption history event data to generate load event type data; Step S1362: extracting the event time sequence of the electricity consumption historical event data to obtain the historical event time axis data; Step S1363: Processing the time series power load data according to the historical load fluctuation trajectory through the fluctuation characteristic point data to generate historical load fluctuation trajectory data; Step S1364: performing time window positioning processing on the historical load fluctuation trajectory data through the historical event time axis data to generate load event window locking data; Step S1365: extracting event load characteristics according to the load event window locking data, and obtaining load characteristic data before the event, load characteristic data during the event, and load characteristic data after the event; Step S1366: Perform load baseline statistical analysis on the load characteristic data before the event to generate the load baseline data before the event; perform load fluctuation analysis on the load characteristic data during the event to generate the load fluctuation data during the event; perform load event difference analysis on the load characteristic data after the event through the load baseline data before the event to obtain the event load recovery index; Step S1367: performing a weighted calculation of the impact degree of the load fluctuation data in the event and the load recovery index after the event based on the load event type data to generate load event impact degree data; Step S1368: Based on the load event impact degree data, the special event feature point division is performed through the load event window locking data to generate special event feature point data.

5. The method for planning and layout of energy storage scale and structure according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: performing enterprise load aggregation processing on the load characteristic vector data based on the time series power load data to generate enterprise load characteristic vector cluster data; Step S22: performing similarity measurement calculation based on the enterprise load characteristic vector cluster data, and performing typical load pattern matching through a preset load pattern library to obtain enterprise load prototype pattern data; Step S23: constructing a load fuzzy membership model according to the enterprise load prototype pattern data; Step S24: Calculate the uncertainty index of the enterprise load characteristic vector cluster data through the load fuzzy membership model, and perform load pattern soft classification processing to generate soft classification load pattern data; Step S25: performing hierarchical clustering optimization on the soft-classified load pattern data to generate load pattern hierarchical data; Step S26: Adaptively perform hierarchical pattern cutting processing on the load pattern hierarchical data using preset load clustering evaluation indicators to generate enterprise load fluctuation pattern data.

6. The method for planning and layout of energy storage scale and structure according to claim 1, characterized in that: Step S32 includes the following steps: Step S321: performing load scenario encoding processing on the load pattern feature space data through a pre-trained variational autoencoder, and performing load scenario deduction using a generative adversarial network to generate load scenario sequence sample data; Step S322: using a preset recurrent neural network model to perform transfer learning on the load scenario sequence sample data to obtain an initial load scenario evolution model; identifying the scenario transition points according to the initial load scenario evolution model, and performing attention mechanism embedding processing to generate a load scenario evolution model; Step S323: extracting dynamic scene evolution rules from load scene sequence sample data using the load scene evolution model to generate dynamic scene evolution rule data; Step S324: performing Markov decision state transformation processing according to the dynamic scene evolution rule data to generate dynamic scene transition probability matrix data; Step S325: performing Monte Carlo tree search processing on the dynamic scenario transition probability matrix data to obtain a multi-path load scenario tree; Step S326: Perform Simpson probability integral calculation based on the multi-path load scenario tree, and perform load scenario combination processing to generate load scenario multi-combination data.

7. The method for planning and layout of energy storage scale and structure according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: performing multi-scenario fractional calculation based on multi-scenario load demand forecast data to generate load demand fractional data; performing generalized extreme load processing based on multi-scenario load demand forecast data to generate extreme load demand data; Step S42: fitting the load demand quantile data with extreme value distribution through the extreme load demand data, and performing multi-level fluctuation interval processing to generate multi-level load demand fluctuation data; Step S43: performing preliminary energy storage demand estimation on the multi-level load demand fluctuation data to generate preliminary energy storage capacity demand data; Step S44: performing energy storage scale configuration on the preliminary energy storage capacity demand data through a preset energy storage technical parameter library to obtain energy storage scale candidate data; Step S45: Performing an energy storage investment economic evaluation on the energy storage scale candidate data to obtain energy storage investment economic scoring data; performing an energy storage technology feasibility evaluation on the energy storage scale candidate data to obtain energy storage scale feasibility scoring data; Step S46: constructing a multi-objective energy storage structure optimization model through energy storage scale candidate data based on energy storage scale feasibility scoring data, energy storage investment economic scoring data, and preset energy storage scale constraint rules; Step S47: using a non-dominated sorting genetic algorithm to solve the Pareto energy storage structure of the multi-objective energy storage structure optimization model, and performing optimal energy storage scale adaptation interval processing to obtain refined energy storage scale interval data.

8. The method for planning and layout of energy storage scale and structure according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: performing energy storage unit configuration processing according to the refined energy storage scale interval data to generate energy storage unit configuration plan data; Step S52: acquiring power network structure diagram data; performing energy storage node mapping processing on the power network structure diagram data using the energy storage unit configuration scheme data to obtain power energy storage layout structure data; Step S53: extracting the power network topology structure according to the power energy storage layout structure data to obtain the power network topology structure data; Step S54: performing a network centrality analysis of energy storage nodes according to the power network topology data, and performing AC power flow calculation of energy storage nodes, to obtain node voltage sensitivity data and node line power flow distribution data respectively; Step S55: Based on the preset energy storage unit layout target, the node voltage sensitivity data and the node line power flow distribution data are used to optimize the energy storage unit planning layout, and the data envelopment efficiency is scored to obtain the optimal energy storage structure layout solution data.

9. A planning and layout system for energy storage scale and structure, characterized in that: Used to execute the energy storage scale and structure planning and layout method as claimed in claim 1, the energy storage scale and structure planning and layout system comprises: The load characteristic analysis module is used to obtain the original historical power load data; draw the power load curve based on the original historical power load data to generate time series power load data; perform multi-scale load characteristic analysis based on the time series power load data to generate load characteristic vector data; The load pattern recognition module is used to match the enterprise load pattern with the load characteristic vector data based on the time-series power load data to obtain the enterprise load prototype pattern data; perform adaptive hierarchical pattern cutting processing according to the enterprise load prototype pattern data to generate the enterprise load fluctuation pattern data; The multi-scenario load demand forecasting module is used to perform load scenario probability processing on the enterprise load fluctuation pattern data to generate multi-combination data of load scenarios; calculate the expected value of load demand based on the multi-combination data of load scenarios to generate multi-scenario load demand forecasting data; The energy storage scale assessment module is used to make a preliminary energy storage demand estimate based on the multi-scenario load demand forecast data and generate preliminary energy storage capacity demand data; based on the preliminary energy storage capacity demand data, the module performs optimal energy storage scale adaptation interval processing to obtain refined energy storage scale interval data; The energy storage structure layout optimization module is used to perform energy storage unit configuration processing according to the refined energy storage scale interval data to generate energy storage unit configuration plan data; optimize the energy storage unit planning layout of the energy storage unit configuration plan data to obtain the optimal energy storage structure layout plan data.

Images