A multi-dimensional constraint modeling energy storage capacity optimization configuration method and system

Abstract

The application discloses a multi-dimensional constraint modeling energy storage capacity optimization configuration method, and particularly relates to the technical field of energy storage capacity planning, and is used for solving the problems of feasible solution space fragmentation and low global optimization efficiency caused by heterogeneous constraint conflicts in the prior art; the method is characterized in that: multi-dimensional constraint data of a continuous technical parameter, operation condition and discrete policy rule of an energy storage system are acquired; the conflict intensity between constraints is quantified and a dynamic weight is distributed; the discrete rule is converted into a continuous penalty function to construct a constraint model; the topology structure of a solution space is analyzed and its dynamic change is monitored to generate a connectivity atlas and an evolution trend atlas; an adaptive search strategy is generated based on the atlas; a global optimal solution set is outputted and a verifiable energy storage capacity configuration scheme is generated; and through dynamic weight distribution and solution space connectivity guidance, the multi-constraint collaborative optimization capability is significantly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy storage capacity planning, and more particularly, to a multi-dimensional constraint modeling energy storage capacity optimization configuration method and system. BACKGROUND

[0002] In the field of energy storage capacity optimization configuration, the prior art usually converts technical parameters, operating conditions and external policy requirements of the energy storage system into constraint conditions through mathematical modeling, and uses an optimization algorithm to solve the optimal solution set of the objective function, which is widely used in power system planning, renewable energy consumption and grid stability improvement scenarios. However, with the complication of energy storage application scenarios, multi-dimensional constraint conditions (such as device physical characteristics, economic indicators, and environmental policy restrictions) have heterogeneity in mathematical form, logical relationship and time scale, resulting in the need to coordinate conflicts and compatibility of different dimensions in model construction.

[0003] When the prior art deals with the energy storage capacity optimization problem under heterogeneous constraint conditions, the implicit conflicts between different constraints are not fully coordinated, resulting in the feasible solution space being divided into multiple isolated regions, making it difficult for traditional optimization algorithms to effectively cross the solution space fragmentation area when searching for the global optimal solution, and easily falling into local optimization or failing to meet the collaborative requirements of multi-dimensional constraints, ultimately leading to significant reduction in the economic, reliability and compliance of the energy storage capacity configuration scheme. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a multi-dimensional constraint modeling energy storage capacity optimization configuration method and system to solve the problems raised in the background art.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0006] A multi-dimensional constraint modeling energy storage capacity optimization configuration method, comprising the following steps:

[0007] S1, obtaining multi-dimensional constraint data of the energy storage system, including continuous technical parameter constraints, operating condition constraints and discrete policy rule constraints;

[0008] S2, evaluating the intersection missing degree of the feasible region through the implicit conflict between the discrete policy rule constraints and the continuous technical parameter constraints to allocate a dynamic weight to the discrete policy rule constraints;

[0009] S3, converting the boundary threshold of the discrete policy rule constraints into a continuous penalty function, and constructing a fusion constraint model that fuses multi-dimensional constraints in combination with the dynamic weight;

[0010] S4, topological structure analysis is performed on the feasible solution space of the fusion constraint model to construct a connectivity atlas, and dynamic changes of the feasible solution space are monitored in real time to generate a solution space evolution trend atlas;

[0011] S5, a search strategy is generated based on the connectivity atlas and the solution space evolution trend atlas;

[0012] S6, a storage capacity configuration scheme is generated according to the global optimal solution set output by the search strategy.

[0013] In a preferred embodiment, multi-dimensional constraint data of the energy storage system is obtained, including continuous technical parameter constraints, operating condition constraints and discrete policy rule constraints, specifically:

[0014] The continuous technical parameter constraints are divided into power limit type parameters and life attenuation model type parameters;

[0015] The text rules of the discrete policy rule constraints are converted into discrete mathematical expressions containing boundary thresholds and trigger conditions;

[0016] The power load demand and renewable energy output fluctuation range in the operating condition constraints are time-aligned to generate a unified timestamp constraint data set.

[0017] In a preferred embodiment, the intersection missing degree of the feasible region is evaluated by the implicit conflict between the discrete policy rule constraints and the continuous technical parameter constraints to allocate a dynamic weight to the discrete policy rule constraints, including:

[0018] The boundary threshold of the discrete policy rule constraint is projected to the feasible region space of the continuous technical parameter constraint, and the geometric intersection missing area between the boundary threshold and the feasible region of the continuous technical parameter constraint is calculated;

[0019] Based on the ratio of the geometric intersection missing area to the total area of the feasible region of the continuous technical parameter constraint, a conflict intensity coefficient is determined;

[0020] According to the conflict intensity coefficient and the preset weight mapping relationship, a dynamic weight is allocated to the discrete policy rule constraint, and the preset weight mapping relationship is an inverse proportional function relationship between the conflict intensity coefficient and the dynamic weight;

[0021] The dynamic weight is embedded as a priority parameter of the discrete policy rule constraint in the objective function of the multi-objective optimization model.

[0022] In a preferred embodiment, the boundary threshold of the discrete policy rule constraint is converted into a continuous penalty function, and a fusion constraint model that fuses multi-dimensional constraints is constructed by combining the dynamic weight, including:

[0023] The boundary threshold of the discrete policy rule constraint is converted into a piecewise exponential continuous function, and a smooth transition interval is set at the boundary threshold to avoid function mutation;

[0024] The dynamic weight is embedded into the piecewise exponential continuous function as a scaling factor of the exponential term to generate a continuous penalty function term with weight;

[0025] Based on the mathematical expression of the continuous penalty function term with weight and the continuous technical parameter constraint, a fusion constraint model with multi-objective weighted sum as the comprehensive optimization objective is constructed;

[0026] The input parameters of the fusion constraint model are processed for dimension alignment.

[0027] In a preferred embodiment, the topology structure of the feasible solution space of the fusion constraint model is analyzed to construct a connectivity atlas, and the dynamic changes of the feasible solution space are monitored in real time to generate a solution space evolution trend atlas, including:

[0028] A candidate solution set is randomly sampled from the feasible solution space of the fusion constraint model, and the Euclidean distance between the candidate solutions is calculated to construct an adjacency matrix, and the edge weight of the adjacency matrix is the inverse of the distance between the candidate solutions;

[0029] Based on the adjacency matrix, an isolated region in the feasible solution space is identified, and the isolated region is defined as a subset of candidate solutions with edge weights lower than a preset connectivity threshold in the adjacency matrix;

[0030] The boundary solution coordinates and connectivity state changes of the isolated region are recorded in real time to generate a spatiotemporal distribution history record of the isolated region;

[0031] The spatiotemporal distribution history record is analyzed by clustering to extract the migration path and convergence characteristics of the isolated region, and a solution space evolution trend atlas is generated.

[0032] In a preferred embodiment, a search strategy is generated based on the connectivity atlas and the solution space evolution trend atlas, including:

[0033] A virtual transition node is set at the boundary of the isolated region in the feasible solution space;

[0034] Based on the migration path and convergence characteristics in the solution space evolution trend atlas, the connectivity of the isolated region is predicted;

[0035] The spanning path from the current solution to the target solution is calculated by a path planning algorithm;

[0036] The search direction is dynamically adjusted according to the convergence speed of the spanning path.

[0037] In a preferred embodiment, the virtual transition node is an intermediate jump point of the spanning path from the current solution to the target solution.

[0038] The pre-judgment includes calculating the success rate of the historical crossing path and a threshold number of iterations;

[0039] The path planning algorithm preferentially selects a direction with a high density of virtual transition nodes to expand the search branch;

[0040] The dynamic adjustment includes reducing the search weight of a low convergence speed path and increasing the exploration probability of a high convergence speed path.

[0041] In a preferred embodiment, a global optimal solution set output according to the search strategy is used to generate a storage capacity configuration scheme, including:

[0042] From the global optimal solution set, a subset of optimal solutions that meet preset economic, reliability and compliance indicators is selected;

[0043] The solution vectors in the subset of optimal solutions are mapped to storage capacity configuration parameters;

[0044] Based on continuous technical parameter constraints and discrete policy rule constraints, the mapped configuration parameters are verified for feasibility;

[0045] A storage capacity configuration scheme containing configuration parameters, verification results and optimization target weights is generated, and the scheme output is an executable engineering deployment instruction.

[0046] In a preferred embodiment, the storage capacity configuration parameters include storage capacity, power limit, life decay coefficient and charge-discharge strategy.

[0047] In another aspect, the present application provides a multi-dimensional constraint modeling storage capacity optimization configuration system, including the following modules:

[0048] A multi-dimensional data acquisition module acquires multi-dimensional constraint data of the storage system, including continuous technical parameter constraints, operating condition constraints and discrete policy rule constraints;

[0049] A conflict weight allocation module allocates dynamic weights to discrete policy rule constraints by evaluating the degree of intersection loss between discrete policy rule constraints and continuous technical parameter constraints based on implicit conflicts between the two;

[0050] A constraint fusion modeling module converts the boundary threshold of discrete policy rule constraints into a continuous penalty function, and constructs a fusion constraint model that fuses multi-dimensional constraints in combination with dynamic weights;

[0051] A topological analysis monitoring module performs topological structure analysis on the feasible solution space of the fusion constraint model to construct a connectivity atlas, and monitors the dynamic changes of the feasible solution space in real time to generate a solution space evolution trend atlas;

[0052] Search strategy generation module: generate search strategy based on connectivity atlas and solution space evolution trend atlas;

[0053] Configuration scheme generation module: generate energy storage capacity configuration scheme according to the global optimal solution set output by the search strategy.

[0054] Compared with the prior art, the present application has the following beneficial effects:

[0055] 1. By coordinating the conflict of multi-dimensional constraints and the dynamic evolution characteristics of the solution space, the global optimization ability and engineering feasibility of the energy storage capacity configuration are significantly improved. For the heterogeneity of multi-dimensional constraints, the dynamic weight distribution mechanism and the continuous penalty function conversion are used to effectively bridge the implicit conflict between discrete policy rules and continuous technical parameters. The dynamic weight quantifies the conflict intensity based on the intersection loss of the feasible region, so that the priority of high conflict constraints is adaptively reduced to avoid excessive fragmentation of the feasible solution space. At the same time, the continuous penalty function converts the discrete threshold into a smooth transition mathematical expression, which reduces the fragmentation of the solution space while maintaining the effectiveness of policy constraints. The connectivity atlas and evolution trend atlas generated by combining topological structure analysis and real-time monitoring can accurately locate the boundary of isolated regions and predict the migration path of the solution space, providing dynamic guidance for the optimization algorithm and significantly improving the quality and search efficiency of global solutions.

[0056] 2. A full-process closed-loop logic from multi-dimensional constraint modeling to configuration scheme generation is constructed to ensure that the optimization results have both technical compliance and engineering operability. By integrating multi-dimensional parameters through constraint model fusion, technical characteristics, operational requirements and policy rules can work together in a unified mathematical framework. The search strategy driven by the solution space evolution trend atlas, combined with the convergence characteristics of historical paths and virtual transition node guidance, realizes intelligent crossing of isolated regions of the solution space and dynamic optimization of search direction. The final generated configuration scheme is verified through multiple levels of feasibility and output as an engineering instruction that can be directly deployed. Furthermore, through dynamic weight and real-time feedback mechanism, the system maintains high adaptability and robustness in complex and variable operating environments. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 A flowchart of a multi-dimensional constraint modeling energy storage capacity optimization configuration method of the present application;

[0058] Figure 2 A structural schematic diagram of a multi-dimensional constraint modeling energy storage capacity optimization configuration system of the present application. DETAILED DESCRIPTION

[0059] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all the other embodiments obtained by those ordinarily skilled in the art without creative effort belong to the scope of the present application.

[0060] Embodiment 1 Figure 1 A multi-dimensional constraint modeling energy storage capacity optimization configuration method is given, which comprises the following steps:

[0061] S1, obtaining multi-dimensional constraint data of the energy storage system, including continuous technical parameter constraints, operating condition constraints and discrete policy rule constraints;

[0062] S2, evaluating the intersection missing degree of the feasible region by the implicit conflict between the discrete policy rule constraints and the continuous technical parameter constraints to allocate a dynamic weight to the discrete policy rule constraints;

[0063] S3, converting the boundary threshold of the discrete policy rule constraints into a continuous penalty function, combining the dynamic weight to build a fusion constraint model integrating multi-dimensional constraints;

[0064] S4, performing topological structure analysis on the feasible solution space of the fusion constraint model to build a connectivity atlas, and monitoring the dynamic change of the feasible solution space in real time to generate a solution space evolution trend atlas;

[0065] S5, generating a search strategy based on the connectivity atlas and the solution space evolution trend atlas;

[0066] S6, generating an energy storage capacity configuration scheme according to the global optimal solution set output by the search strategy.

[0067] S1, obtaining multi-dimensional constraint data of the energy storage system, including continuous technical parameter constraints, operating condition constraints and discrete policy rule constraints, and the specific implementation is:

[0068] The continuous type technical parameter constraints are divided into power limit type parameters and life attenuation model type parameters. The power limit type parameters include the rated power, the maximum charge and discharge rate, and the instantaneous power fluctuation tolerance range of the energy storage device, which are obtained by reading the technical specification document of the energy storage device or based on the statistical results of the measured data; the life attenuation model type parameters include the mapping relationship between the cycle number and the capacity attenuation coefficient of the energy storage device, and the temperature influence coefficient on the life, which are obtained by the fitting curve of the laboratory accelerated aging test or the regression analysis of the historical operation data. The power limit type parameters and the life attenuation model type parameters after the division are stored as independent data tables, and the values in the data tables are normalized. The normalization processing adopts the minimum-maximum value method, and the parameter values are linearly converted to the [0, 1] interval, so as to ensure the comparability of parameters with different dimensions in subsequent modeling. For example, the original data range of the rated power is 0 to 10 megawatts, and after normalization, the corresponding unit value is 0 to 1; the mapping relationship between the cycle number and the capacity attenuation coefficient is fitted by a quadratic function, and the cycle number is normalized to the [0, 1] interval, and the capacity attenuation coefficient is scaled synchronously.

[0069] The text rules of the discrete type policy rule constraints are converted into discrete mathematical expressions containing boundary thresholds and trigger conditions. Specifically, the constraint clauses related to energy storage operation are extracted from the policy file, such as "charging operation is prohibited when carbon emissions exceed 50 kg per megawatt", which is converted into a logical judgment formula. The form of the logical judgment formula is "if condition A is met, then operation B is performed", where condition A is the numerical comparison of the boundary threshold, and operation B is the behavior restriction triggered by the constraint. For policy rules containing multiple conditions, such as "the discharge power of the energy storage is not less than 20% of the total load and the SOC is higher than 40% when the peak period price is low", it is converted into a compound logic expression, which is in the form of "IF (time ∈ peak period) AND (discharge power ≥ total load × 20%) AND (SOC > 40%) THEN allow discharge". The peak period is defined as 8:00 to 12:00 and 18:00 to 22:00 every day, which is realized by time stamp interval matching. The converted discrete mathematical expressions are stored in the database and are associated with the continuous type technical parameter constraints through parameter name indexing.

[0070] The grid load demand and renewable energy output fluctuation range in the operation condition constraint are time-aligned to generate a constraint data set with a unified timestamp. The grid load demand data is provided in the form of historical records or predicted values at an interval of one hour, and the renewable energy output fluctuation range is provided in the form of upper and lower power curves at an interval of 15 minutes. The data with different time resolutions are aligned to a unified time reference through an interpolation algorithm, and the interpolation algorithm selects linear interpolation or cubic spline interpolation according to the data type, for example, linear interpolation is used when the load data changes gently, and cubic spline interpolation is used for photovoltaic output curves due to large fluctuations. The aligned data set includes the load demand value, the upper limit value and the lower limit value of the renewable energy output at each time point, and the timestamp format is "year-month-day hour: minute: second", for example, "2024-10-05 08:15:00" corresponds to a load demand of 100 megawatts, a photovoltaic output upper limit of 30 megawatts, and a wind power output lower limit of 10 megawatts. After the time alignment processing is completed, the integrity of the data set is verified, and the missing data at the time point is filled in according to the following rules: if the consecutive missing time points do not exceed 3, the average value of the adjacent time points is used to fill in; if more than 3, the average value of the same time period of the adjacent date is used to fill in, and the abnormal value is removed according to the following rules: if the load data exceeds 120% of the historical maximum value or is lower than 80% of the historical minimum value, it is considered abnormal, and if the photovoltaic output exceeds 100% of the installed capacity, it is considered abnormal, and the abnormal value is replaced by the linear interpolation of the two valid values before and after it. The final generated constraint data set is associated with the discrete mathematical expressions of the power limit type parameters, the life attenuation model type parameters, the boundary threshold and the trigger condition through the timestamp, forming a unified input interface of multi-dimensional constraint data.

[0071] It is worth noting that in step S1, the operation condition constraint (such as grid load demand, renewable energy output fluctuation range) is the core component of the multi-dimensional constraint data, providing dynamic input for subsequent conflict coordination and model construction; in step S3, the operation condition constraint generates a constraint data set with a unified timestamp through time alignment processing, ensuring the consistency of the variables of the integrated constraint model in the time dimension; in step S4, the dynamic change of the operation condition constraint directly affects the topology structure of the feasible solution space and the generation of the evolution trend map, for example, the load fluctuation causes the real-time adjustment of the solution space form; in step S6, the operation condition constraint is a key basis for verifying the feasibility of the configuration scheme, ensuring that the charging and discharging strategy matches the actual operation demand of the grid.

[0072] S2, through the implicit conflict between discrete policy rule constraints and continuous technology parameter constraints, the intersection missing degree of the feasible region is evaluated to allocate dynamic weights to the discrete policy rule constraints, which is implemented as follows:

[0073] The boundary threshold of the discrete policy rule constraint is projected to the feasible region space of the continuous technical parameter constraint, and the geometric intersection missing area between the boundary threshold and the feasible region of the continuous technical parameter constraint is calculated. Specifically, the projection operation is performed by mapping the boundary threshold of the discrete policy rule constraint to the feasible region coordinate system constituted by the value range of the continuous technical parameter constraint. For example, the power limit feasible region of the continuous technical parameter constraint is 0 to 10 megawatts, and the discrete policy rule constraint requires that the power should not exceed 8 megawatts. Then, the projected feasible region coverage area is 0 to 8 megawatts, the non-overlapping part is 8 to 10 megawatts, and the geometric intersection missing area is the area corresponding to 2 megawatts. The calculation method of the geometric intersection missing area is the polygon area superposition method or the numerical integration method. Specifically, according to the shape of the feasible region, the long and wide product is calculated for a rectangular feasible region, and the integral calculation is performed after the piecewise linear approximation for a curved boundary feasible region.

[0074] The conflict intensity coefficient is determined based on the ratio of the geometric intersection missing area to the total area of the feasible region of the continuous technical parameter constraint. The total area of the feasible region is the area corresponding to the value range of the continuous technical parameter constraint. For example, the total area of the power limit feasible region of 0 to 10 megawatts is the area corresponding to 10 megawatts. The calculation formula of the conflict intensity coefficient is the geometric intersection missing area divided by the total area of the feasible region, and the result is a value between 0 and 1. For example, when the geometric intersection missing area is 2 megawatts corresponding to the area, and the total area of the feasible region is 10 megawatts corresponding to the area, the conflict intensity coefficient is 0.2. When the missing area is 8 megawatts corresponding to the area, the conflict intensity coefficient is 0.8. The conflict intensity coefficient is used to quantify the conflict degree between the discrete policy rule constraint and the continuous technical parameter constraint. The larger the value is, the stronger the conflict is.

[0075] The dynamic weight of the discrete policy rule constraint is allocated according to the conflict intensity coefficient and the preset weight mapping relationship. The preset weight mapping relationship is an inverse proportional function relationship between the conflict intensity coefficient and the dynamic weight. Specifically, the dynamic weight is equal to the reference weight divided by the product of the conflict intensity coefficient and the adjustment factor, wherein the reference weight is a preset fixed value (for example, the reference weight is 1.0), and the adjustment factor is used to control the sensitivity of the weight change with the conflict intensity (for example, the adjustment factor is 2.0). For example, when the conflict intensity coefficient is 0.2, the dynamic weight is 1.0 / (0.2×2.0)=2.5. When the conflict intensity coefficient is 0.8, the dynamic weight is 1.0 / (0.8×2.0)=0.625. Through the inverse proportional function relationship, the discrete policy rule constraint with low conflict intensity is allocated with high weight to strengthen the constraint, and the constraint with high conflict intensity is allocated with low weight to weaken its limitation on the feasible region.

[0076] The dynamic weight is embedded into the objective function of the multi-objective optimization model as a priority parameter of the discrete policy rule constraint. The objective function is composed of multiple-dimensional sub-targets such as economy, reliability, compliance, etc. The original weight of each sub-target is preset according to actual demand (for example, the economy weight is 0.6, the reliability weight is 0.3, and the compliance weight is 0.1). After the dynamic weight adjustment, the weight of the compliance sub-target is the original compliance weight multiplied by the dynamic weight, for example, when the dynamic weight is 2.5, the compliance weight is adjusted to 0.1 x 2.5 = 0.25; when the dynamic weight is 0.625, the compliance weight is adjusted to 0.1 x 0.625 = 0.0625. The adjusted objective function combines each sub-target in the form of weighted summation, and is input into the optimization algorithm for solution, for example, the total objective function expression is: economy index x 0.6 + reliability index x 0.3 + compliance index x adjusted compliance weight.

[0077] S3, the boundary threshold of the discrete policy rule constraint is converted into a continuous penalty function, and a fusion constraint model integrating multiple-dimensional constraints is constructed in combination with the dynamic weight, which is specifically implemented as:

[0078] The boundary threshold of the discrete policy rule constraint is converted into a piecewise exponential continuous function, and a smooth transition interval is set at the boundary threshold to avoid function mutation. Specifically, for the boundary threshold of the discrete policy rule constraint, such as "the discharge power of the energy storage system in the peak period shall not be less than 20 megawatts", first determine its threshold range (such as discharge power ≥ 20 megawatts), and define a smooth transition interval (such as 18 megawatts to 22 megawatts) near the threshold. In the smooth transition interval, the penalty function value starts from zero and increases exponentially: when the discharge power is less than 18 megawatts, the penalty value is zero; in the interval of 18 megawatts to 22 megawatts, the penalty value gradually increases with the decrease of the discharge power according to the exponential function, and the growth rate of the exponential function is determined by the width of the smooth transition interval (for example, when the interval width is 4 megawatts, the growth rate is set to double the penalty value per megawatt); when the discharge power exceeds 22 megawatts, the penalty value remains constant. The setting of the smooth transition interval avoids the mutation of the traditional step function at the threshold, ensuring that the objective function is derivable and facilitating the convergence of the optimization algorithm.

[0079] The dynamic weight is embedded into the piecewise exponential continuous function as a scaling factor of the exponential term to generate a weighted continuous penalty function term. The dynamic weight is calculated based on the conflict intensity coefficient in step S2, and it is embedded by multiplying the dynamic weight with the power parameter of the exponential function, thereby adjusting the growth rate of the penalty function. For example, if the growth rate of the original exponential function is doubled per megawatt of penalty value (i.e., the exponential term coefficient is 0.6931), and the dynamic weight is 2.5, the adjusted growth rate becomes 2.5 times per megawatt of penalty value (i.e., the exponential term coefficient is 0.6931 x 2.5 ≈ 1.7328). Through dynamic weight scaling, when the conflict intensity is low (the dynamic weight is high), the penalty function grows more steeply in the transition interval, strengthening the constraint on behaviors close to the threshold; when the conflict intensity is high (the dynamic weight is low), the function grows more gently, weakening the constraint effect to alleviate the solution space fragmentation.

[0080] Based on the mathematical expressions of the weighted continuous penalty function term and the continuous technical parameter constraints, a fusion constraint model is constructed with a multi-objective weighted sum as the comprehensive optimization objective. The multi-objective weighted sum model linearly superimposes the economic, reliability, compliance, and other sub-objectives according to the preset weights, for example, the economic sub-objective is the minimization of the total cost of the energy storage system, represented as the sum of the costs; the reliability objective is the minimization of the square sum of the power supply gap, represented as the square sum of the difference between the actual load and the power supply capacity; the compliance objective is the minimization of the sum of all weighted penalty function terms. The objective function expression of the fusion constraint model is the weighted sum of the above sub-objectives, and the weights of each sub-objective are preset according to the actual scene requirements, for example, the economic weight is set to 0.6, the reliability weight is set to 0.3, and the compliance weight is set to 0.1. The constraint conditions of the fusion constraint model include the continuous technical parameter constraints obtained in step S1 (such as power limit ≤ 10 megawatts) and the continuous penalty function terms converted from the discrete policy rule constraints (such as the penalty term corresponding to the discharge power ≥ 20 megawatts).

[0081] The input parameters of the fusion constraint model are dimensionally aligned to ensure consistency of discrete policy rules and continuous technical parameters in time and space scales. Time scale alignment is achieved by unifying timestamps, such as aligning peak period definitions in discrete policy rules (e.g., 8:00 to 12:00 daily) with hourly or minute-level time series data of continuous technical parameters to the same time reference (e.g., 15-minute interval timestamps). Spatial scale alignment is achieved by unit unification, such as unifying power units to megawatts and carbon emissions units to kilograms per megawatt-hour. The aligned input parameters are integrated through database association indexing to ensure that all constraint parameters at the same time point can be called synchronously, such as the load demand being 100 megawatts, the photovoltaic output limit being 30 megawatts, and the carbon emissions being 40 kilograms at the time point "2024-10-05 08:15:00", and the time point being in the peak period. Then, the discrete policy rule constraints corresponding to the peak period (e.g., discharge power ≥ 20 megawatts) are called and converted into continuous penalty function terms, which are input into the fusion constraint model together with other parameters.

[0082] S4, topological structure analysis of the feasible solution space of the fusion constraint model is performed to construct a connectivity map, and real-time monitoring of the dynamic changes of the feasible solution space is performed to generate a solution space evolution trend map. The specific implementation is as follows:

[0083] A set of candidate solutions is randomly sampled from the feasible solution space of the fusion constraint model, and the Euclidean distance between the candidate solutions is calculated to construct an adjacency matrix. The edge weight of the adjacency matrix is the inverse of the distance between the candidate solutions. The feasible solution space of the fusion constraint model is defined by the weighted continuous penalty function terms generated in step S3 and the continuous technical parameter constraints. The candidate solution set is obtained by random sampling using the Monte Carlo method. Each candidate solution contains multi-dimensional parameters such as energy storage capacity, power limit, and life decay coefficient. The calculation of Euclidean distance is based on the coordinates of candidate solutions in the feasible solution space. The edge weight of the adjacency matrix is set as the inverse of the Euclidean distance, for example, when the distance is 2.0049, the edge weight is 1 divided by 2.0049. The row and column indices of the adjacency matrix correspond to the candidate solution number, and the matrix element value represents the connectivity strength between the corresponding candidate solutions.

[0084] An isolated region is defined as a subset of candidate solutions in the adjacency matrix whose edge weights are lower than a preset connectivity threshold, which is set according to historical data or experience, for example, the threshold is set to 0.1, indicating that candidate solutions with edge weights lower than 0.1 are considered to have no connectivity, and the identification method of the isolated region is to traverse each row of the adjacency matrix, if all edge weights of a row are lower than the threshold, the candidate solution corresponding to the row is marked as a member of the isolated region, for example, if all edge weights in the adjacency row of candidate solution A are less than 0.1, candidate solution A belongs to an isolated region, and the boundary solution coordinates of the isolated region are determined by geometric space analysis, for example, in a two-dimensional feasible solution space, the boundary solution of the isolated region is the peripheral contour point set.

[0085] The boundary solution coordinates and connectivity state changes of the isolated region are recorded in real time to generate a spatiotemporal distribution history record of the isolated region, which includes a timestamp, boundary solution coordinates and connectivity state. For example, at the timestamp "2024-10-05 08:15:00", the power limit range of the isolated region boundary solution is recorded as 8 to 10 megawatts, the lifetime decay coefficient range is 0.2 to 0.3, and the nearest distance between the isolated region and the adjacent region is marked as 1.5 megawatts, and the real-time record is stored through a time series database, and the state of the isolated region at each time point is aligned with the unified timestamp in step S1.

[0086] The spatiotemporal distribution history record is subjected to cluster analysis to extract the migration path and convergence characteristics of the isolated region, and a solution space evolution trend map is generated, the cluster analysis uses the K-means algorithm, and the spatiotemporal distribution history record is divided into several categories according to the position, area and connectivity state of the isolated region, for example, the cluster result shows that the isolated region presents periodic migration in the power limit dimension and converges to the low decay area in the lifetime decay dimension, the migration path is generated by connecting the center points of the same cluster, and the convergence characteristics are quantified by the cluster variance, and the solution space evolution trend map is displayed in the form of a spatiotemporal heat map, the horizontal axis is the timestamp, the vertical axis is the power limit or the lifetime decay coefficient, and the color depth represents the frequency of the isolated region.

[0087] S5, generating a search strategy based on the connectivity map and the solution space evolution trend map, the specific implementation is:

[0088] A virtual transition node is set at the boundary of the isolated region in the feasible solution space, the virtual transition node is an intermediate jump point of the path from the current solution to the target solution, specifically, the position of the virtual transition node is set based on the isolated region boundary solution coordinates identified in step S4, for example, when the boundary solution of the isolated region is 8-10 megawatts in the power limit dimension and 0.2-0.3 in the life attenuation coefficient dimension, a virtual transition node is inserted near the coordinate point of power=8 megawatts and life attenuation coefficient=0.3, the insertion rule of the virtual transition node is: a plurality of candidate points are generated around the boundary solution, the candidate point with the smallest sum of distances from the current solution and the target solution is selected as the virtual transition node, the virtual transition node does not participate in the calculation of the actual solution, and is only used to guide the search direction of the path.

[0089] Based on the migration path and convergence characteristics in the solution space evolution trend map, the connectivity of the isolated region is predicted, and the prediction process includes calculating the success rate of the historical crossing path and the iteration number threshold. The success rate of the historical crossing path is determined by calculating the ratio of the number of path connectivity to the number of attempts at the same isolated region at past time points in the time and space distribution history record in step S4, for example, if a certain isolated region has a certain proportion of paths successfully crossing in the history search, it is determined that its connectivity is higher, the iteration number threshold is set according to the average iteration number of the historical successful crossing path, for example, the threshold is set to be the average iteration number of the historical successful cases plus twice the standard deviation, if the iteration number of the current search exceeds the threshold and the success rate is lower than the preset lower limit, it is determined that the isolated region is not connectable and the search is terminated.

[0090] The crossing path from the current solution to the target solution is calculated by a path planning algorithm, the path planning algorithm preferentially selects the direction with high virtual transition node density to expand the search branch, the path planning algorithm uses an improved A* algorithm, which introduces the virtual transition node density as a weight factor in the heuristic function, the virtual transition node density is calculated by counting the number of virtual transition nodes in the candidate search direction, for example, if the number of virtual transition nodes in direction A is significantly more than that in direction B, the search branch in direction A is preferentially expanded, and the generation logic of the crossing path is: starting from the current solution, generating child nodes in the direction with high virtual transition node density, and gradually approaching the target solution.

[0091] The search direction is dynamically adjusted according to the convergence speed of the crossing path, the convergence speed is measured by the drop amplitude of the objective function value per unit iteration number, for example, a path with fast drop is identified as a high convergence speed path, the dynamic adjustment rule is: reducing the search weight of the low convergence speed path and increasing the exploration probability of the high convergence speed path. The adjustment method of the search weight is: if the path convergence speed is lower than the preset threshold, the weight is proportionally decayed, if it is higher than the threshold, the weight is proportionally increased, the exploration probability is distributed through the normalized weight, for example, the path with high weight obtains higher exploration priority.

[0092] S6, generating a storage capacity configuration scheme from the global optimal solution set output by the search strategy, specifically implemented as:

[0093] From the global optimal solution set, filter an optimal solution subset that meets the preset economic, reliability and compliance indicators. The global optimal solution set is output by the search strategy in step S5, containing the optimization objective function values and corresponding solution vectors of multiple candidate solutions. The filtering process is based on the preset thresholds of economic, reliability and compliance indicators, for example, the economic indicator requires that the storage investment cost be less than 1000 million yuan, the reliability indicator requires that the power supply gap be less than 5 megawatt hours, and the compliance indicator requires that the carbon emissions be less than 50 kilograms per megawatt hour. When filtering, each candidate solution in the global optimal solution set is traversed, and if its economic, reliability and compliance indicators all meet the preset thresholds, it is added to the optimal solution subset. If there are multiple candidate solutions that meet the conditions, sort them from small to large according to the optimization objective function values, and select the top 10% solutions as the optimal solution subset.

[0094] Map the solution vectors in the optimal solution subset to the storage capacity configuration parameters, including storage capacity, power limit, life attenuation coefficient and charging and discharging strategy. The solution vector is a combination of multiple dimensions, for example, the first dimension of the solution vector corresponds to the storage capacity (unit: megawatt hours), the second dimension corresponds to the power limit (unit: megawatt), the third dimension corresponds to the life attenuation coefficient, and the fourth dimension corresponds to the trigger threshold of the charging and discharging strategy. In the mapping process, each dimension of the solution vector is converted into an actual configuration parameter according to a preset rule, for example, the storage capacity dimension is multiplied by a reference coefficient (such as 10) to obtain the actual capacity value, the power limit dimension is directly used as the maximum charging and discharging rate, the life attenuation coefficient is used as an input parameter of the capacity attenuation model, and the charging and discharging strategy trigger threshold is converted into a time or power condition.

[0095] Based on the continuous type technical parameter constraints and discrete type policy rule constraints, the mapped configuration parameters are verified for feasibility. The verification process includes two levels: the first level checks whether the configuration parameters meet the continuous type technical parameter constraints in step S1, for example, verifying whether the power limit is within the device rated power range (such as 0 to 10 megawatts) and whether the life attenuation coefficient is within the laboratory test allowed interval (such as 0 to 0.5); the second level checks whether the configuration parameters meet the converted discrete type policy rule constraints in step S1, for example, verifying whether the charging and discharging strategy meets the peak period discharging power requirement (such as not less than 20 megawatts) and whether the carbon emissions are below the policy threshold (such as 50 kilograms per megawatt hour). If any level fails the verification, the configuration parameter is excluded and the process is backtracked to step S5 to regenerate candidate solutions.

[0096] The energy storage capacity configuration scheme including the configuration parameters, the verification results and the optimization target weight is generated, and the scheme output is an executable engineering deployment instruction. In the configuration scheme, the parameters such as energy storage capacity and power limit are expressed in standard engineering units (such as megawatt-hour and megawatt), the verification result is marked as “pass” or “fail” and the specific failure reason, and the optimization target weight inherits the dynamic weight distribution result in step S3 (such as economic weight 0.6 and compliance weight 0.25). The format of the engineering deployment instruction is set according to the interface requirements of the target system, for example, converted into a JSON format configuration file, including an energy storage device parameter table, a charging and discharging strategy time table and a constraint compliance declaration. After the deployment instruction is transmitted to the energy storage control system through the industrial communication protocol, the configuration parameters are automatically loaded and executed.

[0097] Embodiment 2: Figure 2 The structure diagram of the energy storage capacity optimization configuration system of the multi-dimensional constraint modeling is given, and the energy storage capacity optimization configuration system of the multi-dimensional constraint modeling comprises the following modules:

[0098] The multi-dimensional data acquisition module acquires the multi-dimensional constraint data of the energy storage system, including continuous technical parameter constraints, operating condition constraints and discrete policy rule constraints;

[0099] The conflict weight distribution module distributes dynamic weights to the discrete policy rule constraints by evaluating the missing degree of the feasible region intersection between the discrete policy rule constraints and the continuous technical parameter constraints;

[0100] The constraint fusion modeling module converts the boundary threshold of the discrete policy rule constraints into a continuous penalty function, and constructs a fusion constraint model of the fused multi-dimensional constraints in combination with the dynamic weights;

[0101] The topology analysis monitoring module performs topology structure analysis on the feasible solution space of the fusion constraint model to construct a connectivity atlas, and monitors the dynamic changes of the feasible solution space in real time to generate a solution space evolution trend atlas;

[0102] The search strategy generation module generates a search strategy based on the connectivity atlas and the solution space evolution trend atlas;

[0103] The configuration scheme generation module generates an energy storage capacity configuration scheme according to the global optimal solution set output by the search strategy.

[0104] The above embodiments can be realized by software, hardware, firmware or any combination thereof, and when realized by software, the above embodiments can be realized in the form of a computer program product.

[0105] Those skilled in the art can understand that the modules and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software manner depends on the specific application of the technical solution and the constraints of the invention. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0106] In addition, each functional module in each embodiment of the present application can be integrated in one processing module, or each module can exist physically alone, or two or more modules can be integrated in one module.

[0107] The above is merely specific embodiments of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0108] Finally: the above is only the preferred embodiment of the present application, and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for multi-dimensional constrained modeling of energy storage capacity optimization configuration, characterized in that, The method comprises the following steps: S1, obtaining multi-dimensional constraint data of the energy storage system, including continuous technical parameter constraints, operating condition constraints and discrete policy rule constraints; S2, evaluating the intersection missing degree of the feasible region through the implicit conflict between the discrete policy rule constraints and the continuous technical parameter constraints to allocate a dynamic weight to the discrete policy rule constraints, comprising: projecting the boundary threshold of the discrete policy rule constraints to the feasible region space of the continuous technical parameter constraints, calculating the geometric intersection missing area between the boundary threshold and the feasible region of the continuous technical parameter constraints; determining a conflict intensity coefficient based on the ratio of the geometric intersection missing area to the total area of the feasible region of the continuous technical parameter constraints; allocating a dynamic weight to the discrete policy rule constraints according to the conflict intensity coefficient and a preset weight mapping relationship, the preset weight mapping relationship being an inverse proportional function relationship between the conflict intensity coefficient and the dynamic weight; embedding the dynamic weight as a priority parameter of the discrete policy rule constraints into an objective function of a multi-objective optimization model; S3, converting the boundary threshold of the discrete policy rule constraints into a continuous penalty function, and constructing a fusion constraint model integrating multi-dimensional constraints in combination with the dynamic weight; S4, performing topological structure analysis on the feasible solution space of the fusion constraint model to construct a connectivity atlas, and monitoring the dynamic change of the feasible solution space in real time to generate a solution space evolution trend atlas; S5, generating a search strategy based on the connectivity atlas and the solution space evolution trend atlas; S6, generating an energy storage capacity configuration scheme according to a global optimal solution set output by the search strategy.

2. The method of claim 1, wherein, The multi-dimensional constraint data of the energy storage system comprises continuous technical parameter constraints, operating condition constraints and discrete policy rule constraints, specifically: dividing the continuous technical parameter constraints into power limit type parameters and life decay model type parameters; converting the text rules of the discrete policy rule constraints into discrete mathematical expressions containing boundary thresholds and trigger conditions; performing time sequence alignment processing on the power grid load demand and renewable energy output fluctuation range in the operating condition constraints to generate a unified timestamp constraint data set.

3. The method of claim 1, wherein, The boundary threshold of the discrete policy rule constraints is converted into a continuous penalty function, and a fusion constraint model integrating multi-dimensional constraints is constructed in combination with the dynamic weight, comprising: converting the boundary threshold of the discrete policy rule constraints into a segmented exponential continuous function, and setting a smooth transition interval at the boundary threshold to avoid function mutation; embedding the dynamic weight as a scaling factor of the exponential term into the segmented exponential continuous function to generate a continuous penalty function item with weight; based on the continuous penalty function item with weight and the mathematical expression of the continuous technical parameter constraints, constructing a fusion constraint model with a multi-objective weighted sum as a comprehensive optimization target; performing dimension alignment processing on the input parameters of the fusion constraint model.

4. The method of claim 1, wherein, The topological structure of the feasible solution space of the fusion constraint model is analyzed to construct a connectivity atlas, and the dynamic change of the feasible solution space is monitored in real time to generate a solution space evolution trend atlas, comprising: Randomly sample from the feasible solution space of the fusion constraint model to generate a candidate solution set, calculate the Euclidean distance between candidate solutions to construct an adjacency matrix, and the edge weight of the adjacency matrix is the inverse of the distance between candidate solutions; Based on the adjacency matrix, identify isolated regions in the feasible solution space, and an isolated region is defined as a subset of candidate solutions in the adjacency matrix with edge weights below a pre-set connectivity threshold; Real-time record the boundary solution coordinates and connectivity state changes of isolated regions, and generate a spatio-temporal distribution history record of isolated regions; Cluster analysis of the spatio-temporal distribution history record, extract the migration path and convergence characteristics of the isolated region, and generate a solution space evolution trend map.

5. The method of claim 1, wherein, Based on the connectivity map and the solution space evolution trend map, generate a search strategy, including: Set a virtual transition node at the boundary of the isolated region in the feasible solution space; Based on the migration path and convergence characteristics in the solution space evolution trend map, predict the connectivity of the isolated region; Calculate the spanning path from the current solution to the target solution through the path planning algorithm; Dynamically adjust the search direction according to the convergence speed of the spanning path.

6. The method of claim 5, wherein, The virtual transition node is an intermediate jump point in the spanning path from the current solution to the target solution; Prediction includes calculating the success rate of historical spanning paths and the iteration number threshold; The path planning algorithm preferentially selects the direction with high density of virtual transition nodes to expand the search branch; Dynamic adjustment includes reducing the search weight of low convergence speed paths and increasing the exploration probability of high convergence speed paths.

7. The method of claim 1, wherein, Generate a storage capacity configuration scheme based on the global optimal solution set output by the search strategy, including: Filter the optimal solution subset that meets the pre-set economic, reliability and compliance indicators from the global optimal solution set; Map the solution vectors in the optimal solution subset to storage capacity configuration parameters; Based on continuous technical parameter constraints and discrete policy rule constraints, verify the feasibility of the mapped configuration parameters; Generate a storage capacity configuration scheme containing configuration parameters, verification results and optimization target weights, and the scheme output is an executable engineering deployment instruction.

8. The method of claim 7, wherein, The storage capacity configuration parameters include storage capacity, power limit, life decay coefficient and charging and discharging strategy.

9. A multi-dimensional constraint modeling energy storage capacity optimization configuration system for implementing the multi-dimensional constraint modeling energy storage capacity optimization configuration method of any one of claims 1-8, characterized in that, Including the following modules: Multi-dimensional data acquisition module: Acquire multi-dimensional constraint data of the energy storage system, including continuous technical parameter constraints, operating condition constraints and discrete policy rule constraints; Conflict weight allocation module: Through the implicit conflict between discrete policy rule constraints and continuous technical parameter constraints, evaluate the degree of intersection loss of feasible region to allocate dynamic weight to discrete policy rule constraints; Constraint fusion modeling module: Convert the boundary threshold of discrete policy rule constraints into a continuous penalty function, and combine the dynamic weight to build a fusion constraint model that fuses multi-dimensional constraints; Topology analysis monitoring module: Perform topology structure analysis on the feasible solution space of the fusion constraint model to construct a connectivity map, and monitor the dynamic changes of the feasible solution space in real time to generate a solution space evolution trend map; Search strategy generation module: Generate a search strategy based on the connectivity map and the solution space evolution trend map; Configuration scheme generation module: Generate a storage capacity configuration scheme based on the global optimal solution set output by the search strategy.

Images