System mode deduction method based on source storage full life cycle characteristics and power grid growth

Abstract

The invention designs a system mode deduction method based on source storage full life cycle characteristics and power grid growth, and belongs to the technical field of power system deduction. According to the method, an evolution boundary model is constructed by analyzing full life cycle characteristics of thermal power, wind power, photovoltaic power and energy storage and considering factors such as load increase and technical progress. Then, a source storage planning model and a power grid growth model based on the complex network theory are constructed, the source storage planning installed capacity can be obtained, and the positions of newly-added load nodes and power supply nodes are determined; furthermore, a source network storage collaborative planning model is constructed, and typical day data is used as input to solve a stage system mode. And finally, determining whether the deduction process is ended or not by judging whether the evolution period is reached or not. The method is suitable for researching and analyzing the future form of the novel power system, and has the advantages of flexible time scale, simple calculation, high expansibility and the like.

Description

Technical Field

[0001] This invention belongs to the field of power system simulation technology and proposes a system mode simulation method based on the full life cycle characteristics of power source and storage and power grid growth. Background Technology

[0002] Power system simulation, as a key means to ensure grid security, optimize planning decisions, and adapt to energy transition, is of great significance for building a reliable, efficient, and low-carbon modern power system. However, traditional power system simulation methods rely on the deterministic characteristics of loads and power sources, making it difficult to reflect the stochastic and dynamic coupling characteristics of the power generation, grid, load, and storage systems, thus having significant limitations.

[0003] Modern power system simulation is essentially a high-dimensional, nonlinear, time-varying dynamic process involving multiple influencing factors, with numerous decision variables, making large-scale computation difficult. Analyzing power grid evolution by mimicking general network growth patterns can avoid solving some optimization problems. Furthermore, the entire lifecycle of power generation and storage exhibits characteristics such as multi-timescale coupling, spatial synergy, and techno-economic mutual benefit. Viewing system evolution as a multi-stage optimization decision-making problem based on the lifecycle characteristics of power generation and storage combined with power grid growth can effectively reflect the stochasticity and dynamic coupling of power generation, grid, load, and storage components during system evolution. Therefore, a system simulation method based on the lifecycle characteristics of power generation and storage and power grid growth is proposed. Summary of the Invention

[0004] To address the problem that traditional power system simulation methods are limited to the deterministic characteristics of load and power source ends, making it difficult to reflect the randomness and dynamic coupling characteristics of each link of source, grid, load and storage, this invention proposes a system mode simulation method based on the full life cycle characteristics of source and storage and grid growth. Its purpose is to provide the grid with a more reliable and accurate modern power system mode simulation method.

[0005] The system mode deduction method proposed in this invention analyzes the full life cycle characteristics of thermal power, wind power, photovoltaic power, and energy storage, and constructs an evolutionary boundary model considering factors such as load growth and technological progress. Then, a source-storage planning model and a power grid growth model based on complex network theory are constructed to obtain the planned installed capacity of source and storage, and to determine the locations of new load nodes and power generation nodes. Furthermore, a source-grid-storage coordinated planning model is constructed, using typical daily data as input to solve for the system mode at each stage. Finally, the deduction process is concluded by determining whether the evolution period has been reached.

[0006] To achieve the above objectives, this invention provides the following technical solution: a system mode deduction method based on the full life cycle characteristics of energy sources and energy storage and grid growth, comprising the following steps: Obtain the initial data of the system's source, grid, load, and storage, and construct an evolutionary boundary model based on the full life cycle characteristics of the source and storage, technological advancements, and load growth trends, and calculate the boundary conditions.

[0007] To avoid direct coupling with grid growth and reduce evolutionary complexity, a source-storage planning model is constructed that minimizes the annualized construction investment, operation, and decommissioning costs, ignoring grid constraints.

[0008] Based on power grid design specifications, a power grid growth model is constructed to simulate the construction of nodes and lines in the power grid.

[0009] Based on the results of source-storage planning and power grid growth, a source-grid-storage collaborative planning model is established with the goal of optimizing the comprehensive annualized construction, operation, maintenance and decommissioning costs. Typical daily data is used as input to solve the system in the solution stage.

[0010] Determine whether the evolutionary stage has been reached and whether the deduction process has ended. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A flowchart illustrating the system deduction method based on the full life cycle characteristics of energy sources and energy storage and grid growth; Figure 2 This is a schematic diagram of the power grid growth process; Figure 3 A schematic diagram illustrating the evolution of wind, solar, thermal, and energy storage capacity; Detailed Implementation

[0013] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are... The embodiments are merely some, not all, of the embodiments of the present invention. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use. Based on the present invention All other embodiments obtained by those skilled in the art without inventive effort are examples of embodiments. The scope of protection by Benming.

[0014] The data required for the design of this invention are as follows: System source-grid-load-storage initial data acquisition: Data such as grid topology, node voltage levels, power generation of various power sources at different nodes, commissioning time and service life, energy storage capacity at different nodes, commissioning time and service life, and load level are obtained through the grid energy management system and asset and equipment management system. Among them, the load data includes the node load peak throughout the year and the regional total load in historical years. The remaining data are statistically analyzed with the last day of the year as the time section. Data such as population and GDP of each region, phased thermal power ratio restrictions, and wind and solar resource distribution are obtained through academic reports.

[0015] (1) Constructing an evolutionary boundary condition model The lifecycle characteristics of source and storage encompass the entire process from construction and operation to decommissioning, involving dynamic evolution patterns across multiple dimensions such as technical performance, economic efficiency, and environmental impact. Based on the lifecycle characteristics of source and storage, an evolutionary boundary model for this invention is constructed.

[0016] The constructed evolution boundary model specifically includes: a load level model, a carbon emission limit model, a decay model of available capacity during the operation and maintenance phases of various source and storage units, and a cost reduction function model for the construction, maintenance, and decommissioning of various source and storage units.

[0017] 1) Utilizing the correlation between population and economic data and load levels, a load level model is constructed based on the obtained historical regional total load regression analysis, which can be expressed as:

[0018] In the formula: L t For the first t Annual load level; β 0 represents the intercept term, which is the load magnitude when the influencing factor is 0; β 1 and β 2 is the regression coefficient, obtained by fitting historical data; D t This is GDP data, in 100 million yuan. P t This is population data, in units of ten thousand people.

[0019] 2) Based on advanced carbon sequestration technologies, a carbon emission limit model can be constructed, which can be expressed as:

[0020]

[0021] In the formula: E t for t Total carbon emission limits for each phase; e 0 represents the initial carbon emission factor; η t fort The stage carbon sequestration efficiency depends on the efficiency of progress over each year. α t Carbon sequestration efficiency and upper limit of carbon sequestration technology η max The minimum value in the range is given. The carbon sequestration efficiency and the technology upper limit are calculated based on the ATB (Advanced Technology Baseline) model constructed by the U.S. Renewable Energy Laboratory. All technology progress coefficients in this invention can be calculated using this model, and therefore will not be elaborated upon further.

[0022] 3) Based on the full lifecycle characteristics of various source and storage systems, a decay model for available capacity during the operation and maintenance phase of source and storage systems is constructed, which can be expressed as:

[0023] In the formula: C t for t stage i Available storage capacity of class source; C 0,i for i The initial capacity of the class source storage; for i Class source storage t The capacity decay rate of the stage; δ j,i for i Class source storage j Capacity recovery rate after maintenance.

[0024] 4) Based on various technological progress coefficients, a cost decay model for the construction, maintenance, and decommissioning of various source and storage units is constructed, which can be expressed as:

[0025] The construction, maintenance, and decommissioning costs of various source and storage units can all be calculated using this general model, where: T t,i for i Class source storage t The unit construction, maintenance, or decommissioning costs at each stage; T 0,i for i The estimated unit construction, maintenance, or decommissioning costs of similar storage facilities; γ t,i for i Class source in t Progress coefficient of a phase in construction, maintenance, or decommissioning; Q t for t Accumulated construction, maintenance, or decommissioning costs in stages; when t= When 1, the exponential term of the model is ignored.

[0026] Using the above methods, the embodiments of the present invention can calculate the evolutionary boundary conditions of each stage according to a set time interval, taking into account factors such as long-term planning and model solution scale, and plan optimally with a five-year time interval.

[0027] (2) Constructing a source-storage planning model To avoid direct coupling with grid growth and reduce evolutionary complexity, the source-storage planning model ignores grid constraints. Its objective function is to minimize the annualized construction investment, operation, and decommissioning costs, and its calculation formula is as follows:

[0028] In the formula: for t Phase 1 i Unit construction investment for this type of equipment; for t Phase 1 i The new capacity of this type of equipment to be constructed; for t Phase 1 i Maintenance investment per unit capacity of this type of equipment; for t Phase 1 i The total capacity of the equipment requiring maintenance; for t Phase 1 i Unit investment in the decommissioning of certain types of equipment; for t Phase 1 i The total capacity of the equipment to be decommissioned; TPU, WTU, PGS and ESU represent thermal power units, wind power units, photovoltaic units and energy storage units, respectively.

[0029] The construction, maintenance, and decommissioning investments for each equipment unit are based on the evolutionary boundary condition model constructed above; the remaining data can be obtained from the initial data of the system's source, grid, load, and storage.

[0030] Considering the upper limit of developable capacity, carbon emission quota constraints, and source-load balance constraints, the calculation formula is as follows:

[0031]

[0032]

[0033] In the formula: This represents the upper limit of the exploitable capacity. e t for t Carbon emission factor per unit of thermal power plant at different stages; E t fort Phased carbon emission limits; for t The magnitude of the load level at each stage.

[0034] By using MATLAB software to call the CPLEX solver, the solution can be obtained. t The planned capacity for energy sources and storage at this stage, considering that carbon emission constraints will only accelerate the retirement of thermal power plants, is essentially the total planned capacity for wind power, photovoltaic power, and energy storage.

[0035] (3) Construct a power grid growth and evolution model The increase in power system load and power generation capacity drives the power grid to grow by adding nodes, exhibiting a regular growth process with a certain degree of randomness. Therefore, this invention designs the construction of the power grid growth evolution model into the following steps: determining whether a node is growing, selecting the location of new nodes, and connecting new nodes. Solving the power grid growth model requires obtaining the initial power grid topology of the power system to be planned.

[0036] 1) Determine if a node is growing According to power grid design specifications, when the power supply capacity of a substation is insufficient, a new substation should be built to support the power supply capacity. Therefore, for power nodes, the remaining grid-connected capacity of the system's power nodes at a certain stage should be calculated. C re,t If the newly added power capacity is insufficient compared to the current capacity, additional power nodes are required. The formula for calculating the remaining grid-connected capacity of system power nodes is as follows:

[0037] In the formula: for t Phase 1 j Maximum capacity allowed per power node; for t Phase 1 j One power node has been connected to the capacity. For the first j Reserved capacity for each power node; k This represents the number of power supply nodes. It can be calculated from the initial capacity state of the system, using the following formula:

[0038] For load nodes, the load increase for each node in this phase is allocated according to the proportion of each node's load to the total load in the initial system. The calculation formula is as follows:

[0039] In the formula: Indicates the first j Initial load of each node;L t express t The total system load for a given period is obtained from the load level model designed and constructed in this invention; L 0 represents the initial total load of the system.

[0040] Each load node is checked to see if its downstream load exceeds the substation's power supply capacity. If it does, it indicates that the node needs to be split into a new load node, and the location and capacity of the new node are determined.

[0041] 2) Location selection for new nodes Complex networks are a mathematical framework for describing various types of networks in the real world (such as social networks, power grids, and biological networks). Their core characteristics lie in their irregular topologies and dynamic evolutionary behavior. Their community structure suggests that new nodes are more likely to appear near existing nodes and, more likely, in candidate positions with optimal characteristic parameters. For power generation nodes, new power sources are considered only for wind and solar power construction.

[0042] Clustering of power grid topology nodes to delineate power resource regions: First, nodes with relatively far geographical distances are selected as initial cluster centers; then, the exploitable wind and solar power capacity of each node is used as a node characteristic indicator; finally, nodes are aggregated using Euclidean distance to obtain the power resource region delineation results. The Euclidean distance calculation method is as follows:

[0043] In the formula: Represents cluster center points i Developable wind power capacity; Represents a node j Developable wind power capacity; Represents cluster center points i The exploitable photovoltaic capacity; Represents a node i The exploitable photovoltaic capacity.

[0044] The initial cluster centers and thresholds are set relatively randomly to reflect the uncertainties inherent in power grid growth. The characteristic parameters for newly added nodes are defined as the difference between the exploitable wind and solar power capacity and the existing capacity within the region. ζ v , m The calculation formula is as follows:

[0045] In the formula: For the region m The exploitable wind and solar confidence capacity can be obtained by multiplying the unit exploitable wind and solar confidence capacity by the area of ​​the region; For the regionm The first j The power nodes already have wind and solar capacity.

[0046] Select ζ v , m The largest area is designated as the geometric center point of the planned area, which will be used as the new power node for this growth.

[0047] For load nodes whose load exceeds the substation's power supply capacity, the transferable load range is determined with the node to be split as the center. This transferable load range is determined by the voltage drop tolerance radius, which refers to the grid topology radius where, after load transfer, the voltage drop between the new node and the overcapacity node does not exceed ±5%. The calculation formula is as follows.

[0048] Assuming a new node j Connect to the node to be split via a virtual line. i Connect them, and we get:

[0049] In the formula: R ij and X ij These represent the resistance and reactance of the virtual circuit, respectively. P ij + jQ ij To transfer the load; V i This represents the voltage of the node to be split.

[0050] The maximum impedance of the virtual line can be calculated using the above formula. Z max And based on the impedance per unit length of the line z Inverse topological radius d ij , which is the voltage drop tolerance radius, is calculated using the following formula:

[0051] With the node to be split as the center, the range formed by the voltage drop tolerance radius is the range within which the power supply load can be transferred.

[0052] Several nodes are randomly selected as candidate points within the load transferable range. The difference between the total load within the power supply radius of the candidate points and the existing power supply capacity is calculated to form a power supply deficit. The power supply radius is the topological distance between the node to be split and the candidate points. Power supply deficit selection. ζ d , i The maximum value, as the newly added load node, is calculated using the following formula:

[0053] In the formula: P d all,i New nodes to be selected i The total load within the power supply area can be obtained by multiplying the unit area by the area of ​​the power supply area; Pv j In the region m The first j Each load node already has a load.

[0054] 3) Adding new nodes The small-world property is another characteristic of complex networks, referring to the network's simultaneous existence of a small average distance and a large clustering coefficient. A small average distance indicates short energy transmission distances, facilitating rapid information propagation, while a high clustering coefficient suggests that nodes tend to form tightly connected groups. This theory was initially applied to social networks, but the small-world property has been found in many complex systems, and numerous experimental studies in research papers have demonstrated that power systems exhibit small-world network characteristics.

[0055] Based on the characteristics of small networks, newly added nodes are more likely to connect to nodes that are closer in distance and have a higher degree. Node degree represents the number of nodes connected to a given node and is a key indicator of whether a node is critical in the topology. In addition, newly added load nodes should connect to power supply nodes with surplus power whenever possible, and newly added power supply nodes should connect to nodes with power deficits whenever possible to achieve local balancing.

[0056] Therefore, construct node optimization factors. φ p Characterizes the connection point of a newly added node. p The relative probability is calculated using the following formula:

[0057] Where: Ω c Range of potential access points; α , β , χ These are the weighting coefficients for characterization, distance, and the degree of influence of power, respectively. k p Access point p The degree is calculated from the edges directly connected to it, that is, the number of nodes connected to that node; l p Access point p Euclidean distance between the new node and the node; G p Access point p The net power is calculated from the difference between the total confidence capacity of the power supply at the access point and the maximum load. When the newly added node is a load node, only positive numbers are counted; when the newly added node is a power supply node, only negative numbers are counted.

[0058] The weighting coefficients are determined by the subjective weighting method, and their calculation method is as follows: The importance of each indicator in the evaluation system was determined by soliciting expert opinions, and a fuzzy evaluation matrix R was established with a numerical scale of 1-9, where the higher the value, the greater the importance.

[0059] A judgment matrix is ​​established based on the importance ratios of various indicators. M And calculate the largest eigenvector. λ max :

[0060]

[0061] In the formula: m ij For the first i The first indicator and the first j The importance ratio of each indicator; ρ for λ max The corresponding feature vector.

[0062] Due to the limitations of expert subjective experience, it is necessary to verify the consistency among matrix elements, using a consistency ratio. C R The measurement and calculation formula is as follows:

[0063]

[0064] In the formula: n To determine the matrix M The order of; C I As a consistency indicator, R I This is a random consistency index, obtained through a table lookup. If the calculation result... C R <0.1 indicates that the consistency standard has been met, matrix M If it is reasonable, otherwise the matrix needs to be modified. M The element values ​​are processed until the consistency requirement is met.

[0065] After passing the consistency check, the subjective weights can be calculated. j The subjective weight calculation formula for each indicator is as follows:

[0066] In the formula: For matrix M The Middle i The product of all elements in a row.

[0067] The optimization factor for each potential access point is calculated and normalized to form the relative access probability. The calculation formula is as follows:

[0068] In the formula, φ max This represents the maximum value of the preference factor for all nodes within the potential access point range; φ min It is the minimum value of the selection factor for all nodes within the potential access point range.

[0069] To reflect the uncertainties in the evolution, a roulette wheel algorithm is used to select the connection point for newly added nodes. The core idea of ​​the roulette wheel algorithm is to map the probability of each option onto a sector of the roulette wheel, and to determine the selected item by randomly rotating the roulette wheel. Based on this idea, a node random selection model is constructed: generating random numbers with values ​​in the range of 0-1. r Through analysis r Determine which node's sector area it belongs to, and then identify the connection point for the new node to connect to.

[0070] (4) Construct a source-grid-storage collaborative planning model After obtaining the phased source-storage planning and grid growth scheme, a source-grid-storage collaborative planning model is constructed to calculate the current phase's source-storage site selection and capacity determination, as well as the grid upgrade scheme.

[0071] The objective function of the model is to minimize the annualized comprehensive cost of construction, operation, maintenance, and decommissioning, and its calculation formula is as follows:

[0072] C s i,j,t for T Time period nodes j The Middle i The increased capacity of this type of equipment; C p i,j,t for T Time period nodes j The Middle i The capacity required for maintenance of this type of equipment; C r i,j,t for T Time period nodes j The Middle i The capacity of this type of equipment that needs to be decommissioned; N This indicates the total number of nodes.

[0073] Compared to the source-storage planning model, it takes into account more factors. F line and F o This refers to the cost of line construction and the cost of system operation. The formula for calculating line construction cost is as follows:

[0074] In the formula, c l i,t for T Phase 1 i Construction cost per unit length of the line; Le i,t express t Phase 1 i The construction length of the line.

[0075] In the established objective function, the system construction, maintenance, and decommissioning costs are expressed on an annual time scale. T The computational cost is measured in minutes. t The calculation involves 96 data points spaced 15 minutes apart throughout the day, primarily considering coal consumption and carbon emission costs for thermal power plants, as well as unit start-up and shutdown costs. The calculation formula is as follows:

[0076] In the formula: c caol , These are unit coal consumption cost and unit carbon emission cost, respectively; For thermal power units j The coal consumption coefficient; and thermal power units j exist t The on / off state variable at any given time; and thermal power units j The start-up and shutdown costs.

[0077] Consideration should be given to constraints such as node installed capacity, power balance constraints at each node, power injection constraints for the entire system, line transmission capacity not exceeding limits, power supply active power output constraints, and energy storage charging and discharging power and state of charge safety range constraints.

[0078] 1) The calculation formula for node installed capacity constraints is as follows:

[0079]

[0080] In the formula: C s i,j,T,max for T Time period nodes j No. i The maximum capacity that can be added to a certain type of equipment is equal to the planned capacity plus the decommissioned capacity.

[0081] 2) The calculation formulas for node power balance constraints and total system injected power constraints are as follows:

[0082]

[0083]

[0084] In the formula: P k i,j,t For nodes i The j The device is t Efforts made at all times; P in i, t For nodes i exist t Injection power at any given moment; P d i, t For nodes i exist t Active power of the load at any given time; P line i, t For the line l exist t Active power at any given time; Ω to and Ω from Each is based on a node i A set of routes with start and end nodes.

[0085] 3) The calculation formula for ensuring that the line transmission capacity does not exceed the limit constraint is as follows:

[0086] In the formula: P line i, max For the line l Maximum transmission capacity.

[0087] The constraint relationship between nodal injected power and line power flow is constructed using DC power flow relationships, and the calculation formula is as follows:

[0088] In the formula: For each line in t The active power matrix transmitted at any given time. L Total number of lines; The system power flow distribution factor matrix, I This represents the total number of nodes; For each node in t The active power matrix injected at any given time.

[0089] 4) Power output constraints need to be divided into constraints for thermal power units and constraints for wind and solar power units. The calculation formulas are as follows:

[0090]

[0091]

[0092] In the formula: and thermal power units j The minimum and maximum output; and thermal power units j The maximum upward and downward power changes per unit time; and Wind and solar turbines j exist t Maximum and minimum output at any given time.

[0093] 5) Constraints on energy storage charging and discharging power and the safe range of state of charge are calculated using the following formulas:

[0094]

[0095] In the formula: and Energy storage units j Maximum discharge power and maximum charging power, with the discharge power set to a negative value; and Energy storage units j Minimum state of charge and maximum state of charge.

[0096] The model designed and constructed in this invention solves a multi-timescale collaborative optimization planning problem. To reduce the computational complexity of the model, this invention uses the natural month as a dividing point, selecting a typical day each month to convert it into a monthly load operation scenario, thereby improving the efficiency of multi-year decision-making computation. The typical day load scenario is extracted based on historical load data analysis using the k-means clustering algorithm. The specific steps are as follows: T The system node load time-series data for each phase is scaled up proportionally according to the system load evolution variables to generate a load dataset, which is then divided according to calendar months. Therefore, the required dataset dimension is... N The dimensionality reduction factor is still quite large (×12×96). Therefore, we first use the PCA (Principal Component Analysis) algorithm to reduce the dimensionality of the dataset. The basic idea of ​​PCA is to reduce the dimensionality of the data by maximizing the variance of the data, that is, to find a new set of dimensions that maximizes the variance, thereby transforming the original data into low-dimensional data. The calculation method is as follows: For the dataset of this invention X It can be represented as:

[0097] In the formula: n This invention represents the number of datasets. N ×12.

[0098] For high-dimensional data, covariance is used as a constraint, as it represents the correlation between two features. To ensure that two features represent as much of the original information as possible, they should not be linearly correlated, because correlation implies that the two features are not completely independent and inevitably contain duplicate feature information.

[0099] Centering the original feature set: This involves removing the mean, i.e., calculating the average of each feature, and then subtracting its own mean from each feature for all samples.

[0100] In the formula, x i For centralized data; x ’ i For the original number i Data points.

[0101] In mathematics, variance is used to describe the dispersion of projected values ​​after projection. The variance of a feature can be seen as the average of the differences between each feature and its mean. Since the data has already been centered, the variance can be directly expressed as the sum of the squares of each feature divided by the number of features.

[0102] Covariance represents the correlation between two features. A centered feature dataset is converted into a covariance matrix, and the formula for calculating the covariance matrix is:

[0103] In the formula, ω is the unit vector of the projection direction.

[0104] Each eigenvector of the covariance matrix is ​​a projection surface, and the eigenvalue corresponding to each eigenvector is the variance of the original features projected onto this projection surface.

[0105] Calculate the eigenvalues ​​and eigenvectors of the covariance matrix, and arrange the eigenvectors in descending order of their corresponding eigenvalues. Select the first few eigenvectors. n The eigenvectors form a new feature space. Then, the original data is projected onto the new feature space, and the vectors are... x Projected to n dimensional subspace W The formula for calculating coordinates is:

[0106] By decomposing the feature covariance matrix to obtain the principal components (i.e., eigenvectors) and corresponding weight values ​​(i.e., eigenvalues) of the data, the data can be transformed from... n Dimensional reduction to dimensionality dDimension, the calculation formula is:

[0107] In the formula, x ’’ i This is the data after dimensionality reduction.

[0108] After dimensionality reduction N ×12× d Dimension Dataset k -means clustering, k The -means algorithm is a typical partition-based clustering algorithm. Its principle is simple: for a given sample set, Euclidean distance is used as a metric to measure the similarity between data objects. Similarity is inversely proportional to the distance between data objects; the greater the similarity, the smaller the distance. Its calculation method is as follows: In spatial domains, the Euclidean similarity distance between any sample object and the cluster center is calculated using the following formula:

[0109] In the formula: X For data objects; C i For the first i Each cluster center; m For the dimensions of data objects; X j -C ij )for X and C i The j The difference between the attributes.

[0110] k The objective function to be achieved by the -means algorithm is:

[0111] In the formula: c 1, c 2, …, c k It is not k The central point of each cluster; C ( X i )express X i This point is the center point of the cluster to which it belongs; d 2 This indicates calculating the square.

[0112] For each category c kRecalculate its cluster centers using the following formula:

[0113] k -The means algorithm requires selection k The values ​​and initial random centroids are used, and the results of each clustering will not be exactly the same, so it is necessary to determine the selected values. k Is the value reasonable?

[0114] Therefore, the SSE (Sum of Squared Error) formula is used to judge the performance of clustering, which is the sum of the squared distances from each point to its cluster centroid. A smaller SSE value indicates that the data points are closer to their centroids, and the better the clustering effect. Because the error is squared, more emphasis is placed on points far from the center. The goal of clustering is to improve the quality of clusters while keeping the number of clusters constant. The calculation formula is as follows:

[0115] In the formula: the size of SSE indicates the quality of the clustering result; k The number of clusters.

[0116] In practical applications, SSE (Segregational Segmentation) is often used in conjunction with silhouette coefficients to evaluate the effectiveness of clustering models. Each sample has a silhouette coefficient, which consists of two parts: a and b. a represents the average distance between the sample and other samples in the same cluster (i.e., quantifying cohesion), and b represents the average distance between the sample and all samples in the nearest cluster (i.e., quantifying separation). The silhouette coefficient for each sample is defined as follows:

[0117] In the formula, s The value range is from -1 to 1.

[0118] To determine the number of cluster centers, the elbow curve method, a common cluster analysis technique, is used. This involves setting several cluster numbers, calculating the sum of the inter-cluster distances, and plotting an elbow curve. Observing the elbow curve, the number of clusters at the bend is considered a suitable number of cluster centers.

[0119] After clustering, typical low-dimensional load scenarios for each month can be determined according to different needs. Then, an inverse transformation of these typical low-dimensional load scenarios is performed to obtain load scenarios for 96 data points. The inverse transformation remaps the dimensionality-reduced data back to the original feature space to obtain approximately reconstructed data. The calculation formula is as follows:

[0120] In the formula: D TTranspose of the covariance matrix By inputting typical load scenarios for each month into the model and converting them into equivalent annual operating data, the time scale can be unified. The final solution of the model can then be completed by calling the CPLEX solver using MATLAB software.

[0121] (5) Determine whether the evolutionary stage has been reached. The deduction method designed in this invention takes 5 years as a stage. By setting the evolution period to reach the final demand, it can be determined whether the evolution has reached the evolution period. If the evolution period has been reached, the result is output.

[0122] If the evolution period has not been reached, the evolution will continue. For systems that need to continue evolving, the system mode data of the updated stage is used as the initial data, and then combined with the load evolution results of the next stage and the source-grid-storage planning data are re-input into the source-grid-storage collaborative planning model.

[0123] Repeat the above evolutionary process until the evolutionary stage is reached.

Claims

1. A system approach for deriving based on source storage life cycle characteristics and grid growth, characterized in that, The method comprises the following steps: a. Constructing an evolution boundary model based on the full life cycle characteristics of the source and storage, technological progress and load growth trend, calculating boundary conditions, b. Constructing a source and storage planning model and a power grid growth model based on complex network theory, obtaining stage source and storage planning, and determining the location of new load nodes and power supply nodes, c. Establishing a source-grid-storage collaborative planning model, using typical day data as input to solve the stage system mode to determine whether the evolution period is reached and whether the deduction process is ended.

2. The method as claimed in claim 1, characterized in that, The step a comprises: a1. Obtain system source-grid-load-storage initial data, including but not limited to power grid topology, different node types of power generation capacity, operation time and service life, storage capacity, operation time and service life, load level, etc.; obtain system each stage vision design data, including but not limited to population and economic data, thermal power proportion limit source and wind and light resource distribution, etc.; divide the source and storage full life cycle into development and construction stage, operation and maintenance stage and decommissioning and recycling stage, and construct an evolution boundary model. Specifically, it includes: load level model, carbon emission limit model, decay model of available capacity of various types of source and storage in operation and maintenance stage, and unit construction, maintenance and decommissioning cost reduction function model of various types of source and storage, a2. Calculate boundary conditions, specifically: based on historical data, use the correlation regression of population and economic data and load level to calculate the load growth curve; based on the carbon fixation technological progress coefficient, calculate the carbon emission limit curve; based on the efficiency decay rate, use the initial capacity and operation time data to calculate the available capacity decay curve of various types of source and storage; based on the technological progress coefficient of each type, calculate the unit construction, maintenance and decommissioning cost decay curve of various types of source and storage.

3. The method as claimed in claim 1, wherein, The step b comprises: b1. Construct a source and storage planning model with the optimal annual construction, operation and decommissioning comprehensive cost as the target, the calculation solving time scale is year, and the stage source and storage planning capacity is obtained; traverse the power grid each node under the grid load and the power supply capacity shortage, judge whether the load node needs to grow; calculate the difference between the remaining grid-connected capacity and the power supply demand in this stage, and judge whether the power supply node needs to grow, b2. Determine the possible range of new load and power supply nodes based on complex network theory. Calculate the power supply capacity shortage to determine the new load node; calculate the confidence capacity difference to determine the new power supply node; construct a comprehensive node optimization factor to calculate the relative probability of the new node access location, and select the location with the maximum probability as the new node access location.

4. The method as claimed in claim 1, wherein, The step c comprises: c1. Establish a source-grid-storage collaborative planning model, set the objective function as the optimal annual construction, operation, maintenance and decommissioning comprehensive cost, and also consider the line construction cost in the power grid, the calculation solving time scale is minute level; set the constraint condition, specifically, the node power balance constraint and the whole system injection power constraint, the line transmission capacity constraint, the power supply active power output range constraint, the storage charging and discharging power and state of charge safety range constraint, etc., c2. Use PCA dimensionality reduction to process historical load data set, and use k-means clustering to obtain typical day data as input to generate stage system mode, c3 judging whether the evolution time reaches the evolution period, if reaching the evolution period, outputting the result, if not reaching the evolution period, continuing evolution; for the system needing to continue evolution, updating the stage system mode data, inputting the source-grid-storage collaborative planning model again in combination with the next stage load evolution result and the source-storage planning data; repeating the evolution process until the evolution reaches the evolution period.

Images