CN-122020604-A - Photovoltaic power station maximum power generation calculation method integrating topography and meteorological features

CN122020604ACN 122020604 ACN122020604 ACN 122020604ACN-122020604-A

Abstract

A calculation method of maximum power generation of a photovoltaic power station integrating topography and meteorological features belongs to the technical field of power of photovoltaic power stations and comprises the following steps of constructing a meteorological-photovoltaic grid node system, based on a spatial association field fused by a physical element mechanism, used for quantifying comprehensive association strength between any two grid nodes, outputting a topological embedded vector by the meteorological-photovoltaic grid node system, carrying out multi-scale space-time feature extraction on the topological embedded vector of the grid nodes and weather forecast information, dividing a photovoltaic power station array into multi-level subareas and constructing a multi-grid system, constructing a hierarchical dynamic graph structure, deploying a space-time graph neural network on the hierarchical dynamic graph structure, and outputting a maximum power generation prediction value of the photovoltaic power station array by the grid nodes. Through the improvement of the algorithm, the maximum power generation capacity of the photovoltaic unit is modulated, and a solid data and model foundation is laid for safe, efficient and economic operation of the power system.

Inventors

ZHEN ZHAO
TIAN YUXUAN
TONG ZIHAO
WANG FEI

Assignees

华北电力大学

Dates

Publication Date: 20260512
Application Date: 20251231

Claims (10)

1. The method for calculating the maximum power generation of the photovoltaic power station by integrating the topographic and meteorological features is characterized by comprising the following steps of: Step S10, collecting weather forecast information and static terrain information of a photovoltaic power station array serving as grid nodes, and generating power, longitude and latitude position information and installed capacity of the photovoltaic power station array in a photovoltaic base, constructing a weather-photovoltaic grid node system, and mapping discrete photovoltaic power station arrays and weather forecast information to a uniform geographic space grid; Step S20, based on a spatial association field fused by a physical element mechanism, the physical element is jointly determined by geographic distance, historical radiation correlation, dominant wind direction and azimuth angle relation and is used for quantifying the comprehensive association strength between any two grid nodes; Step S30, adopting a CNN terrain-aware cavity convolutional network model and a structure-aware converter double-stage coding architecture to enable a meteorological-photovoltaic grid node system to output a topological embedded vector; Step S40, performing multi-scale space-time feature extraction on topology embedded vectors and weather forecast information of grid nodes, and step S50, dividing a photovoltaic power station array into multi-level subregions and constructing a multi-grid system; Step S60, constructing a hierarchical dynamic graph structure; and step S70, a space-time diagram neural network is deployed on the hierarchical dynamic diagram structure, multi-source weather forecast data of a future period is taken as input, forward propagation is carried out through the hierarchical dynamic diagram model, and finally the grid node outputs the maximum power prediction value of the photovoltaic power station array.
2. The method according to claim 1, wherein the weather forecast information in the step S10 includes total surface radiant flux, cloud cover, air temperature and wind speed; The static terrain information comprises altitude, gradient and slope direction; The formula of mapping the discrete photovoltaic sites and weather forecast information to the unified geospatial grid is S ij (t): In the above-mentioned method, the step of, As a characteristic of the weather-state, Is the feature of the terrain state, T is the moment, G i is the total radiant flux of the earth surface, E i is the cloud cover, the value range is 0-100%, T i is the air temperature, Is the wind speed; For altitude, α i and β i are slope and slope direction, respectively.
3. The method for calculating the maximum power that can be generated by a photovoltaic power plant according to claim 1, wherein said step S20 comprises the steps of: Step S201, collecting geographic distances d nm between grid nodes, a Pearson correlation coefficient rho nm of a historical total radiation sequence between grid nodes, a dominant wind direction angle psi n of a meteorological system on the ground near the grid nodes, an azimuth angle theta nm between grid nodes, and a typical propagation scale L of cloud clusters or meteorological disturbance; Step S202, calculating the weather association strength between each grid node, In the formula, For adjusting the weights, λ 1 、λ 2 、λ 3 is an adjustable weight coefficient.
4. The method for calculating the maximum power that can be generated by a photovoltaic power plant according to claim 1, wherein said step S30 comprises the steps of: step S301, adopting CNN topography aware cavity convolution network to perform local feature aggregation on input tensors, capturing a multi-scale topography-meteorological coupling mode, Where e (l) represents the convolution operation with expansion ratio e (l) , output Contains rich local context information, the terrain shading effect and weather cooperative change characteristics of the adjacent area are preliminarily fused, the ReLU () is an activation function, Is a parameter operator, which is a parameter operator, Is a bias operator, and S represents a set of static terrain information and weather forecast information obtained in the step S10; step S302, introducing a structure-aware transducer dual-stage coding architecture, injecting the spatial correlation field as a structure bias into an attention mechanism, modeling a long-range dependency under physical constraint, Where Attention m () represents an Attention calculation operator, MHSA () represents a multi-head Attention, head n represents an nth Attention head, Q is a query vector, K is a key vector, V is a value vector, For the original attention score, B ε R N×N is the bias matrix, defined as: Wherein B nm represents an element of the nth row and the mth column, Representing the weighting parameters; Step S303, each grid node position outputs a topology embedded vector fusing local detail and global structure context Where MLP () represents a full connection layer neural network.
5. The method according to claim 1, wherein the multi-scale spatio-temporal feature extraction in step S40 includes feature extraction on a time scale and a space scale.
6. The method for calculating the maximum power that can be generated by a photovoltaic power plant according to claim 1, wherein said step S50 comprises the steps of: Step S501, performing normalization processing on the actual output of the photovoltaic power plant array, In the formula, Representing normalized power output, P i t being the photovoltaic output of the actual ith photovoltaic array at time t, P i,capacity being the installed capacity of the photovoltaic array; Taking the longitude and latitude position information of the photovoltaic power station array and the average value, the maximum value and the standard deviation of the photovoltaic power annual output sequence as characteristic input features (P i ), constructing a sample matrix P, feature(p i )＝[p i,longitude ,p i,latitude ,mean(p i,annual ),max(p i,annual ),std(p i,annual )], Wherein p i,annual represents a normalized photovoltaic power sequence of one year, and p i,longitude and p i,latitude represent longitude and latitude of an ith photovoltaic array respectively; Step S502, constructing a graph structure based on the similarity matrix, Constructing an undirected graph based on graph nodes; if there is a connection between two graph nodes, the weight w ij of the edge is not zero, otherwise it is zero, the formula of the element w ij of the adjacency matrix is, Dividing the graph structure into k sub-graphs by means of Ratio-cut criterion, namely dividing the photovoltaic power station array into sub-areas, In the formula, k represents the number of subgraphs, Representing the complement of C x , Represents to be And Cutting; step S503, converting and solving the spectral clustering optimization problem, Λ=D-l, Wherein the degree matrix Λ is a Laplace matrix, H is a matrix formed by eigenvectors corresponding to the first k minimum non-zero eigenvalues of the Laplace matrix L, D is a degree matrix, and the diagonal element D ii of the degree matrix D is the sum of weights of all connected edges of the grid node i; By means of the K-means clustering algorithm, features (p i ) are divided into corresponding sub-regions, Step S504, aggregation and hierarchy construction of secondary subregions; on the basis of the divided subareas, further evaluating the output cooperativity, the spatial adjacency and the weather consistency among the subareas, and carrying out secondary clustering on the subareas to form a middle-level area; The process can be recursively carried out, and a multi-layer grid system of 'photovoltaic array → subarea → intermediate-level area → base whole' is finally constructed.
7. The method according to claim 1, wherein the hierarchical dynamic graph structure in the step S60 includes graph nodes and edge weights; The graph nodes are divided sub-areas of each layer, and comprise a bottom-layer node which is taken as a photovoltaic power station array, a middle-layer node which is taken as a primary sub-area, and a high-layer node which is taken as a middle-level area or a base as a whole; the edge weight is the weather association strength between each grid node.
8. The method for calculating the maximum power that can be generated by a photovoltaic power plant according to claim 1, wherein said step S70 comprises the steps of: Step S701, introducing an LSTM long-term memory network to explore the time sequence change rule of the sub-region output mode, wherein the sub-region output mode is as follows h t ＝o t ⊙tanh(s t ), Wherein A t ,J t ,o t ,s t ,h t represents the input gate, forget gate, output gate, memory gate and hidden gate of the module respectively, { Θ x ,Θ h , b } is the specific model parameter of the gate, σ (·) is the activation function, X t represents the module output, As an input of the spatio-temporal correlation information X t t=1, 2,3, -, represents a matrix multiplier, Θ W is a corresponding matrix multiplier; step S702, the matrix containing the time sequence information Input to a space-time diagram neural network and based on space-time information of the current time step Mining to obtain space-time associated information X sub1 with dimension of 48 multiplied by 1; X sub1 ＝X⊙Θ sub1 , Where X represents the output of the convolutional layer, Is the input to the convolutional layer, Θ is the convolutional kernel parameter to be trained, In the form of a node degree matrix, For the adjacency matrix at time t, I is an identity matrix, and Θ sub1 is a corresponding sub-region multiplier; Step S703, defining the extracted space-time correlation information of the different sub-regions as X sub1 ,X sub2 ,X sub3 . Through global space-time correlation modeling and fusion of the characteristics of all nodes through a full connection layer, X region is obtained, the maximum power generation predicted value of the photovoltaic power station can be obtained, Θ p is a corresponding global space-time correlation multiplier, p t：t+15 ＝X region ⊙Θ p 。
9. A non-transitory computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the steps of the method of any of claims 1-8.
10. A computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, implements the steps of the method according to any of claims 1-8.

Description

Photovoltaic power station maximum power generation calculation method integrating topography and meteorological features Technical Field The invention belongs to the technical field of photovoltaic power station power, and particularly relates to a calculation method of maximum power generation of a photovoltaic power station integrating topography and meteorological features. Background Currently, china is accelerating to construct a novel power system mainly based on new energy, and the construction of a large-scale wind-solar base is comprehensively accelerated. Under the drive of a 'two-carbon' strategic target, western regions such as Qinghai, ningxia, gansu, inner Mongolia and the like are planned and built into a plurality of millions of kilowatt-level light Fu Daji lands. The bases are wide, the occupied area can reach hundreds or even thousands of square kilometers, the internal topography is complex and various, and the bases cover various geographic units such as highland, mountain areas, hills, deserts and the like. However, it is this combination of very large scale and complex terrain that presents unprecedented challenges to the fine-scale operation management of photovoltaic power plants. Different areas in the base show obvious meteorological state space heterogeneity due to the difference of topography factors such as altitude, gradient, slope direction and the like and the influence of meteorological processes such as local circulation, cloud cluster movement and the like. For example, at the same time, one side of the base may be sunlit and the other side covered by clouds, the ridges being irradiated sufficiently while the valleys are under shadows or fog cages. This "ten different day" microclimate pattern directly results in a large difference in the actual maximum power that can be generated (i.e., theoretical upper power generation limit) for each sub-region photovoltaic power plant. Most of the existing photovoltaic power prediction and evaluation methods are based on point-to-point modeling ideas, and data of single or few meteorological observation points are directly used for power estimation of the whole power station or an adjacent area. The method ignores the dynamic propagation rule of the meteorological field in space and the coupling effect of the meteorological field and complex terrain, and cannot accurately describe the fine meteorological-power mapping relation in the base. As a result, the assessment of the maximum power that can be generated by the light Fu Da base is often not rough, either overestimating the power generation potential of the area affected by the obscuration or clouds, or underestimating the output power of the area of superior illumination. The inaccurate evaluation not only affects the accurate grasp of the available power resources by the power grid dispatching department and reduces the new energy consumption level, but also restricts the efficiency of the electric power market transaction and even threatens the safe and stable operation of a large power grid. Therefore, development of a novel calculation method capable of deeply fusing the space-time correlation characteristics of the topographical features and the meteorological elements is needed. The method must be able to systematically characterize how meteorological disturbances propagate, evolve under complex terrain and ultimately modulate the maximum power generation capacity of the individual photovoltaic units. Only then can the maximum power generation panoramic sensing with high precision and high resolution be provided for the ultra-large scale photovoltaic base, and a solid data and model foundation is laid for the safe, efficient and economic operation of the novel power system. This patent is presented in this context, aims at solving the key bottleneck problem that prior art faced under the complex scene. Disclosure of Invention The invention aims to solve the technical problem of providing a calculation method of the maximum power generation of a photovoltaic power station, which is integrated with the topography and the meteorological features, and by improving an algorithm, the method can systematically characterize how meteorological disturbance propagates and evolves under complex topography, and finally modulates the maximum power generation capacity of each photovoltaic unit, so that the panoramic perception of the maximum power generation with high precision and high resolution can be provided for a super-large-scale photovoltaic base, and a solid data and model foundation is laid for the safe, efficient and economic operation of a novel power system. The technical scheme includes that the method comprises the steps of S10, collecting weather forecast information and static weather information of a photovoltaic power station array serving as grid nodes and generating power, longitude and latitude position information and installed capacity of the photovoltaic power st