A weather-based dynamic graph network photovoltaic power station group ultra-short-term prediction method

Abstract

The application discloses a meteorology-based dynamic graph network photovoltaic power station group ultra-short-term prediction method, which constructs a dynamic space-time graph network, regards photovoltaic power stations in a region as a complex network system connected by meteorological fluctuations, takes each power station as a node, and represents the propagation relationship of meteorological fluctuations between power stations through edges of the graph. The weight and time lag of the edge are dynamically adjusted according to real-time meteorological data, and a hybrid model combining a graph neural network and a recurrent neural network is used to accurately capture the dynamic interaction between power stations, the overall trend under stable meteorological conditions and the rapid fluctuation caused by sudden weather events. Compared with traditional methods, the application significantly improves the prediction accuracy, real-time response capability and physical interpretability, integrates dispersed photovoltaic power stations into a dynamically correlated system, provides an efficient and reliable solution for power station group power prediction, and enhances the operation stability of large-scale photovoltaic systems and the efficiency of power grid dispatching.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of new energy power generation technology and power system technology, and particularly relates to a dynamic graph network photovoltaic power station group ultra-short-term prediction method based on meteorology. BACKGROUND

[0002] Photovoltaic power generation technology, with its clean and renewable advantages, plays a crucial role in global energy structure transformation and climate change response. With continuous technological progress and cost reduction, photovoltaic power generation has become the core force of global energy sustainable development. In China, the rapid development of the photovoltaic industry not only promotes the green transformation of the energy structure, but also provides strong support for achieving the carbon peak and carbon neutralization goals, and sets an example of technological innovation and cost optimization in the global industry chain. However, with the expansion of large-scale centralized photovoltaic power stations, the intermittent, volatile and uncertain nature of their grid-connected power output poses a huge challenge to power grid operation and dispatching. Therefore, accurate power prediction for photovoltaic power station clusters, especially ultra-short-term (e.g. 15 minutes to 4 hours in the future) prediction, has become a key foundation for ensuring the efficient and stable operation of the power system.

[0003] Traditional power prediction methods mainly focus on statistical and shallow machine learning models, such as autoregressive moving average models and their variants. The focus of these models is on classical time series analysis. For example, some recent research has used time series models such as ARIMA to predict daily solar power generation, which assumes a linear relationship between past and present values. However, complex time series data in the real world are not always linear, which leads to suboptimal model prediction accuracy and data with strong limitations. Another approach is to extend a series of autoregressive models such as SARIMA by introducing external meteorological data such as solar irradiance to achieve short-term photovoltaic power generation prediction. The performance of this type of method usually depends on the accuracy of meteorological data, and once there is a large error in the meteorological data, the prediction result will also be directly affected.

[0004] Although these methods mine the seasonal and periodic variation behind photovoltaic power generation, they can be effective in predicting the stable power generation mode of a specific power station that is not affected. However, in actual application, regional photovoltaic power stations are usually located close to each other, and their power output is also affected by similar weather conditions. Therefore, another research approach also attempts to introduce modeling and analysis methods for power station clusters. For example, a time convolution network (TCN) method is used for regional distributed photovoltaic power prediction. This method can process historical data of multiple power stations and predict the power of other power stations in the region based on the data. However, although this method can process historical data of multiple power stations, the selection of representative power stations and their impact are still lacking in-depth discussion, and the correlation between power stations has not been fully analyzed.

[0005] From a physical point of view, meteorological events in a certain area, such as the movement of cloud clusters and the advection of wind fields, are continuous and interact with each other. The movement of a cloud cluster above a power station will inevitably affect the downstream or adjacent power stations within a certain time. This causal relationship caused by the transmission of meteorological events is physical, not just a simple statistical correlation. Due to its inherent design flaws, the traditional prediction model cannot capture this physical causal chain, resulting in a significant decrease in prediction accuracy when the weather changes suddenly or fluctuates dramatically, which is a fundamental flaw. In order to overcome the limitations of traditional methods, some studies have begun to use graph structures to process spatio-temporal data. These methods treat individual units within the divided area as nodes in the graph and use graph structures to capture the correlation between nodes. However, many existing graph neural network models still use static graph structures, i.e., constructing adjacency matrices based on pre-set topologies, fixed relationships, or historical data-level correlations. This static modeling approach, while having some reference value, is significantly different from the dynamic meteorological environment in which photovoltaic power stations actually operate. This static modeling approach has its own basis and reference value, but it is significantly different from the rapidly changing meteorological environment that photovoltaic power stations actually operate in. Meteorological conditions have the characteristic of evolving over time, and changes in wind direction and speed can cause the mutual influence relationship between power stations to change significantly. For example, between two photovoltaic power stations, the relationship between the "upstream power station" and the "downstream power station" is dynamically affected by the real-time wind direction. Assuming that the dominant wind direction is north, the power station located at the starting point of the wind direction can be defined as the upstream power station, and its meteorological fluctuations (such as cloud movement or irradiance changes) may affect the power output of the downstream power station located on the downwind side of the wind direction; when the wind direction changes to south, the roles of "upstream" and "downstream" power stations will also reverse, and the influence relationship will dynamically adjust. Static graphs are difficult to capture this dynamic change caused by the advection and disturbance transmission of meteorological conditions, making it difficult for the model to accurately reflect the real propagation path of meteorological fluctuations when dealing with sudden weather events, and the robustness and interpretability of the prediction results are also limited. Therefore, how to effectively represent and utilize the dynamic evolution of the correlation between power stations with changing meteorological conditions in the prediction model is still a problem that needs to be solved by the existing technology. SUMMARY

[0006] Therefore, the purpose of the present application is to overcome the shortcomings of the prior art and provide a meteorology-based dynamic graph network photovoltaic power station group ultra-short-term prediction method. This method captures the meteorological fluctuation transmission relationship between photovoltaic power stations in a region through a dynamic spatio-temporal graph network, combines a hybrid model of graph neural networks and recurrent neural networks, and realizes ultra-short-term power prediction for photovoltaic power station clusters, improving the accuracy, real-time performance, and physical interpretability of the prediction.

[0007] To achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0008] A meteorology-based dynamic graph network photovoltaic power station group ultra-short-term prediction method, comprising the following steps:

[0009] S1, obtaining the ultra-short-term time series data of a plurality of geographically dispersed photovoltaic power stations, the data comprising the grid-connected power data, irradiance and regional meteorological data of each power station, the regional meteorological data comprising wind speed, wind direction, temperature, humidity and air pressure, and performing missing value processing, abnormal value detection and data preprocessing on the data;

[0010] S2, defining the photovoltaic power stations as nodes in a graph, and establishing the relative static coupling relationship between the nodes based on geographical location similarity, power station distance relationship, historical power similarity or other correlation analysis; at the same time, using the meteorological data, calculating the directed edge weight between the power station nodes based on meteorological physical factors to dynamically construct a dynamic adjacency matrix reflecting the physical propagation effect of meteorological fluctuations, and finally combining the static coupling relationship and the dynamic adjacency matrix to construct a dynamic graph network G(t)=(V, E(t)), wherein V represents a set composed of N nodes, and E(t) represents the dynamic edge of the graph structure;

[0011] S3, constructing a hybrid model composed of a graph neural network and a recurrent neural network, inputting the dynamic graph network G(t) and the preprocessed time series data of each power station into the hybrid model for training until the preset training accuracy requirement is met, and completing model training;

[0012] S4, judging whether the power prediction result meets the preset prediction accuracy requirement; if not, returning to the model construction and training stage to retrain or optimize the hybrid model until the prediction accuracy meets the corresponding situation requirement;

[0013] S5, when the hybrid model meets the prediction performance accuracy threshold, using the latest real-time meteorological data and historical power data, combining the real-time updated dynamic graph network G(t), and performing power prediction in a specific time period in the future through the hybrid model meeting the accuracy requirement, the power prediction adopts a prediction mechanism, including rolling prediction, future k-step power is calculated by iteratively inputting real-time data, supporting multi-step prediction, and the corresponding prediction result and confidence interval or other uncertainty estimation method prediction result can be output.

[0014] As a further improvement of the application, in step S1, the data processing step comprises:

[0015] Missing value processing:

[0016] The missing values in the data are processed using linear interpolation or forward filling method, for time series data points x t , the missing values x mIt can be calculated by the following formula (1):

[0017] (1)

[0018] Wherein, x m is the missing value at time m; x m-1 is the data value at the previous time m-1 of the missing data; x m+1 is the data value at the next time m+1 of the missing data;

[0019] Outlier detection:

[0020] Data outliers are detected for all data, and abnormal data obviously beyond the physical range is corrected. Z-score method is adopted, if , it is regarded as an anomaly and replaced with median or other appropriate value, wherein Z is the result of Z-score value; x is the data of the detected data point; μ is the mean of the data set; σ is the standard deviation of the data set;

[0021] Data preprocessing:

[0022] All feature data is scaled to the required range to eliminate the influence of different dimensions on model training. For irradiance data, the following formula (2) is used for normalization method:

[0023] (2)

[0024] Wherein, x norm is the normalized result; x is the original data; x min is the minimum value of the data in the data set; x max is the maximum value of the data in the data set.

[0025] As a further improvement of the application, in step S2, the historical power similarity in the static coupling relationship is constructed based on the historical power correlation between power plants according to the correlation coefficient formula, using the Pearson correlation coefficient formula as follows:

[0026] (3)

[0027] Wherein, R ij is the power correlation of node i and node j, and are the average power of the power plants represented by node i and node j, respectively; and Pit and Pjt are the power of the power station represented by node i and node j at time t, respectively, based on which the correlation between the output powers of the nodes can be verified, and an adjacency matrix A is constructed, and the correlation between two nodes is established if the correlation exceeds a certain threshold value p, and the edge value of the correlation in the adjacency matrix is set to 1, otherwise, the value is set to 0, and the mathematical expression relationship can be simplified as follows:

[0028] (4)

[0029] Wherein, A is an adjacency matrix; a i，j is the element value between node i and node j in the adjacency matrix; R ij is the historical power correlation coefficient between node i and node j; p is the preset correlation threshold.

[0030] As a further improvement of the application, in step S2, the geographical position similarity in the static coupling relationship is constructed based on the invariable geographical distance between power stations according to the distance calculation formula, and the geographical distance between photovoltaic power stations can be calculated using the Haversine formula:

[0031] (5)

[0032] Wherein, φ and λ are the longitude and latitude, respectively; the subscripts i and j represent nodes i and j, respectively; r is the radius of the earth, and if d ij < δ, wherein δ is the distance threshold, then the edge between node i and j is established in the static coupling relationship, indicating that there is significant power correlation between the two power stations, and the weight of the edge is calculated by the following formula to reflect the attenuation effect of distance on the coupling strength between power stations:

[0033] (6)

[0034] Wherein, ω ij is the edge weight between node i and j, and β is an attenuation parameter for controlling the influence of distance on the weight, and the greater the weight value, the stronger the coupling relationship between the two power stations, and the subsequent static adjacency matrix is constructed based on the edge and the weight, providing a basis for the topology of the dynamic graph network G(t).

[0035] As a further improvement of the application, in step S2, the definition of the power station nodes in the constructed graph is that the N photovoltaic power stations in the region are defined as N node sets V = {v1, v2,..., v N} in the graph, and each node v i has a feature vector X i (t) at each time step t, which contains the time sequence characteristics of the power station itself including historical power, irradiance and temperature, and the directed edge is defined as any two power station nodes v i and v j at any time t.i and v j There is a directed edge e ij between i and j, the directionality of the edge is determined by the meteorological propagation direction, which includes the wind direction, based on the cosine value of the angle θ ij between the wind direction and the direction of the line connecting the power station i to j, if cos θ ij > ρ, ρ is a certain threshold value, then a directed edge from i to j is established, indicating that meteorological fluctuations can propagate from i to j; if cos θ ij ≤ ρ, then according to the wind direction, it is determined whether to establish a reverse direction edge from j to i, or not to establish an edge, depending on whether the angle meets the threshold condition of reverse propagation.

[0036] As a further improvement of the application, in step S2, the dynamic graph network G(t) introduces a time lag to reflect the delayed impact of meteorological fluctuations in upstream power stations on downstream power stations, calculated by formula (7) and formula (8):

[0037] (7)

[0038] (8)

[0039] Where τ ij is the introduced time lag; S t is the wind speed at height H0; v t is the wind speed at height H; α is the wind shear exponent, which is taken as an appropriate value according to local ground conditions; the time lag is based on the distance D ij and the equivalent wind speed S t at high altitude; the time lag can be used to adjust the edge weight of the future time step, distributing the influence at the current time t to t+k steps, where k is calculated based on the time lag and the time step interval Δt, as formula:

[0040] (9).

[0041] As a further improvement of the application, in step S2, the dynamic adjacency matrix considers the quantification of the impact caused by meteorological fluctuations, based on the difference between the current grid-connected power and the expected power, the power fluctuation ΔP% when a fluctuation event has occurred is calculated as formula (10) to quantify the real-time impact of different meteorological fluctuations on power output:

[0042] (10)

[0043] Where P i (t) is the output power of node i power station at time t; is the expected output power of node i power station at time t; ΔP% is the quantified power output impact, based on which the time lag Derive the influence on the future to simulate the physical process of wave propagation.

[0044] As a further improvement of the application, in step S2, the construction of the dynamic graph network G(t) is realized by fusing the static coupling relationship and the dynamic adjacency matrix, first, a static adjacency matrix A static is constructed based on geographical location, distance or historical correlation dynamic ; then a dynamic weight matrix W static (t) is calculated based on real-time weather and time lag; finally, an edge matrix E(t) = A dynamic ·W (t) or other fusion methods are generated, where · represents element-level multiplication, to ensure that the static matrix provides the basic topology, while the dynamic matrix adjusts the real-time weight and direction to form a reasonable space-time graph network.

[0045] As a further improvement of the application, in step S3, the hybrid model is constructed by combining graph neural network and recurrent neural network, GNN variant processes the spatial dependence of dynamic graph G(t), and aggregates node features; RNN variant processes the time sequence output by GNN, and captures time dynamics; the input includes G(t) and preprocessed data, and the training is optimized by end-to-end loss function.

[0046] As a further improvement of the application, in step S4, the prediction accuracy requirement is evaluated by corresponding error indicators, including root mean square error RMSE, mean absolute error MAE and mean absolute percentage error MAPE, and the requirement index is required to be less than a preset threshold; if not satisfied, the model is optimized;

[0047] (11)

[0048] (12)

[0049] (13)

[0050] Wherein, RMSE is the root mean square error to be solved; MAE is the mean absolute error to be solved; MAPE is the mean absolute percentage error to be solved; M is the total number of data samples; y m is the actual value of the mth sample; is the predicted value of the mth sample.

[0051] Compared with the prior art, the application has the following advantages:

[0052] The application integrates the dispersed photovoltaic power station into a dynamically correlated system, significantly improves the prediction accuracy, real-time response capability and physical interpretability compared with the traditional method, provides a reliable solution for power management of large-scale photovoltaic power station group, and improves the stability and efficiency of the actual photovoltaic system. Attached Figure Description

[0053] Figure 1 : Overall system flowchart of the method of the present invention.

[0054] Figure 2 Heat map of power correlation between power plants.

[0055] Figure 3 Schematic diagram of the correlation of irradiance between power plants.

[0056] Figure 4 : Schematic diagram of the geographical distribution of photovoltaic power stations.

[0057] Figure 5 Figure : Partial experimental results using only recurrent neural networks (RNNs) for prediction.

[0058] Figure 6 Figure 1 shows some experimental results of prediction based on static graph neural networks (GNN).

[0059] Figure 7 The results of the ultra-short-term forecasting method for photovoltaic power plant clusters based on meteorological dynamic graph network proposed in this invention are shown in the figure. Detailed Implementation

[0060] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments and accompanying drawings. It should be understood that the described embodiments are only some examples of the present invention and not all of the present invention. Other embodiments obtained by those skilled in the art without creative effort should also be considered within the scope of protection covered by the present invention.

[0061] The specific implementation of this invention is described in detail using a regional cluster containing three photovoltaic power plants as an example. The core idea of ​​this method is to treat each power plant as a node in a graph and dynamically construct directed weighted edges between nodes using meteorological physics principles. This dynamic graph network is then combined with the actual operating data of the power plants and input into a deep learning model for power prediction. The technical solution will be elaborated below to provide sufficient technical details to ensure operability and meet the corresponding writing requirements.

[0062] Figure 1 This is a flowchart illustrating the overall system process of the method of this invention, showing the complete flow from data preparation, data preprocessing, dynamic graph network construction to the final power prediction output. Figure 1 As shown, this embodiment of the invention provides a meteorological-based dynamic graph network method for ultra-short-term forecasting of photovoltaic power plant clusters, including the following steps:

[0063] Step S1: Obtain ultra-short-term time series data of a plurality of geographically dispersed photovoltaic power stations, the data including grid-connected power data, irradiance and regional meteorological data (including wind speed, wind direction, temperature, humidity, air pressure) of each power station, and clean, handle missing values and normalize the data;

[0064] Step S2: Define the photovoltaic power stations as nodes in the graph, and establish a relatively static coupling relationship between the nodes based on geographical location, distance or correlation analysis; at the same time, use the meteorological data to calculate the directed edge weight between the power station nodes based on meteorological physical factors (such as wind direction, wind speed, cloud cover), to dynamically construct a dynamic adjacency matrix reflecting the physical propagation effect of meteorological fluctuations, and finally combine the static coupling relationship and dynamic adjacency matrix to construct a dynamic graph network G(t)=(V, E(t));

[0065] Step S3: Construct a hybrid model composed of a graph neural network (GNN) and a recurrent neural network (RNN), input the dynamic graph network G(t) and the preprocessed time series data of each power station into the hybrid model for training until the preset training accuracy requirement is met, and complete model training;

[0066] Step S4: Determine whether the power prediction result meets the preset prediction accuracy requirement; if not, return to the model construction and training stage to retrain or optimize the hybrid model until the prediction accuracy meets the specific situation requirement;

[0067] Step S5: When the hybrid model meets the prediction performance accuracy threshold, use the latest real-time meteorological data and historical power data, combined with the real-time updated dynamic graph network G(t), to perform power prediction in a specific future time period through the hybrid model that meets the accuracy requirement.

[0068] The application discloses a dynamic graph network photovoltaic power station group ultra-short-term prediction method based on weather. With the rapid development of distributed photovoltaic power generation technology, the scale of photovoltaic power station groups is continuously expanding. Photovoltaic power generation has the characteristics of intermittency and volatility, is greatly affected by weather conditions, and leads to the increase of the difficulty of photovoltaic power station group power prediction. In order to solve the problems of low prediction accuracy, poor real-time performance and weak physical interpretability of the photovoltaic power station group power prediction method in the prior art, the application constructs a dynamic space-time graph network, regards photovoltaic power stations in a region as a complex system connected by weather fluctuations, dynamically adjusts the edge weight and time lag of the graph by using real-time weather data, captures the overall trend under stable conditions and rapid fluctuations caused by sudden events, and effectively integrates spatial dependence and time dynamics. Compared with the prior art, the application can improve prediction accuracy, real-time response capability and physical interpretability, support multi-step rolling prediction and uncertainty estimation, better adapt to the fluctuation characteristics of large-scale photovoltaic systems, and improve power grid scheduling efficiency and system robustness.

[0069] The application can refer to a variety of different feature variables, which mainly depend on specific sensor deployment. In the embodiment, the feature variables referred to are shown in Table 1.

[0070] Table 1 Feature variables that can be referred to

[0071]

[0072] The data required by the application mainly comes from real-time monitoring systems of three photovoltaic power stations in a region. The data acquisition frequency should be high enough, for example, 15 minutes or shorter intervals, to support ultra-short-term power prediction.

[0073] Before using historical power data and irradiance data to establish a static coupling relationship graph, the historical data needs to be preprocessed. The preprocessing operation specifically includes missing value processing, for example, using a linear interpolation or forward filling method to process the null values in the data. For time series data points x t , the missing values x m can be calculated by the formula ; the Z-score method can be used for anomaly detection, and if |z| > 3, it is considered to be abnormal and is replaced by the median or other appropriate values; at the same time, data normalization is performed to scale all feature data to a unified range to eliminate the influence of different dimensions on model training. For convenience of display, Table 2 is a part of the data processing flow of a certain day.

[0074] Table 2 Data and data processing table

[0075]

[0076] In the establishment of static coupling relationship graph for distributed photovoltaic power station cluster, a single correlation coefficient index can be used as the basis for decision, for example, using Pearson correlation coefficient formula to calculate:

[0077] ;

[0078] where, R ij is the power correlation of node i and node j, and are the average power of the power stations represented by node i and node j respectively; and are the power of the power stations represented by node i and node j at time t. Based on this, the correlation relationship between the output power of each node power station can be verified, and the adjacency matrix A is constructed. In the case where the correlation exceeds a certain threshold value p, the connection between the two node power stations is established, and the related edge value in the adjacency matrix is set to 1, otherwise, the value can be set to 0. The mathematical expression relationship can be simplified as follows:

[0079] ;

[0080] Figure 2 is the power correlation heat map between power stations, which shows the static coupling relationship based on Pearson correlation coefficient, as shown in Figure 2 , the heat map can be established according to the correlation coefficient and the adjacency matrix can be constructed according to this. If the threshold value p is 0, the adjacency matrix A can be constructed as follows.

[0081] ;

[0082] Not only the power correlation, but also the irradiance as an important indicator of photovoltaic power generation often has a certain correlation between regions, Figure 3 is a schematic diagram of irradiance correlation between power stations, which shows the correlation distribution of irradiance characteristics of each photovoltaic power station, which can be used as another basis for constructing the adjacency matrix, as shown in Figure 3 , the same relationship can also be used to construct a more stable adjacency matrix.

[0083] In addition, the static coupling relationship can also be integrated with the similarity of geographical location, and the Haversine formula can be used to calculate the geographical location distance:

[0084] ;

[0085] where, φ and λ are the longitude and latitude respectively; subscripts i and j represent nodes i and j respectively; r is the radius of the earth, about 6371 km. Based on this, if d ijIf the distance threshold is less than δ (δ is the distance threshold, such as 50km), then an edge from node i to j is established in the static coupling relationship, indicating a significant power correlation between the two power stations. The edge weight is calculated using the following formula to reflect the attenuation effect of distance on the coupling strength between the power stations:

[0086] ;

[0087] Where, ω ij Let β be the edge weight between nodes i and j, and β be the attenuation parameter (e.g., 100km) used to control the influence of distance on the weight. The larger the weight value, the stronger the coupling relationship between the two power stations. The subsequent static adjacency matrix is ​​constructed based on the above edges and weights, providing the foundation for the topology of the dynamic graph network G(t).

[0088] Figure 4 This is a schematic diagram showing the geographical distribution of photovoltaic power stations. The specific geographical locations of the three photovoltaic power stations in the example are marked. Using the Haversine formula, distances can be further calculated to form a distance weight matrix, as shown below. Figure 4 Based on the Haversine formula, the geographical locations of the power stations referenced in this embodiment are calculated as follows: the distance between power stations N1 and N2 is 61.51 km, the distance between power stations N2 and N3 is 51.69 km, and the distance between power stations N1 and N3 is 106.73 km. If the attenuation coefficient β is taken as 100, the distance weight matrix can be obtained as follows:

[0089] ;

[0090] To construct a dynamic spatiotemporal graph network, the three photovoltaic power stations in the region are first defined as a node set V = {v1, v2, v3} in the graph, with each time step being 15 minutes. Each of the three nodes has a feature vector composed of the processed data shown in Table 2, containing the power station's own temporal characteristics (such as historical power, irradiance, temperature, etc.). Directed edges are defined at any time t, where any two power station nodes v1, v2, v3, v3, v4, v5, v6, v7, v8, v9, v1, v2, v3 ...9, v1, v2, v3, v3, v6, v9, v1, v2, v3, v3, v6, v1, v2, v3, v3, v6, v1, v2, v3, v3, v6, v i and v j There exists a directed edge e ij Its directionality is determined by the direction of meteorological propagation (especially wind direction), for example, based on the angle θ between the wind direction and the direction of the line connecting power station i to j. ij Calculation of cosθ value, if cosθ ij If ρ (ρ is a set threshold), then a directed edge is established from i to j, indicating that meteorological fluctuations can propagate from i to j; if cosθ ij If ≤ ρ, then it is determined whether to establish an edge pointing in the opposite direction from j to i, or not to establish an edge, based on the wind direction. The specific decision depends on whether the included angle satisfies the threshold condition for backpropagation. For example... Figure 4The specific geographical locations of the power plants are shown, and according to the wind direction data shown, the edge direction matrix can be completed at each moment.

[0091] Table 3 Wind direction data and edge direction table

[0092]

[0093] ;

[0094] The construction of the dynamic graph network further introduces a time delay to reflect the delayed influence of meteorological fluctuations of upstream power plants on downstream power plants, for example, by calculating the time delay according to the formula:

[0095] ;

[0096] ;

[0097] wherein τ ij is the introduced time delay; S t is the wind speed at height H0; v t is the wind speed at height H; and a is the wind shear exponent, which is taken as an appropriate value according to local ground conditions; the time delay is based on the distance D ij and the equivalent wind speed S t at high altitude; the time delay can be used to adjust the edge weight of the future time step, for example, to distribute the influence of the current t moment to t+k step, wherein k is based on the time delay and the time step interval Δt.

[0098] At the same time, the dynamic adjacency matrix considers the inversion of meteorological fluctuations, by calculating the percentage difference ΔP%

[0099] ;

[0100] Then combine the time delay to deduce the influence on the future to simulate the physical process of fluctuation propagation, in the embodiment, taking power plant N1 at noon as an example, its power fluctuation ΔP% is 2.4%, and combining the actual situation, its time delay result τ 12 is similar to the influence on power plant N2 after 2 hours, i.e. 8 time steps, and by the same reasoning, the fluctuation at this moment is also affected by another two power plants k time steps ago, which can be constructed as a matrix:

[0101] ;

[0102] The final construction of the dynamic graph network G(t) is realized by fusing the static coupling relationship and the dynamic adjacency matrix, for example, first construct a static adjacency matrix A static based on geographical location, distance or historical correlation, then calculate the dynamic weight matrix Wdynamic (t) (based on real-time weather and time lag):

[0103] ;

[0104] The edge matrix E(t) = A static ·W dynamic (t) is finally generated, so as to ensure that the static matrix provides a basic topology, and the dynamic matrix adjusts real-time weights and directions to form a reasonable spatio-temporal graph network.

[0105] Further, in the hybrid model construction and training phase, the model can be realized by the combination of a graph neural network and a recurrent neural network, for example, a GNN variant (such as GCN) processes the spatial dependence of the dynamic graph G(t) and aggregates node features; an RNN variant (such as LSTM) processes the time sequence output by the GNN and captures time dynamics. The input includes G(t) and preprocessed data, and the training is performed through end-to-end optimization of the loss function.

[0106] Model verification and tuning are performed by evaluating the prediction accuracy through multiple error indicators, such as root mean square error RMSE, mean absolute error MAE, and mean absolute percentage error MAPE:

[0107] ;

[0108] ;

[0109] ;

[0110] If the evaluation indicators are better than the preset threshold, the training is completed; if not, the training phase is returned to for retraining or tuning. The above indicators can also be used to prove the performance of the network. In this example, the prediction accuracy of the model without introducing spatial dependence is shown in Table 1, and the prediction results of the static graph neural network are shown in Table 2. The prediction results of the network constructed by the proposed dynamic graph network photovoltaic power station group ultra-short-term power prediction method based on weather are shown in Table 3. Figure 5 Figure 6 Figure 7 Figure 5 Figure 6

[0111] ​​​​​It can be seen that the application integrates the dispersed photovoltaic power station into a dynamically correlated system, significantly improves the prediction accuracy, real-time response capability and physical interpretability compared with the traditional method, provides a reliable solution for power management of large-scale photovoltaic power station group, and improves the stability and efficiency of the actual photovoltaic system.

[0112] Table 4 Comparison of prediction effects of different methods

[0113]

[0114] Further, in the last real-time power prediction stage, the trained network can be saved, the future k-step power is calculated through iterative input of real-time data for multi-step prediction, and the corresponding prediction results and confidence intervals can be output, so as to complete the complete work flow of the system.

[0115] The above is only a preferred embodiment of the application, and is not intended to limit the application in any other form, and any modification or equivalent change made according to the technical essence of the application still falls within the scope of the application claimed.

Claims

1. A meteorology-based dynamic graph network photovoltaic power station group ultra-short-term prediction method, characterized in that, The method comprises the following steps: S1, obtaining ultra-short time series data of a plurality of photovoltaic power stations dispersed in geographical positions, the data comprising grid-connected power data, irradiance and regional meteorological data of each power station, the regional meteorological data comprising wind speed, wind direction, temperature, humidity and air pressure, and performing missing value processing, abnormal value detection and data preprocessing on the data; S2, defining the photovoltaic power stations as nodes in a graph, and establishing a relatively static coupling relationship between the nodes based on geographical position similarity, power station distance relationship, historical power similarity or other correlation analysis; at the same time, using the meteorological data, calculating the directed edge weight between the power station nodes based on meteorological physical factors to dynamically construct a dynamic adjacency matrix reflecting the physical propagation effect of meteorological fluctuations, and finally combining the static coupling relationship and the dynamic adjacency matrix to construct a dynamic graph network G(t) = (V, E(t)), wherein V represents a set composed of N nodes, and E(t) represents a dynamic edge of the graph structure; S3, constructing a hybrid model composed of a graph neural network and a recurrent neural network, inputting the dynamic graph network G(t) and the preprocessed time series data of each power station into the hybrid model for training until a preset training accuracy requirement is met, and completing model training; S4, judging whether the power prediction result meets a preset prediction accuracy requirement; if not, returning to the model construction and training stage to retrain or optimize the hybrid model until the prediction accuracy meets the corresponding situation requirement; S5, when the hybrid model meets the prediction performance accuracy threshold, using the latest real-time meteorological data and historical power data, combining the real-time updated dynamic graph network G(t), and performing power prediction in a specific time period in the future through the hybrid model meeting the accuracy requirement, the power prediction adopting a prediction mechanism, including rolling prediction, future k-step power calculated by iterative input of real-time data, supporting multi-step prediction, and being capable of outputting corresponding prediction results and confidence intervals or prediction results of other uncertainty estimation methods. 2.The weather-based dynamic graph network photovoltaic power station group ultra-short-term prediction method according to claim 1, characterized in that: In the step S1, the data processing steps comprise: missing value processing: Using linear interpolation or forward fill method to handle missing values in data, for time series data point x t , missing value x m may be calculated by the following equation (1): (1) Wherein, x m is the missing value at time m; x m-1 is the data value at the previous time m-1 of the missing data; x m+1 is the data value at the next time m+1 of the missing data; abnormal value detection: Data outliers detection is performed on all data and abnormal data that obviously exceed the physical range are corrected. Z-score method is adopted. If then it is considered as abnormal and replaced by median or other appropriate value, wherein Z is the result of Z-score value; x is the data of the detected data point; μ is the mean of the data set; σ is the standard deviation of the data set. data preprocessing: scaling all feature data to a required range to eliminate the influence of different dimensions on model training, and using the following formula (2) for normalization for irradiance data: (2) where x norm is the normalized result; x is the original data; x min is the minimum value of the data in the data set; x max is the maximum value of the data in the data set. 3.The weather-based dynamic graph network photovoltaic power station group ultra-short-term prediction method according to claim 2, characterized in that: In the step S2, the historical power similarity in the static coupling relationship is constructed based on the historical power correlation between power stations according to a correlation coefficient formula, and the Pearson correlation coefficient formula is used as follows: (3) wherein R ij is the power correlation of node i and node j, is the average power of the power station represented by node i and node j, respectively; is the power of the power station represented by node i and node j at time t, respectively; is the power of the power station represented by node i and node j at time t, respectively; Based on this, the correlation relationship between the output powers of the nodes and power stations can be verified, and an adjacency matrix A is constructed. In the case where the correlation exceeds a certain threshold value p, a connection between the two nodes and power stations is established, and the relevant edge value in the adjacency matrix is set to 1, otherwise, the value is set to 0. The mathematical expression relationship can be simplified as follows: (4) wherein A is an adjacency matrix; a i，j is an element value between node i and node j in the adjacency matrix; R ij is a historical power correlation coefficient between node i and node j; and p is a preset correlation threshold. 4.The weather-based dynamic graph network photovoltaic power station group ultra-short-term prediction method according to claim 3, characterized in that: In the step S2, the geographical position similarity in the static coupling relationship is constructed based on the invariable geographical distance between power stations according to a distance calculation formula, and the Haversine formula can be used to calculate the geographical distance between photovoltaic power stations: (5) where φ and λ are the longitude and latitude, respectively; the subscripts i and j represent nodes i and j, respectively; r is the earth radius, and d is the distance between nodes i and j. ij where δ is the distance threshold, then an edge between nodes i and j is established in the static coupling relationship, indicating that there is a significant power correlation between the two power plants, and the weight of the edge is calculated by the following formula to reflect the attenuation effect of distance on the coupling strength between power plants: (6) where ω ij is the edge weight between nodes i and j, and β is the decay parameter used to control the strength of the distance effect on the weight. The larger the weight value, the stronger the coupling relationship between the two power plants. The subsequent static adjacency matrix is based on the above edge and weight, providing a basis for the topology of the dynamic graph network G(t).

5. The dynamic graph network meteorology-based ultra-short-term prediction method for a photovoltaic power plant group according to claim 4, characterized in that: The definition of power station nodes in the graph constructed in step S2 is: N photovoltaic power stations in the region are defined as N node set V = {v1, v2,..., vN} in the graph. N} each node v i In each time step t, there is a feature vector X i (t) containing the time series characteristics of the power station itself including historical power, irradiance and temperature. The directed edge is defined as: at any time t, there is a directed edge e i between any two power station nodes v j and v ij The directionality of the edge is determined by the meteorological propagation direction, which includes the wind direction. The cosine value based on the angle θ ij between the wind direction and the connecting direction of the power station i to j is calculated. If cosθ ij > ρ, ρ is a certain threshold value, then a directed edge from i to j is established, indicating that meteorological fluctuations can propagate from i to j. If cosθ ij ≤ ρ, then according to the wind direction, it is judged whether to establish an opposite direction edge from j to i or not to establish an edge, depending on whether the angle meets the threshold condition of reverse propagation. 6.The weather-based dynamic graph network photovoltaic power station group ultra-short-term prediction method according to claim 5, characterized in that: In the step S2, the dynamic graph network G(t) introduces time delay To reflect the delayed influence of the weather fluctuation of the upstream power station on the downstream power station, the formula (7) and the formula (8) are calculated: (7) (8) where τ ij is the introduced time lag; S t is the wind speed at height H0; v t is the wind speed at height H; a is the wind shear exponent, which is taken as an appropriate value according to local ground conditions; the time lag is based on the distance D ij and the equivalent wind speed at high altitude S t ; the time lag can be used to adjust the edge weight of the future time step, distributing the influence of the current time t to the t+k step, where k is calculated based on the time lag and the time step interval At, as follows: (9)。 7.The weather-based dynamic graph network photovoltaic power station group ultra-short-term prediction method according to claim 6, characterized in that: In the step S2, the dynamic adjacency matrix considers the quantification of the influence caused by meteorological fluctuations, and calculates the power fluctuation ΔP% when a fluctuation event has occurred based on the difference between the current grid-connected power and the expected power as follows (10) to quantify the real-time influence of different meteorological fluctuations on power output: (10) where P i (t) is the output power of the power plant at node i at time t; is the forecasted output power of the power plant at node i at time t; ΔP% is the quantized power output impact, based on which the future impacts can be derived to simulate the physical process of the wave propagation. to derive the future impacts to simulate the physical process of the wave propagation. 8.The weather-based dynamic graph network photovoltaic power station group ultra-short-term prediction method according to claim 7, characterized in that: In the step S2, the construction of the dynamic graph network G(t) is achieved by fusing the static coupling relationship and the dynamic adjacency matrix. First, a static adjacency matrix A is constructed based on geographical location, distance or historical correlation static ; then a dynamic weight matrix W dynamic (t) is calculated based on real-time weather and time lag; finally, an edge matrix E(t) = A static ·W dynamic (t) or other fusion methods are generated, wherein · represents element-level multiplication, to ensure that the static matrix provides the basic topology, while the dynamic matrix adjusts the real-time weight and direction to form a reasonable space-time graph network. 9.The weather-based dynamic graph network photovoltaic power station group ultra-short-term prediction method of claim 8, wherein: In the step S3, the hybrid model is constructed by combining a graph neural network and a recurrent neural network. The GNN variant processes the spatial dependence of the dynamic graph G(t) and aggregates node features. The RNN variant processes the time sequence output by the GNN and captures time dynamics. The input includes G(t) and preprocessed data, and the training is performed by end-to-end optimization of a loss function. 10.The weather-based dynamic graph network photovoltaic power station group ultra-short-term prediction method of claim 9, wherein: In the step S4, the prediction accuracy requirement is evaluated by a corresponding error indicator, including root mean square error RMSE, mean absolute error MAE, and mean absolute percentage error MAPE. The requirement indicator is required to be less than a preset threshold. If not satisfied, the model is tuned. (11) (12) (13) Wherein, RMSE is the root mean square error; MAE is the mean absolute error; MAPE is the mean absolute percentage error; M is the total number of data samples; y m is the actual value of the mth sample; is the predicted value of the mth sample.

Images