Artificial intelligence-based power equipment meteorological monitoring and early warning system

Abstract

The application provides an artificial intelligence-based power equipment weather monitoring and early warning system. The artificial intelligence-based power equipment weather monitoring and early warning system comprises a data acquisition module, a data preprocessing module, a space-time feature fusion module, a weather disaster prediction model, the weather disaster prediction model adopts a deep reinforcement learning framework, inputs the multi-dimensional space-time feature matrix, and outputs weather disaster risk levels and key parameter prediction values in a future preset period, including wind speed, precipitation, temperature anomaly, tropical cyclone path probability, a dynamic early warning threshold generation module, an early warning decision module and a model optimization module. The artificial intelligence-based power equipment weather monitoring and early warning system provided by the application has the advantages of being capable of improving the data interpolation accuracy of complex terrain areas, being capable of realizing high-precision prediction of typhoon paths, short-time strong winds and icing risks, and being capable of comprehensively improving the early warning accuracy and defense response efficiency of the power system in response to weather disasters.

Description

Technical Field

[0001] This invention relates to the field of power technology, and in particular to a meteorological monitoring and early warning system for power equipment based on artificial intelligence. Background Technology

[0002] With the expansion of new energy power systems, meteorological disasters pose an increasingly severe threat to power transmission and transformation equipment, wind farms, and photovoltaic power stations. Traditional meteorological monitoring systems rely heavily on data from single meteorological stations, resulting in problems such as low spatial coverage and difficulties in integrating multi-source heterogeneous data, making it difficult to capture the evolution characteristics of disasters such as localized severe convection and sudden changes in typhoon paths in a timely manner.

[0003] In existing technologies, spatial kriging is commonly used for meteorological data interpolation, but it does not fully consider the spatiotemporal coupling characteristics of meteorological elements, leading to high interpolation errors in complex terrain areas such as mountains and coastlines. Disaster prediction models are mostly based on statistical methods or single deep learning architectures, lacking constraints from physical laws, and are prone to misjudgments such as wind speed prediction deviations exceeding 5 m / s under extreme weather conditions. Meanwhile, warning thresholds for power equipment typically use fixed standards, failing to consider dynamic factors such as equipment aging and grid load fluctuations, resulting in a high false alarm rate. Regarding defense strategies, existing systems mostly provide general contingency plans, failing to generate targeted response plans based on equipment geographical distribution and operational status. Furthermore, existing models generally lack online learning capabilities and cannot adapt to weather pattern migrations caused by climate change. Although some studies have attempted to integrate the complementary characteristics of wind and solar resources, they have not established dynamic optimization models that consider minimizing grid fluctuations, resulting in insufficient renewable energy consumption efficiency.

[0004] Therefore, it is necessary to provide an artificial intelligence-based meteorological monitoring and early warning system for power equipment to solve the above-mentioned technical problems. Summary of the Invention

[0005] The technical problem solved by this invention is to provide an artificial intelligence-based meteorological monitoring and early warning system for power equipment that can improve the data interpolation accuracy in complex terrain areas, achieve high-precision prediction of typhoon paths, short-term strong winds and icing risks, and comprehensively improve the early warning accuracy and defense response efficiency of power systems in response to meteorological disasters.

[0006] To address the aforementioned technical problems, the present invention provides an artificial intelligence-based meteorological monitoring and early warning system for power equipment, comprising: a data acquisition module, wherein the data acquisition module is used to acquire multi-source meteorological data and power equipment operation data in real time;

[0007] A data preprocessing module is connected to the data acquisition module. The data preprocessing module is used to perform quality control, missing value imputation, and spatiotemporal consistency correction on the raw data to generate a standardized meteorological dataset.

[0008] The spatiotemporal feature fusion module, based on convolutional neural networks and graph attention networks, extracts the spatial correlation features and temporal series dependencies of meteorological data, and fuses the geographical distribution information of power equipment to construct a multidimensional spatiotemporal feature matrix.

[0009] A meteorological disaster prediction model, which adopts a deep reinforcement learning framework, takes the multi-dimensional spatiotemporal feature matrix as input and outputs the predicted values ​​of meteorological disaster risk level and key parameters for a future preset period, including wind speed, precipitation, temperature anomalies and tropical cyclone path probability.

[0010] A dynamic early warning threshold generation module, which adaptively calculates meteorological risk thresholds for different types of power equipment, regional locations, and operational stages based on historical disaster data and equipment disaster resistance indicators;

[0011] The early warning decision module compares the prediction results with dynamic thresholds, generates graded early warning signals and defense strategy suggestions, and pushes them to the terminal device in real time through a 3D GIS visualization platform.

[0012] The model optimization module is based on an online learning mechanism and uses real-time feedback data on early warning effects to update the prediction model parameters, thereby improving the system's iterative accuracy.

[0013] Preferably, the missing value imputation method of the data preprocessing module includes:

[0014] For the missing portions of meteorological station observation data, a spatiotemporal kriging interpolation algorithm is used, and the interpolated values ​​are calculated according to the following relationship:

[0015]

[0016] Where Z(s0,t0) is the interpolated value of the target position s0 at time t0, and λ i and γ j The spatial and temporal weighting coefficients are determined through semi-variogram optimization. This represents the time gradient operator, used to capture the temporal variation trend of meteorological elements.

[0017] Preferably, the workflow of the spatiotemporal feature fusion module includes the following steps:

[0018] Spatial texture features of radar echo images are extracted by CNN branches, and dilated convolutional layers are used to expand the receptive field to capture the structure of large-scale weather systems.

[0019] A topological graph of meteorological stations is constructed using GAT branches. Node characteristics include wind speed, temperature, and humidity, and edge weights are jointly determined by the geographical distance between stations and meteorological correlation.

[0020] A multi-head attention mechanism is used to dynamically weight and fuse spatiotemporal features to generate a feature matrix that contains the regional meteorological evolution patterns.

[0021] Preferably, the meteorological disaster prediction model includes the following sub-models:

[0022] Typhoon track prediction sub-model: Based on the improved U-Net architecture, it takes satellite cloud images and sea surface temperature field data as input and outputs a probability distribution map of the movement trajectory of the tropical cyclone center.

[0023] Short-term strong wind prediction sub-model: A bidirectional LSTM network is coupled with physical constraint equations, and its wind speed prediction values ​​satisfy the following dynamic correction conditions:

[0024]

[0025] Where α is the machine learning weight coefficient. For pressure gradient force, v obs The current observed wind speed;

[0026] Icing risk assessment sub-model: Based on the random forest algorithm, the icing thickness growth rate of the conductor is calculated by taking into account temperature, humidity, wind speed and equipment surface material parameters.

[0027] Preferably, the dynamic early warning threshold generation module performs the following operations:

[0028] According to the IEC wind resistance rating standard for wind turbine generators, the maximum tolerable wind speed threshold v is classified for different wind turbine types. max ;

[0029] Based on equipment aging test data, a lifespan decay factor β(t) = e is introduced. -kt The threshold is dynamically lowered, where k is the material fatigue coefficient;

[0030] Based on real-time grid load demand, the high temperature warning threshold can be raised during power supply emergencies to avoid unnecessary shutdowns.

[0031] Preferably, the method for generating defense strategy suggestions in the early warning decision module includes:

[0032] When the predicted wind speed exceeds 80% of the design threshold, a pre-warning is triggered and the wind turbine yaw system self-check is initiated.

[0033] When the predicted precipitation triggers the regional flood risk index R flood When the value is greater than 0.7, a substation waterproof gate closing command and a backup power supply switching scheme are generated.

[0034] For photovoltaic power plants, the priority queue of cleaning robots is dynamically adjusted based on the predicted optical thickness τ of sandstorms.

[0035] Preferably, the model optimization module adopts a federated learning framework to realize distributed model training for multiple regional power companies, and its parameter aggregation process satisfies:

[0036]

[0037] in, Let D be the weight parameters of the i-th local model during the t-th round of training. i For local data volume, D total This represents the total amount of global data.

[0038] Preferably, the 3D GIS visualization platform integrates the following functions:

[0039] The thermal layer displays the real-time wind speed field and predicted path, and overlays terrain elevation data to analyze areas where the canyon effect is enhanced.

[0040] Dynamically render the power grid topology, highlighting transmission lines and substations at risk of weather conditions;

[0041] An embedded AR interface allows inspection personnel to obtain real-time micro-meteorological information overlaid around equipment via smart glasses.

[0042] Preferably, it also includes a wind-solar hybrid optimization module, which performs the following:

[0043] The regional wind-solar complementarity index C was calculated based on Kendall's rank correlation coefficient. ws :

[0044]

[0045] Where, N c N represents the number of periods in which wind and solar power output change in the same direction. d This represents the number of periods of reverse change;

[0046] With the goal of minimizing grid fluctuations, the optimal installed capacity ratio is solved. Its objective function is:

[0047]

[0048] in, These represent the variances of wind and solar power output, σ ws Let its covariance be.

[0049] Preferably, the multi-source meteorological data includes meteorological station observation data, radar reflectivity data, satellite remote sensing data, numerical weather prediction data, and atmospheric reanalysis data; the power equipment operation data includes the operating status of wind turbine generators, the output power of photovoltaic power plants, and grid load parameters.

[0050] Compared with related technologies, the artificial intelligence-based meteorological monitoring and early warning system for power equipment provided by this invention has the following beneficial effects:

[0051] This invention provides an artificial intelligence-based meteorological monitoring and early warning system for power equipment. The system integrates multi-source heterogeneous data from meteorological stations, radar, satellites, and power equipment through a data acquisition module, addressing the problem of single data dimensions in traditional monitoring systems. The data preprocessing module employs a spatiotemporal kriging interpolation algorithm fused with spatiotemporal gradient operators, significantly improving the data interpolation accuracy in complex terrain areas. The spatiotemporal feature fusion module extracts multi-scale meteorological features based on a CNN-GAT hybrid network, enhancing the ability to identify small- and medium-scale weather systems. The meteorological disaster prediction model combines a deep reinforcement learning framework with physical constraint equations to achieve high-precision prediction of typhoon paths, short-term strong winds, and icing risks. The dynamic early warning threshold generation module introduces equipment aging factors and a grid load coordination mechanism to establish adaptive hierarchical early warning standards. The early warning decision module uses a 3D GIS platform to visualize and locate the disaster impact range and generate equipment-level defense strategies. The model optimization module uses a federated learning framework to achieve collaborative evolution of multi-regional models, ensuring continuous optimization of the prediction model. The wind-solar hybrid optimization module establishes an optimal capacity ratio model based on wind-solar output correlation analysis, effectively mitigating fluctuations in new energy grid connection. The synergistic effect of these modules comprehensively improves the accuracy of early warnings and the efficiency of defense response to meteorological disasters in the power system. Attached Figure Description

[0052] Figure 1 The schematic diagram of the meteorological monitoring and early warning system for power equipment based on artificial intelligence provided by the present invention. Detailed Implementation

[0053] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0054] Please refer to the following: Figure 1 ,in, Figure 1 This invention provides a principle block diagram of an artificial intelligence-based meteorological monitoring and early warning system for power equipment. The artificial intelligence-based meteorological monitoring and early warning system for power equipment includes:

[0055] The data acquisition module is used to acquire multi-source meteorological data and power equipment operation data in real time.

[0056] The multi-source meteorological data includes meteorological station observation data, radar reflectivity data, satellite remote sensing data, numerical weather prediction data, and atmospheric reanalysis data; the power equipment operation data includes the operating status of wind turbine generators, the output power of photovoltaic power plants, and grid load parameters.

[0057] A data preprocessing module is connected to the data acquisition module. The data preprocessing module is used to perform quality control, missing value imputation, and spatiotemporal consistency correction on the raw data to generate a standardized meteorological dataset.

[0058] The missing value imputation methods in the data preprocessing module include:

[0059] For the missing portions of meteorological station observation data, a spatiotemporal kriging interpolation algorithm is used, and the interpolated values ​​are calculated according to the following relationship:

[0060]

[0061] Where Z(s0,t0) is the interpolated value of the target position s0 at time t0, and λ i and γ j The spatial and temporal weighting coefficients are determined through semi-variogram optimization. This represents the time gradient operator, used to capture the temporal variation trend of meteorological elements.

[0062] The spatiotemporal feature fusion module, based on convolutional neural networks and graph attention networks, extracts the spatial correlation features and temporal series dependencies of meteorological data, and fuses the geographical distribution information of power equipment to construct a multidimensional spatiotemporal feature matrix.

[0063] The workflow of the spatiotemporal feature fusion module includes the following steps:

[0064] Spatial texture features of radar echo images are extracted by CNN branches, and dilated convolutional layers are used to expand the receptive field to capture the structure of large-scale weather systems.

[0065] A topological graph of meteorological stations is constructed using GAT branches. Node characteristics include wind speed, temperature, and humidity, and edge weights are jointly determined by the geographical distance between stations and meteorological correlation.

[0066] A multi-head attention mechanism is used to dynamically weight and fuse spatiotemporal features to generate a feature matrix that contains the regional meteorological evolution patterns.

[0067] A meteorological disaster prediction model, which adopts a deep reinforcement learning framework, takes the multi-dimensional spatiotemporal feature matrix as input and outputs the predicted values ​​of meteorological disaster risk level and key parameters for a future preset period, including wind speed, precipitation, temperature anomalies and tropical cyclone path probability.

[0068] The meteorological disaster prediction model includes the following sub-models:

[0069] Typhoon track prediction sub-model: Based on the improved U-Net architecture, it takes satellite cloud images and sea surface temperature field data as input and outputs a probability distribution map of the movement trajectory of the tropical cyclone center.

[0070] Short-term strong wind prediction sub-model: A bidirectional LSTM network is coupled with physical constraint equations, and its wind speed prediction values ​​satisfy the following dynamic correction conditions:

[0071]

[0072] Where α is the machine learning weight coefficient. For pressure gradient force, v obs The current observed wind speed;

[0073] Icing risk assessment sub-model: Based on the random forest algorithm, the icing thickness growth rate of the conductor is calculated by taking into account temperature, humidity, wind speed and equipment surface material parameters.

[0074] A dynamic early warning threshold generation module, which adaptively calculates meteorological risk thresholds for different types of power equipment, regional locations, and operational stages based on historical disaster data and equipment disaster resistance indicators;

[0075] The dynamic early warning threshold generation module performs the following operations:

[0076] According to the IEC wind resistance rating standard for wind turbine generators, the maximum tolerable wind speed threshold v is classified for different wind turbine types. max ;

[0077] Based on equipment aging test data, a lifespan decay factor β(t) = e is introduced. -kt The threshold is dynamically lowered, where k is the material fatigue coefficient;

[0078] Based on real-time grid load demand, the high temperature warning threshold can be raised during power supply emergencies to avoid unnecessary shutdowns.

[0079] The early warning decision module compares the prediction results with dynamic thresholds, generates graded early warning signals and defense strategy suggestions, and pushes them to the terminal device in real time through a 3D GIS visualization platform.

[0080] The methods for generating defense strategy recommendations in the early warning decision-making module include:

[0081] When the predicted wind speed exceeds 80% of the design threshold, a pre-warning is triggered and the wind turbine yaw system self-check is initiated.

[0082] When the predicted precipitation triggers the regional flood risk index R flood When the value is greater than 0.7, a substation waterproof gate closing command and a backup power supply switching scheme are generated.

[0083] For photovoltaic power plants, the priority queue of cleaning robots is dynamically adjusted based on the predicted optical thickness τ of sandstorms.

[0084] The 3D GIS visualization platform integrates the following functions:

[0085] The thermal layer displays the real-time wind speed field and predicted path, and overlays terrain elevation data to analyze areas where the canyon effect is enhanced.

[0086] Dynamically render the power grid topology, highlighting transmission lines and substations at risk of weather conditions;

[0087] An embedded AR interface allows inspection personnel to obtain real-time micro-meteorological information overlaid around equipment via smart glasses.

[0088] The model optimization module is based on an online learning mechanism and uses real-time feedback data on early warning effects to update the prediction model parameters, thereby improving the system's iterative accuracy.

[0089] The model optimization module adopts a federated learning framework to achieve distributed model training across multiple regional power companies. Its parameter aggregation process satisfies the following:

[0090]

[0091] in, Let D be the weight parameters of the i-th local model during the t-th round of training. i For local data volume, D total This represents the total amount of global data.

[0092] It also includes a wind-solar hybrid optimization module, which performs the following:

[0093] The regional wind-solar complementarity index C was calculated based on Kendall's rank correlation coefficient. ws :

[0094]

[0095] Where, N c N represents the number of periods in which wind and solar power output change in the same direction. d This represents the number of periods of reverse change;

[0096] With the goal of minimizing grid fluctuations, the optimal installed capacity ratio is solved. Its objective function is:

[0097]

[0098] in, These represent the variances of wind and solar power output, σ ws Let its covariance be.

[0099] The artificial intelligence-based meteorological monitoring and early warning system for power equipment provided by this invention is deployed in a hybrid architecture of edge computing nodes and cloud platforms, wherein: the edge performs real-time data preprocessing and low-latency early warning generation; the cloud performs large-scale model training and cross-regional meteorological model assimilation calculation; and the entire process of early warning decision-making is recorded through blockchain evidence storage technology to ensure the traceability of operations.

[0100] Compared with related technologies, the artificial intelligence-based meteorological monitoring and early warning system for power equipment provided by this invention has the following beneficial effects:

[0101] This invention provides an artificial intelligence-based meteorological monitoring and early warning system for power equipment. The system integrates multi-source heterogeneous data from meteorological stations, radar, satellites, and power equipment through a data acquisition module, addressing the problem of single data dimensions in traditional monitoring systems. A data preprocessing module employs a spatiotemporal kriging interpolation algorithm fused with spatiotemporal gradient operators, significantly improving data interpolation accuracy in complex terrain areas. A spatiotemporal feature fusion module extracts multi-scale meteorological features based on a CNN-GAT hybrid network, enhancing the ability to identify small- and medium-scale weather systems. A meteorological disaster prediction model combines a deep reinforcement learning framework with physical constraint equations to achieve high-precision prediction of typhoon paths, short-term strong winds, and icing risks. A dynamic early warning threshold generation module introduces equipment aging factors and a grid load coordination mechanism to establish adaptive hierarchical early warning standards. An early warning decision module uses a 3D GIS platform to visualize and locate the disaster impact range and generate equipment-level defense strategies. A model optimization module uses a federated learning framework to achieve collaborative evolution of multi-regional models, ensuring continuous optimization of the prediction model. A wind-solar hybrid optimization module establishes an optimal capacity ratio model based on wind-solar output correlation analysis, effectively mitigating fluctuations in new energy grid connection. The synergistic effect of these modules comprehensively improves the accuracy of early warnings and the efficiency of defense response to meteorological disasters in the power system.

[0102] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. An artificial intelligence-based power equipment meteorological monitoring and early warning system, characterized in that, The utility model relates to a kind of meteorological disaster prediction system based on deep learning, comprising: Data acquisition module, the data acquisition module is used to obtain multi-source meteorological data and power equipment operation data in real time; Data preprocessing module, the data preprocessing module is connected to the data acquisition module, the data preprocessing module is used to carry out quality control, missing value interpolation and space-time consistency correction to original data, generate standardized meteorological data set; Spacetime feature fusion module, the spacetime feature fusion module is based on convolutional neural network and graph attention network, extracts the spatial correlation feature and time series dependency of meteorological data, and fuses the geographical distribution information of power equipment, constructs multidimensional spacetime feature matrix; Meteorological disaster prediction model, the meteorological disaster prediction model uses deep reinforcement learning framework, inputs the multidimensional spacetime feature matrix, outputs meteorological disaster risk level and key parameter prediction value in future preset period, including wind speed, precipitation, temperature anomaly and tropical cyclone path probability; Dynamic early warning threshold generation module, the dynamic early warning threshold generation module adaptively calculates meteorological risk threshold of different power equipment types, regional location and operation stage according to historical disaster data and equipment disaster resistance capability index; Early warning decision module, the early warning decision module compares prediction result with dynamic threshold, generates graded early warning signal and defense strategy suggestion, and pushes to terminal equipment through three-dimensional GIS visualization platform in real time; Model optimization module, the model optimization module updates prediction model parameter using real-time feedback early warning effect data based on online learning mechanism, improves system iteration accuracy; The missing value interpolation method of the data preprocessing module includes: For missing part of meteorological station observation data, interpolation value calculation satisfies the following relationship based on spatiotemporal kriging interpolation algorithm: wherein is the target position at time the interpolated value, and are spatial and temporal weight coefficients determined by semi-variogram optimization; denotes a temporal gradient operator to capture the temporal variation trend of the meteorological element; n is the number of spatial samples; m is the number of time samples, is the meteorological element observation value at time is the meteorological element observation value at time 2. The power equipment weather monitoring and early warning system based on artificial intelligence according to claim 1, characterized in that, The working procedure of the spacetime feature fusion module includes the following steps: Through CNN branch, the spatial texture feature of radar echo image is extracted, and the empty convolution layer is used to expand the receptive field to capture large-scale weather system structure; Through GAT branch, the topological relationship graph between meteorological stations is constructed, node feature includes wind speed, temperature, humidity, and edge weight is determined by joint determination of geographical distance and meteorological correlation between stations; Multi-head attention mechanism is used to dynamically weight and fuse spacetime features, and feature matrix containing regional meteorological evolution law is generated. 3.The power equipment weather monitoring and early warning system based on artificial intelligence according to claim 1, characterized in that, The meteorological disaster prediction model includes the following submodels: Typhoon path prediction submodel: based on improved U-Net architecture, input satellite cloud picture and sea temperature field data, output probability distribution map of tropical cyclone center moving track; Short-time strong wind prediction submodel: bidirectional LSTM network and physical constraint equation are coupled, and wind speed prediction value satisfies the following dynamic correction condition: wherein, are machine learning weight coefficients, / are barometric pressure gradient forces, is the current observed wind speed; is the LSTM network predicted wind speed value, is the time step or prediction time interval used in the physical constraint equation to calculate the wind speed change; Icing risk assessment submodel: based on random forest algorithm, temperature, humidity, wind speed and equipment surface material parameters are integrated to calculate conductor icing thickness growth rate. 4.The power equipment weather monitoring and early warning system based on artificial intelligence according to claim 1, characterized in that, The dynamic early warning threshold generation module performs the following operations: According to the IEC wind resistance level standard of wind turbine generator, the maximum wind speed threshold of different wind turbine types is divided ; Based on the device aging detection data, a life attenuation factor is introduced Dynamic downshift threshold, wherein is the material fatigue coefficient, t represents the time variable, indicating the device running time or aging period; In power supply emergency state, high temperature early warning threshold is improved to avoid unnecessary shutdown in combination with real-time power grid load demand. 5.The power equipment weather monitoring and early warning system based on artificial intelligence according to claim 1, wherein, The defense strategy suggestion generation method of the early warning decision module includes: When the predicted wind speed exceeds 80% of the design threshold, a preliminary warning is triggered and a fan yaw system self-check is initiated; When the predicted precipitation amount triggers a regional flood risk index A substation floodgate closing instruction and a standby power switching scheme are generated. For a photovoltaic power station, a cleaning robot operation priority queue is dynamically adjusted according to a dust storm forecast optical thickness. , dynamic adjustment cleaning robot operation priority queue. 6.The power equipment weather monitoring and early warning system based on artificial intelligence according to claim 1, wherein, The model optimization module adopts a federated learning framework to realize distributed model training of multi-regional power companies, and the parameter aggregation process meets: wherein, is the weight parameter of the th local model in the th round of training, is the local data amount, is the total global data amount; is the global model weight parameter in the federated learning framework, and N is the total number of local models. 7.The power equipment weather monitoring and early warning system based on artificial intelligence according to claim 1, characterized in that, The three-dimensional GIS visualization platform integrates the following functions: The heat map layer displays the real-time wind speed field and the predicted path, and superimposes the terrain elevation data to analyze the canyon effect enhancement area; The dynamic rendering of the power grid topology highlights the transmission lines and substations that are at meteorological risk. An AR interface is embedded to support inspection personnel in obtaining device surrounding micro-meteorological live superimposed information through smart glasses. 8.The power equipment weather monitoring and early warning system based on artificial intelligence according to claim 1, wherein, It also includes a wind-solar complementation optimization module, which performs: Calculating a regional wind-solar complementation index based on a kendall rank correlation coefficient : wherein, Nwind-solar is the number of wind-solar power output co-variation periods, Nwind-solar is the number of wind-solar power output co-variation periods, Solve the optimal installed capacity ratio with the goal of minimizing grid fluctuations The objective function is: wherein, , are the wind, light output variance, respectively, is the covariance thereof, is the minimum or target value of the optimal installed capacity ratio solved in the wind-solar complementary optimization module, used for minimizing the grid fluctuation. 9.The power equipment weather monitoring and early warning system based on artificial intelligence according to claim 1, wherein, The multi-source meteorological data includes meteorological station observation data, radar reflectivity data, satellite remote sensing data, numerical weather prediction data, and atmospheric reanalysis data; The power equipment operation data includes wind turbine generator set operation state, photovoltaic power station output power and power grid load parameter.

Images