Research On Wind Power Forecasting And Optimal Scheduling Of Interregional Wind Power Consumption Based On Artificial Intelligence

With the continuous development of new energy,a large number of large-capacity wind turbines are continuously integrated into the grid operation,which poses new challenges to the reliable,continuous and stable operation of the grid.so accurate wind power prediction and optimal dissipation of wind power dispatch are of great significance.In recent years,wind power industry has developed rapidly.However,large-scale wind power grid connection has brought new problems to the operation of the power grid.The low annual utilization hours and the large amount of abandoned wind power not only lead to a reduction in the economic benefits of the wind power industry,but also seriously affect the healthy growth.The biggest problem with the existing wind power forecasting methods at this stage is that the prediction accuracy is not enough.The first reason is the depth of the model.Second,the scale and dimensions of the training samples are not large enough,and the data cleaning is not strict enough.It is difficult for the model to adapt to complex scenes and flow field relationships in the terrain.How to improve wind power forecasting accuracy and wind power dissipation capacity has become a big challenge.This paper summarizes the current research status of wind power forecasting and regionally optimized wind power dispatching at home and abroad,analyzes wind power output characteristics and its impact on the power grid,and current forecast levels of wind power.The effective reserve capacity of the interconnected area and the remaining transmission capacity of the tie lines are of great significance for the optimization of the entire network.Aiming at wind power prediction,the characteristics of numerical weather forecast(NWP)data are analyzed first,and the space-time offset of NWP data is demonstrated.A historical input of wind farm wind power and wind speed and wind direction is used as the data input structure.Component analysis extracts valid data and reduces data size.A short-term wind power prediction model based on CNN-LSTM is constructed.The processed meteorological data and historical wind power data are used to extract the features of the data and further reduce the data dimension through a convolution network.In the training process of neural network,Drop Connect technology is introduced to reduce the over-fitting phenomenon in the model,and finally to achieve accurate prediction of wind power.Aiming at the optimal dispatch of wind power,a Nash-Q learning method based on large-scale coordinated wind power consumption in interconnected regions is proposed.An economic dispatch model of the interconnected system considering the uncertainty of wind power is established.The optimal goal is to absorb wind power.Firstly,according to the equivalent costs of the interests of the participants in different regions,construct their own state decision models,and establish a non-cooperative game Nash equilibrium model with the goal of balancing the interests among the participants in multiple regions.High-dimensional decision variables,the introduction of Q-learning algorithms,and the establishment of an interconnected system scheduling solution based on a combination of non-cooperative game Nash equilibrium and Q-learning algorithms;Finally,the improved IEEE39 node interconnection system verifies that the proposed method can achieve inter-regional benefits.Reasonable distribution to promote large-scale consumption of wind power.