Short-term Wind Power Prediction Based On Similar Time Period Meteorological Data

In today’s increasingly serious world energy crisis and environmental problems,in order to ensure that economic development and environmental protection can go hand in hand,China has formulated the vision of "30 carbon peaks" and "60 carbon neutral" and started to promote the development of clean energy,and the construction of a new power system with The construction of a new type of power system with clean energy as the mainstay is the only way to achieve the goal.In recent years,China’s installed wind power capacity has been growing at an explosive rate.However,wind energy is not as readily available as fossil fuels,and its intermittent and volatile nature makes the output of electricity more unstable,making it extremely difficult to connect wind power to the grid on a large scale.Accurate wind power forecasting can be used to improve the utilization of wind energy,reduce the rate of wind abandonment and ensure the safety and stability of power equipment operation.The current short-term wind power prediction model still cannot meet the accuracy requirements of real-time grid dispatch.In order to make the short-term wind power prediction more accurate,the short-term wind power prediction method is improved by further processing of historical wind power data and meteorological data,and by using a more matching model in two aspects.The main research efforts and innovations of this paper are:Firstly,to address the adverse effects of the oscillatory instability characteristics of wind speed signals and wind power signals on model construction and the difficulty of selecting adjustable parameters for the traditional least squares support vector machine(LSSVM)model.The wavelet transform is chosen to process the raw data,and the LSSVM parameters are optimally selected using the viscous bacteria optimization algorithm,and air pressure,temperature and humidity are added to the input data.The proposed prediction model is proved to be significant in solving the above-mentioned difficulties through case analysis,and the prediction accuracy of wind power is improved to a certain extent.Secondly,in order to further improve the prediction accuracy,the seasonal characteristics and similarities in the meteorological data and historical wind power data,as well as the insufficient ability of the LSSVM algorithm to mine deep features and the difficulties in handling large-scale data,are addressed.Affinity Propagation clustering algorithm,which can automatically obtain the number of clusters,is adopted to find a subset of similar time periods of meteorological data and wind power data by season,incorporate multi-position numerical weather forecast information,combine with a long and short-term memory neural network(LSTM)algorithm with stronger non-linear mapping capability,and use the dimensionality reduction of the input data by using principal component analysis to reduce the dimensionality of the model input data and the amount of network computation at the same time This paper also uses principal component analysis to reduce the dimensionality of the input data and the computational effort of the network,while reducing the data noise,so as to build a prediction model.Finally,simulation experiments were carried out using the measured data from a wind farm in Ningxia,and the results showed that the proposed model could fit better and have higher prediction accuracy.