Research On Wind Power Short-term Prediction Based On Hidden Markov Model

Today fossil fuels are still the main source of global energy,because of its non renewability and huge consumption making the energy crisis serious.Meanwhile,the use of fossil fuels has led to the deterioration of the environment.In order to defuse the energy crisis and improve the ecological environment,all countries in the world have adjusted their energy structure.As renewable energy,wind energy is mostly used in power generation industry.Due to the instability of the wind turbine,causing wind power is difficult to connect to the power grid and consume,which further restricts the development of wind power industry.Wind power predict helps to include wind generation into economic scheduling,optimize unit commitment and reserve allocation,and reduce the operating cost of the system.At the same time,it can provide basis of planning for maintenance for wind farm,and use wind energy effectively.In this paper,HMM is applied to find out the relationship between NWP wind speed and wind power,so as to realize short-term prediction of wind power.The fuzzy C means clustering method is used to classify the historical wind power data and get the initial state of wind power.According to the classification results,the maximum likelihood estimation method is used to calculate the initial probability distribution and the state transition probability matrix of the model.The mean and variance of the Gauss probability density function are calculated from the NWP data corresponding to each state of the wind power.Based on the calculated model parameters and combined with the NWP wind speed data in the next 72 hours,the Viterbi algorithm is applied to predict the wind power state in the next 72 hours.From the predicted power state,the power predicted value is calculated by interpolation method.The single turbine abnormal and missing data of power and NWP data obtained from wind farm are preprocessed,and then an example is analyzed.The prediction results are evaluated by MAE,RMSE and MAPE,the validity of the HMM for short-term wind power prediction is verified.The influence of the number of samples and the number of hidden states on the prediction accuracy of the model is analyzed,and the conclusion is that with the increase of the number of samples and the number of hidden states,the prediction error first decreases and then increases.Therefore,when the HMM is applied to short-term wind power forecasting,better prediction results can be achieved by selecting the appropriate number of samples and the number of hidden states.