Short-term Power Forecasting Model For Photovoltaic Plants Based On Deep Learning

Solar energy is widely recognized as an ideal renewable energy source in the future.Photovoltaic power generation,as an important way to use solar energy,can greatly alleviate energy crisis and mitigate environmental problems caused by fossil energy burning.However,the photovoltaic power generation is affected by the intensity of solar radiation and meteorological conditions,and is highly random,fluctuating and intermittent,which poses a severe challenge to large-scale photovoltaic grid-connected power generation.Therefore,the accurate forecasting of photovoltaic power generation has considerable practical significance.The thesis summarizes the research progress in related fields in recent years,and conducts a detailed study on short-term photovoltaic power forecasting.This paper mainly discusses the following aspects:(1)The thesis designs a database for photovoltaic power forecasting through abnormal point monitoring,missing data interpreting and value normalization.(2)A short-term point prediction model based on deep learning is proposed.Firstly,three different model structures are proposed and the advantages and disadvantages of the three are compared experimentally.At the same time,the input data requirements are given,that is,the solar power data of the current day and the next day's weather data.Then,the influence of model capacity on the prediction results is explored.The number of layers and nodes are adjusted separately.Suggestions on the model capacity for single-user photovoltaic power prediction are given.Further the thesis explores the impact of Dropout mechanism and activation function on the model prediction.In addition,different deep models are compared in terms of prediction accuracy and test time.Finally,the three-dimensional weather features that have the greatest impact on the forecasting accuracy are extracted,and the superiority of the three-dimensional feature input is verified on the overall data set.(3)The method of interval forecasting is proposed based on the point prediction of solar photovoltaic power generation.Firstly,the evaluation index of interval prediction is given.Then a heuristic interval prediction algorithm is proposed to achieve the global optimality and low computational cost with the given accuracy.Finally,the validity of the interval prediction algorithm is verified on the test set.