The Short-term Solar Power Forecasting Research Based On Deep Learning Theory

Accurate solar photovoltaic power forecasting can help mitigate the potential risk caused by the uncertainty of photovoltaic out power in systems with high penetration levels of solar photovoltaic generation.Weather classification based photovoltaic power forecasting modeling is an effective method to enhance its forecasting precision because photovoltaic output power strongly depends on the specific weather statuses in a given time period.However,the most intractable problems in weather classification models are the insufficiency of training dataset(especially for the extreme weather types)and the selection of applied classifiers.Given the above considerations,a generative adversarial networks and convolutional neural networks-based weather classification model is proposed in this paper.First,33 meteorological weather types are reclassified into 10 weather types by putting several single weather types together to constitute a new weather type.Then a data-driven generative model named generative adversarial networks is employed to augment the training dataset for each weather types.Finally,the convolutional neural networks-based weather classification model was trained by the augmented dataset that consists of both original and generated solar irradiance data.In the case study,we evaluated the quality of generative adversarial networks-generated data,compared the performance of convolutional neural networks classification models with traditional machine learning classification models such as support vector machine,multilayer perceptron,and k-nearest neighbors algorithm,investigated the precision improvement of different classification models achieved by generative adversari al networks,and applied the weather classification models in solar irradiance forecasting.The simulation results illustrate that generative adversarial networks can generate new samples with high quality that capture the intrinsic features of the original data,but not to simply memorize the training data.Furthermore,convolutional neural networks classification models show better classification performance than traditional machine learning models.And the performance of all these classification models is indeed improved to the different extent via the generative adversarial networks-based data augment.In addition,weather classification model plays a significant role in determining the most suitable and precise day-ahead photovoltaic power forecasting model with high efficiency.As the main influence factor of PV power generation,solar irradiance and its accurate forecasting are the prerequisite for solar PV power forecasting.However,previous forecasting approaches using manual feature extraction(MFE),traditional modeling and single deep learning(DL)models could not satisfy the performance requirements in partial scenarios with complex fluctuations.Therefore,an improved DL model based on wavelet decomposition(WD),the Convolutional Neural Networ k(CNN),and Long Short-Term Memory(LSTM)is proposed for day-ahead solar irradiance forecasting.Given the high dependency of solar irradiance on weather status,the proposed model is individually established under four general weather type(i.e.,sunny,cloudy,rainy and heavy rainy).For certain weather types,the raw solar irradiance sequence is decomposed into several subsequences via discrete wavelet transformation.Then each subsequence is fed into the CNN based local feature extractor to automatically learn the abstract feature representation from the raw subsequence data.Since the extracted features of each subsequence are also time series data,they are individually transported to LSTM to construct the subsequence forecasting model.In the end,the final solar irradiance forecasting results under certain weather types are obtained via the wavelet reconstruction of these forecasted subsequences.This case study further verifies the enhanced forecasting accuracy of our proposed method via a comparison with traditional and single DL models.