Research On Short-term Forecasting Of New Energy Power Generation And Load In Microgrid

The intermittence and randomness of photovoltaic output and the fluctuation of power load of microgrid will have a great impact on the stable operation of microgrid system,which makes the energy exchange between energy supply and demand of microgrid extremely complex.Therefore,in order to ensure the safe and stable operation of microgrid and the normal production activities of users,it is of great practical significance to make accurate and reliable short-term prediction of photovoltaic output and load.In terms of short-term forecasts of photovoltaic output.First,effective analysis of photovoltaic output characteristics is carried out,and after feature engineering processing,the time characteristics,meteorological characteristics and historical photovoltaic output characteristics required for microgrid photovoltaic output prediction are obtained;secondly,according to the obtained characteristic attributes,combined with the width and depth model(Wide and Depth Model)Deep,Wide &Deep)and Long Short Term Memory(Long Short Term Memory,LSTM)have their respective advantages in data processing,and design a short-term prediction model for photovoltaic output based on Wide & Deep-LSTM;again,in order to optimize the model under different weather conditions K-means clustering algorithm is used to design the screening module of similar days to optimize the model training data.Finally,this paper uses the historical photovoltaic output and meteorological data of a photovoltaic power station to build a prediction model on the Python software platform.Comparing the prediction effects of BP,Wide & Deep and the prediction model of this paper under the weather type verifies the accuracy of the proposed prediction model.In the short-term forecast of microgrid load.In order to effectively mine historical data and improve the accuracy of short-term load forecasting,this paper firstly analyzes the periodicity and meteorological factors related to the power load of the microgrid to determine the input variable characteristics of the model.Then channel attention(CA)and temporal attention mechanism(TA)were added to the CNN and LSTM respectively,to build a CA-CNN module and a TA-LSTM module.Combining the characteristics of the two modules,the short-term load prediction model of the hierarchical attention mechanism of TCA-CNN-LSTM is constructed.Finally,the TCA-CNN-LSTM model based on the optimization of the attention mechanism is realized,and the electricity load forecast for four campus microgrids is carried out for a continuous week.At the same time,in order to reflect the superiority of the model proposed in this article,the model in the article is compared with the CNN-LSTM and the CNN-LSTM model optimized based on a single attention mechanism.The experimental verification shows that the method in this article has certain application prospects...