Multi-Input Deep Convolutional Neural Network Model for Short-Term Power Prediction of Photovoltaics

Abstract

Along with the increasing prominence of energy and environmental issues, solar energy has received more and more extensive attention from countries around the world, and the installed capacity of photovoltaic power generation, as one of the main forms of solar energy development, has developed rapidly. Solar energy is by far the largest available source of energy on Earth, the use of solar power photovoltaic system has the advantages of flexible installation, simple maintenance, environmentally friendly, etc., by the world’s attention, especially the grid-connected photovoltaic power generation system has been rapid development. However, photovoltaic power generation itself is intermittent, affected by irradiance and other meteorological factors very drastically, and its own randomness and uncertainty are very large, and its grid connection affects the stability of the entire power grid. Therefore, the short-term prediction of photovoltaic power generation has important practical significance and guiding meaning. Multi-input deep convolutional neural networks belong to deep learning architectures, which use local connectivity, weight sharing, and subpolling operations, making it possible to reduce the number of weight parameters that need to be trained so that convolutional neural networks can perform well even with a large number of layers. In this paper, we propose a multi-input deep convolutional neural network model for PV short-term power prediction, which provides a short-term accurate prediction of PV power system output power, which is beneficial for the power system dispatching department to coordinate the cooperation between conventional power sources and PV power generation and reasonably adjust the dispatching plan, thus effectively mitigating the adverse effects of PV power system access on the power grid. Therefore, the accurate and reasonable prediction of PV power generation power is of great significance for the safe dispatch of power grid, maintaining the stable operation of power grid, and improving the utilization rate of PV power plants.

PV Power Forecasting in the Hexi Region of Gansu Province Based on AP Clustering and LSTNet

“…Machine learning methods mainly include support vector machine [5], random forest [6], extreme learning machine, neural network [7], and so on. In recent years, deep neural networks such as convolutional neural networks (CNN) [8,9] and long short-term memory (LSTM) networks [10,11] have also been introduced to improve the ftting ability of models. Combining the advantages of diferent neural networks, hybrid network prediction models such as CNN-LSTM [12,13] and recurrent neural network (RNN)-LSTM [14,15] further improve the prediction accuracy of the model by capturing the time series and spatial correlation between PV power generation power sequence and related infuencing factors.…”