Photovoltaic-Power Prediction Model Based on Quantum Long Short-Term Memory Network

Abstract

Owing to the rapid development of new energy-generation systems，accurate photovoltaic （PV）-power forecasting is crucial in enhancing the grid’s ability to integrate solar energy. To address the insufficient accuracy of existing methods，this study proposes a quantum long short-term memory （LSTM） network PV-power forecasting model that is more lightweight in terms of parameters，more stable in training，and yields better results. First，data decomposition is performed based on a singular spectrum analysis. Subsequently，a quantum LSTM network is constructed to capture high-dimensional data features，followed by the utilization of dual attention mechanisms to capture features and temporal importance，which culminates in results output via a decision layer. Case studies show that compared with conventional methods，quantum PV-power forecasting can effectively improve the accuracy of such forecasts. Furthermore，empirical validation underscores the feasibility and effectiveness of utilizing quantum computers for PV-power forecasting.As quantum computers continue to develop，there is hope for the future application of these systems to achieve rapid and precise forecasting of power generation from large-scale photovoltaic （PV） power stations，This would assist in the safe scheduling and reliable operation of the power grid.