A DNI Prediction Method for Solar Thermal Power Plant based on SCSSA-CNN-BiLSTM Model

Solar thermal power plants rely on Direct Normal Irradiance (DNI) for energy conversion to complete the power generation process, and the sudden change in DNI itself can affect their reliability and power generation efficiency. To better operate and regulate solar thermal power plants, a DNI prediction method with improved accuracy and stability is proposed based on SCSSA-CNN-BiLSTM. A new hybrid neural network is formed by combining a convolutional neural network (CNN), a bi-directional long short-term memory (BiLSTM), and a sine-cosine and Cauchy mutation sparrow search algorithm (SCSSA). The research object is introduced as the real engineering data of a solar thermal power plant in Qinghai Province. Through evaluation of the prediction results and corresponding performance evaluation indicators, the proposed prediction model is shown to accurately predict DNI variation trends. The DNI prediction method has the potential of enhancing the operation flexibility of solar thermal power plants by maximizing its solar potential.