Based on the rapid expansion of the production scale of the photovoltaic industry, this paper discusses the lack of accuracy in photovoltaic power prediction, the lack of tracking efficiency of maximum power points, and the backwardness of operation and maintenance technology of photovoltaic power plants. Combined with the artificial neural network (ANN) model, meta-heuristic algorithm and photovoltaic health state architecture technology, the internal mechanism of digital-real integration driving the high-quality development of the photovoltaic industry is deeply discussed. Multilayer Perceptron (MLP), Recurrent Neural Network (RNN), Long Short-Term Memory Neural Network (LSTM), Gated Recurrent Unit Network (GRU) and Convolutional Neural Network (CNN) were used to optimize the existing technology and improve the accuracy of PV power prediction. The meta-heuristic algorithm is introduced to help solve the problems of fast response, continuous oscillation at the maximum power point and easy locking of local peak points, and improve the tracking efficiency of the maximum power point. Data Quality Routines (DQRs), digital twins, and AI-driven fault diagnosis algorithms are used to solve the problem of lack of accurate, general, and location -independent data-driven PV diagnostic algorithms in the field of PV system fault detection, and improve the operation and maintenance procedures of large-scale PV power plants.