LSTM (Long Short Term Memory) ultra-short-term photovoltaic power prediction method and system fused with time mode characteristics

The invention discloses an LSTM ultra-short-term photovoltaic power prediction method and system fused with time mode characteristics, and the method comprises the steps: carrying out the feature learning of photovoltaic power measured data and numerical weather forecast data after dual-branch input preprocessing through employing an improved LSTM model, and obtaining a feature sequence; and fusing the characteristic sequence of the photovoltaic power measured data and the characteristic sequence of the numerical weather forecast data, and outputting an ultra-short-term photovoltaic power prediction result through a full connection layer. According to the invention, two-channel feature learning is carried out on photovoltaic power measured data and numerical weather forecast data, a traditional LSTM method is improved, a first interaction link between current input and a previous training state is added, input quantity information income is enhanced, and the real-time performance of the system is improved. An