Electricity constitutes an indispensable source of secondary energy in modern society. Accurate and robust short-term load forecasting is essential for more effective scheduling of load generation, minimizing the gap between generation and demand, and reducing electricity waste. This study proposes a theory guided deep-learning load forecasting (TgDLF) framework to predict the future load through load ratio decomposition, in which dimensionless trends are obtained based on domain knowledge, and the local fluctuations are estimated via data-driven models. The historical load, weather forecast and calendar effect are considered in the model, and the modelâ€™s robustness to inaccurate weather forecast data is improved by adding synthetic disturbance during the training process. Experiments demonstrate that TgDLF is 23% more accurate than LSTM, and the TgDLF with enhanced robustness can effectively extract information from weather forecast data with up to 40% noise.
Keywords load forecast, domain knowledge, neural network, theory guided, physics informed