Short-term Power Load Forecasting Based on WNR-LSTM — Take the Singapore Region as an Example

In recent years, as the electrical energy of distributed storage has gradually increased, randomness load demand increased. This makes it more difficult to rationally dispatch and store loads. How quickly accurately dig out effective information objective laws from massive power data, effectively clarify instability timing changes, then reduce consumption accidents in dispatching is great significance. paper uses wavelet analysis stochastic sparrow search algorithm optimize LSTM neural networks construct long- short-term forecasting models, Singapore December 2021 Empirical studies were conducted on electricity 24 January 23, 2022, well corresponding peak-to-valley prices, meteorological data other related data. The results show that long-term prediction model based WNR-SSA-LSTM compared with traditional terms networks, RMSE MAPE reduced by 4.000% 1.7871%, goodness-of-fit R2 increased 0.0196, It suitable for forecasting.