Three-level Planning and Scheduling of Comprehensive Energy Based on Random Forest R egression-Long- and Short-Term Memory Network Multivariate Load Forecasting

In allusion to the problem of high cost and low efficiency in power planning and dispatching due to load uncertainty in integrated energy system, a multivariate load forecasting-based three layer planning and dispatching model, in which the forecasting layer, the planning layer and the scheduling layer were included, was proposed. Based on random forest regression network and long- and short-term memory network a long-term and short-term multiple load prediction model was constructed. Taking the minimum integrated planning and dispatching cost and the minimum dispatching and operation cost as objectives, adopting improved particle swarm optimization algorithm and CPLEX solver, the optimal system comprehensive cost and allocation scheme was obtained. By means of planning and dispatching under different scenarios the equipment status and system cost were analyzed. Comparing the planning and dispatching results obtained by the constructed three layer model with that obtained by conventional two layer model, it is proved that the three layer planning and dispatching model possesses better economy and reliability.