Multiple Energy Flow Modeling of Integrated Energy System Based on Heterogeneous Learner Integration Strategy

Abstract

To solve the problems of Newton’s method in the multiple energy flow (MEF) calculation of integrated energy systems (IES), such as the convergence solution depends on the selection of initial values, and the high dimension of Jacobi matrix leads to slow iterative calculation, a MEF modeling method for IES based on heterogeneous learner integration strategy is proposed. Firstly, considering the complex characteristics of the IES, the MEF model is trained using a variety of data-driven algorithms which are proven successful in related literatures. Secondly, based on the learner selecting strategy of ‘’accurate but different’’ and the quantitative indexes of accuracy and divergence of each model, partial least squares and deep neural network are selected as the basic learning algorithms to construct the heterogeneous learner integration model. Finally, the case study shows that the model established by the proposed method can achieve better accuracy than the model created by single algorithm. The model can calculate the energy flow of IES quickly and accurately without relying on the initial value and iteration, and the speed of solving this model is 47.8 times that of the traditional Newton’s method. The proposed method provides a new approach for the accurate and rapid calculation of MEF in a large-scale IES.