Polynomial-dynamic electric arc furnace model combined with ANN

Two approaches exist for modeling electric arc furnace, the largest electrical load. Distinct mathematical equations are used as the first approach. However, the process is nonstationary and the equation is limited to 1 form and unable to extract all nonlinear components. Black-box models such as artificial neural networks are used as the second approach. However, using a black-box model for a process with partially known equations is not efficient, as it cannot benefit from known knowledge of the process. This paper proposes a combination of the 2 mentioned approaches. The first phase consists of nonlinear and dynamic terms. The appropriate model orders relate to dynamic and nonlinear terms are attained. In the second phase, artificial neural network is used to model the rest of the nonlinear components, which are not captured in the first phase. Numerous actual records are used to compare the proposed model with several electric arc furnace models.