Short-term load forecasting for microgrids based on DA-SVM

• Anan Zhang ^[1] ; Pengxiang Zhang ^[2] ; Yating Feng ^[1]
  1. [1] Southwest Petroleum University

     Southwest Petroleum University

     China

     
  2. [2] National Engineering Laboratory for Big Data Application on Improving Government Governance Capabilities, China
• Localización: Compel: International journal for computation and mathematics in electrical and electronic engineering, ISSN 0332-1649, Vol. 38, Nº 1, 2019, págs. 68-80
• Idioma: inglés
• Enlaces
  • Texto completo
• Resumen

  • Purpose – The study aims to accomplish the short-term load forecasting for microgrids. Short-term load forecasting is a vital component of economic dispatch in microgrids, and the forecasting error directly affects the economic efficiency of operation. To some extent, short-term load forecasting is more difficult in microgrids than in macrogrids.

    Design/methodology/approach – This paper presents the method of Dragonfly Algorithm-based support vector machine (DA-SVM) to forecast the short-term load in microgrids. This method adopts the combination of penalty factor C and kernel parameters of SVM which needs to be optimized as the position of dragonfly to find the solution. It takes the forecast accuracy calculated by SVM as the current fitness value of dragonfly and the optimal position of dragonfly obtained through iteration is considered as the optimal combination of parameters C and s of SVM.

    Findings – DA-SVM algorithm was used to do short-term load forecast in the microgrid of an offshore oilfield group in the Bohai Sea, China and the forecasting results were compared with those of PSO-SVM, GA-SVM and BP neural network models. The experimental results indicate that the DA-SVM algorithm has better global searching ability. In the case of study, the root mean square errors of DA-SVA are about 1.5 per cent and its computation time is saved about 50 per cent.

    Originality/value – The DA-SVM model presented in this paper provides an efficient and effective method of short-term load forecasting for a microgrid electric power system.