Implementation of advanced functionalities for Distribution Management Systems: Load forecasting and modeling through Artificial Neural Networks ensembles

• Published in: Electric power systems research Vol. 167; pp. 230 - 239
• Main Authors: Saviozzi, M., Massucco, S., Silvestro, F.
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01-02-2019; Elsevier Science Ltd
• Subjects: Algorithms; Alternative energy; Artificial Neural Networks; Clean energy; Distribution management; Distribution Management System; Electric fields; Electric power systems; Electricity consumption; Electricity distribution; Energy policy; Energy sources; Forecasting; Forecasts; Load forecasting; Load modeling; Management systems; Mathematical models; Modelling; Neural networks; Stress concentration; Uncertainty; Distribution Management System; Load forecasting; Artificial Neural Networks; Load modeling
• ISSN: 0378-7796; 1873-2046
• DOI: 10.1016/j.epsr.2018.10.036

•The paper presents the implementation in a real DMS of two advanced functionalities.•Load forecasting and modeling are based on ANNs ensemble.•A formal procedure for the ANNs parameters selection is presented.•Field results validate the proposed procedures.
Electric power systems are undergoing significant changes in all sectors at all voltage levels. The growing penetration of Renewable Energy Resources (RES), the liberalization of energy markets, the spread of active customers, the increasing diffusion of green energy policies to foster sustainable and low-emission policies, represent the main drivers in the evolution of the electric system. For these reasons, Distribution System Operators (DSO) are asked to adopt modern Distribution Management Systems (DMS) in order to manage RES uncertainties for an efficient, flexible and economic operation of distribution systems. In this context, the paper presents the design and the implementation in a real DMS of two advanced functionalities: load forecasting and load modeling. These two algorithms are based on an ensemble of Artificial Neural Networks (ANN). The good performances obtained on a real distribution network encourage the exploitation of the two proposed techniques to deal with demand uncertainties, in order to use efficiently the controllable resources and to face the stochastic behavior of RES.