Power forecasting for a photovoltaic system based on the multi-agent adaptive fuzzy neuronet

[摘要] This article presents a multi-agent adaptive fuzzy neuronet for a two days ahead forecasting of the hourly power from a photovoltaic system under random perturbations. In this research we consider a 5 KW Solar Power Plant for a residential building (model SA-5000M). The main objective of this research is to fulfil the multi-agent adaptive fuzzy neurone for hourly power forecasting for a photovoltaic system. The agents of the multi-agent adaptive fuzzy neuronet are fulfilled as two-layered recurrent networks. The standard Levenberg-Marquardt algorithm is described. The analysis of the evolving errors shows the potential of the multi-agent adaptive fuzzy neuronet in the hourly power forecasting for a photovoltaic system.

[发布机构] JSC Central Construction Bureau Geofizika, 89 Kirenskogo street, Krasnoyarsk, Russia^1;JSC Academician M F Reshetnev Information Satellite Systems, 52 Lenin street, Zheleznogorsk, Krasnoyarsk region; 662972, Russia^2;Katanov Khakass State University, 92, Lenina ave., Abakan; 655017, Russia^3;Reshetnev Siberian State University of Science and Technology, 31, Krasnoyarsky Rabochy Av., Krasnoyarsk; 660037, Russia^4

[效力级别] 工业技术

[关键词] Adaptive fuzzy;Levenberg-Marquardt algorithm;Multi agent;Photovoltaic systems;Power forecasting;Random perturbations;Recurrent networks;Residential building