In this paper we describe a neural network- based approach for automatic prioritization of objectives to solve the multi-objective economic dispatch (MOED) problem in the framework of smart microgrids. Four objectives are considered: energy cost, distance of supply, load balancing, and environmental impact. The proposed system tries to reproduce the preference function used by an expert to prioritize the objectives by assigning weights to the objectives themselves. To this aim, we use a multi-layer perceptron neural network whose inputs are four operating condition indicators sensed, with a regular time frequency, by the information network of the microgrid. Such indicators represent the current state of the microgrid. Learning has been performed by using a dataset composed of 150 samples, each one composed by a combination of the operating condition indicators, associated with a configuration of weights assigned to the objectives by an expert. Accuracies of 99.203% and 98.547% on the training and test sets, respectively, were achieved, with mean squared errors of 3.24 · 10^-4 and 6.59 · 10^-4 on the training and test sets, respectively.

Neural Network-Based Objectives Prioritization for Multi-Objective Economic Dispatch in Microgrids

LAZZERINI, BEATRICE;PISTOLESI, FRANCESCO
2014

Abstract

In this paper we describe a neural network- based approach for automatic prioritization of objectives to solve the multi-objective economic dispatch (MOED) problem in the framework of smart microgrids. Four objectives are considered: energy cost, distance of supply, load balancing, and environmental impact. The proposed system tries to reproduce the preference function used by an expert to prioritize the objectives by assigning weights to the objectives themselves. To this aim, we use a multi-layer perceptron neural network whose inputs are four operating condition indicators sensed, with a regular time frequency, by the information network of the microgrid. Such indicators represent the current state of the microgrid. Learning has been performed by using a dataset composed of 150 samples, each one composed by a combination of the operating condition indicators, associated with a configuration of weights assigned to the objectives by an expert. Accuracies of 99.203% and 98.547% on the training and test sets, respectively, were achieved, with mean squared errors of 3.24 · 10^-4 and 6.59 · 10^-4 on the training and test sets, respectively.
File in questo prodotto:
File Dimensione Formato  
SI2014.pdf

accesso aperto

Descrizione: Articolo principale
Tipologia: Documento in Post-print
Licenza: Tutti i diritti riservati (All rights reserved)
Dimensione 1.25 MB
Formato Adobe PDF
1.25 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11568/678324
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact