Artificial neural network-based distribution substation and feeder load forecast

Abstract

Artificial Neural Networks (ANNs) have been successfully applied to the problem of forecasting future load values, especially in the short term framework (a few minutes to a few hours ahead). Traditional analytical models have shown difficulties when dealing with (i) the highly variable demand curve shapes, (ii) some independent variables that exhibit random behaviour, and (iii) the identification of variables that could explain relevant load variations, such as weather variables. Current available ANN applications to this problem are by far aimed at a systemwide level, where the load behaviour is more regular than at substation or even primary feeder levels.

Assessment of the allowable short-term overloads in power transformers including condition management

“…This procedure aims at supplying more accurate conditions of the loading behaviour, every 15 minutes, for five future intervals -¼ hour, ½ hour, 1 hour, 2 hours and 4 hours ahead. This module has been extensively studied [1]. Figure 2 shows one of such forecast.…”

Adjustment of discrete load changes in feeder databases for improving medium‐term demand forecasting

“…In this reference it is shown that data shifts, due to feeders switching, contaminates registers and reduces forecasting accuracy in the medium and long term. In addition, feeders' reconfiguration has been recognised as an important element for estimating future demand values in the shortterm (24 h ahead) [15]. In that reference the authors conclude that switching operation in primary networks has a substantial impact on load profiles, hence, producing errors in future demand forecasting.…”

A New Method for Substation Planning Problem Based on Weighted K-Means

“…In [2], a branch-and-bound model is proposed for the selection of the substation location. In general, different mathematical programming formulations and techniques have been proposed and applied to the location problem of electrical distribution substations: simulating annealing [4,5], genetic algorithms [6,7], evolution strategies [8,9], particle swarm optimization [10], bacterial colony chemotaxis [11,12], evolutionary programming [13] and neural network [14]. In general, different mathematical programming formulations and techniques have been proposed and applied to the location problem of electrical distribution substations: simulating annealing [4,5], genetic algorithms [6,7], evolution strategies [8,9], particle swarm optimization [10], bacterial colony chemotaxis [11,12], evolutionary programming [13] and neural network [14].…”