Um modelo de previsão de carga por barramento

AUTOR(ES)
DATA DE PUBLICAÇÃO

2004

RESUMO

In the operation of an electric power system, an important stage is the determination of the daily operation programme, which determines a plan of electric power production for the following day(s) for each of the generating units of the system, usually on an hourly or a half-hourly basis. This programme is used by the system s real time operation as an operational reference, and therefore it is important that the proposed solution should assure an appropriate operation of the system. To evaluate the impact of any given operation programme on the transmission system, the distribution of the load along the net must be known, since the loading in the transmission lines and transformers depends on the load demand in each bus (point of electric power delivery). In a daily operation planning context, it is necessary to know the load in each bus in each time interval considered in the programme. In other words, a short-term load forecast per bus is necessary. The main goal of this work was to develop a daily active load forecast model, on an hourly basis, per bus. Two types of methodologies were implemented: individual forecast methodology (MPI) that adresses each bus in an isolated way and aggregated forecast methodology (MPA) in which the forecast is made a single time for a given group of buses. The aggregate model seeks to reduce the need of forecasts being made for each bus, proposing for that a single forecast in an aggregated way. This model is composed of three stages: (i) aggregation phase - where the loads of the buses are joined; (ii) forecast phase - which is made through the aggregated series in (i); (iii) disaggregation phase - where the forecast is distributed through the joined buses. Two data c1ustering techniques were used for aggregating the buses. The individual forecast methodology considers the buses that present little similarity in their demand profile, when compared to the other buses, thus their aggregation results in high error rates. For the buses that present high similarity to other buses, the aggregated methodologies were efficient in the forecast, providing a smaller number of forecasts and results of good quality. The database used to test the models features real data measured in the transmission system and subtransmission of the Brazilian northeast region

ASSUNTO(S)

analise de series temporais - processamento de dados sistemas de energia eletrica - distribuição de carga algoritmos difusos redes neurais (computação)

Documentos Relacionados