A novel adaptive day-ahead load forecast method, incorporating non-metered distributed generation: a comparison of selected European countries

Sinkovics, Bálint ✉ [Sinkovics, Bálint (villamosmérnöki t...), szerző] Villamos Energetika Tanszék (BME / VIK); Táczi, István [Táczi, István (Villamos Energetika), szerző] Villamos Energetika Tanszék (BME / VIK); Vokony, István [Vokony, István (Villamos energetika), szerző] Villamos Energetika Tanszék (BME / VIK); Hartmann, Bálint [Hartmann, Bálint (Villamos energetika), szerző] Villamos Energetika Tanszék (BME / VIK)

Angol nyelvű Tudományos Könyvfejezet (Könyvrészlet)
    Azonosítók
    • MTMT: 31857114
    Chapter 3, authored by B. Sinkovics, I. Taczi, I. Vokony, and B. Hartmann, provides a novel adaptive day-ahead load forecasting method, incorporating nonmetered distributed generation. The method provides a comparison of selected European countries. This chapter reveals, similar to Chapter 1, the need for tailored solutions for each application, due to the considerably varying conditions of each country. Long short-term memory networks are a subtype of recurrent neural networks using a unique node architecture called cells. This is one of the most widely used state-of-the-art models for predictions, where temporal dependencies are important. The chapter makes a hypothesis that the proliferation of nonmetered residential photovoltaic plants significantly affects the accuracy of load forecasts. The introduced long short-term memory load predictor was applied to transmission system operator frequency control areas with various load scales and renewable penetration. The developed model has outperformed national forecasts and has identified potential effects of nonmetered distributed generation. This comparison reinforced the hypothesis of the authors that residential photovoltaic entities play an important role in the characteristics of the daily load curves, and the performance of load forecasts can be improved by taking such data into consideration, when developing forecasting algorithms.
    Hivatkozás stílusok: IEEEACMAPAChicagoHarvardCSLMásolásNyomtatás
    2021-10-19 01:43