Short-Term Power Forecasting Model for Photovoltaic Plants Based on Historical Similarity

Energies

Claudio Monteiro²  Tiago Santos²  L. Alfredo Fernandez-Jimenez³  Ignacio J. Ramirez-Rosado¹ 

[1] Electrical Engineering Department, University of Zaragoza, Maria de Luna 3, Zaragoza 50018, Spain;Faculty of Engineering, University of Porto, Dr. Roberto Frias, Porto s/n 4200-465, Portugal; E-Mails:;Electrical Engineering Department, University of LaRioja, Luis de Ulloa 20, Logroño 26004, Spain; E-Mails:

关键词: power forecasting;    solar energy;    data mining;    genetic algorithm;   

DOI  :  10.3390/en6052624

来源: mdpi

【 摘 要 】

This paper proposes a new model for short-term forecasting of electric energy production in a photovoltaic (PV) plant. The model is called HIstorical SImilar MIning (HISIMI) model; its final structure is optimized by using a genetic algorithm, based on data mining techniques applied to historical cases composed by past forecasted values of weather variables, obtained from numerical tools for weather prediction, and by past production of electric power in a PV plant. The HISIMI model is able to supply spot values of power forecasts, and also the uncertainty, or probabilities, associated with those spot values, providing new useful information to users with respect to traditional forecasting models for PV plants. Such probabilities enable analysis and evaluation of risk associated with those spot forecasts, for example, in offers of energy sale for electricity markets. The results of spot forecasting of an illustrative example obtained with the HISIMI model for a real-life grid-connected PV plant, which shows high intra-hour variability of its actual power output, with forecasting horizons covering the following day, have improved those obtained with other two power spot forecasting models, which are a persistence model and an artificial neural network model.

【 授权许可】

CC BY   
© 2013 by the authors; licensee MDPI, Basel, Switzerland.