A novel heuristic method for wind farm power prediction: A case study M. Jabbari Ghadi ⇑ , S. Hakimi Gilani, H. Afrakhte, A. Baghramian Dept. of Electrical Engineering, Faculty of Engineering, University of Guilan, Rasht, Iran article info Article history: Received 30 July 2013 Received in revised form 23 June 2014 Accepted 6 July 2014 Available online 24 July 2014 Keywords: Imperialistic competitive algorithm – neural network Numerical weather predictions Wind farm Wind power prediction abstract Integration of wind power has a broad impact on power system operations, ranging from short-term sys- tem operations to long-term planning. The traditional deterministic unit commitment and economic dis- patch algorithms that power generation operators are currently using in the power system operations cannot capture uncertainties from wind power. To this end, an accurate wind farm power forecasting can highly support distribution and transmission system managers to improve power system manage- ment. In this research, authors present a novel hybrid method based on combination of imperialistic com- petitive algorithm (ICA) and artificial neural network (ANN) method to boost the short-term wind farm power prediction exactness using data from a numerical weather prediction (NWP) as well as measured data from an online SCADA. Besides, a very short-term generation power forecasting is implemented based on the values of wind speed and wind generation. An extensive comparative literature survey on presented methods in cases of short-term and very short-term is provided in this paper. At first step, considering environmental factors (i.e. geographical conditions, wind speed, humidity, temperature and other factors) a predictive model for wind speed forecasting is provided by the means of a multilayer per- ceptron (MLP) artificial neural network. At next step, ICA is employed in order to update weights of neural network. Validation of proposed method is confirmed using data of an actual wind farm. Ó 2014 Elsevier Ltd. All rights reserved. Introduction Kyoto Protocol with the aim of achieving the ‘‘stabilization of greenhouse gas concentrations in the atmosphere at a level that would prevent dangerous anthropogenic interference with the cli- mate system’’ is a socialism environmental covenant. By adopting this protocol, due to environmental advantages of renewable resources particularly wind power generation, utilization of these types of energy has acquired noticeable consideration in eye- catching number of countries, recently. In comparison with the environmental damages of traditional sources of energy, the eco- logical influences of wind power are proportionately negligible; in fact, wind power fuel usage is incomparable with fossil power plants as well as fuel emission. Furthermore, wind power enjoys infinitesimal progress expenses, in addition to an average cost of investigation. However, despite remarkable environmental bene- fits, the continuous and chaotic fluctuations of the wind speed make output power of wind farms completely stochastic and dif- ferent from those of conventional units. Whereas, instant electrical generation must be equal to the grid consumption to maintain net- work stability and reduce the spare capacity in the power system. Due to this fickleness, it may inject ample challenges to composi- tion of large quantities of wind power into a network system. How- ever, this challenge is not insuperable. In order to increase the economic efficiency and acceptability of the wind power, and to diminish punishments associated with participating in instanta- neous spot markets resulted from extra estimation or underrating of the production, the exact prediction of wind power as well as wind velocity is requested. Surely, a reliable prediction system can help distribution system operators and power marketers to make better decisions on critical situations such that guaranty power market revenues and decrease costs. Nowadays, several methods have been developed to predict wind power and speed. Existing methods can be arranged as statis- tical, physical and time series modeling methods, depending on the prediction model [1]. The physical methods are based on local meteorological service or NWP model data of the lower atmo- sphere (in relation with atmospheric pressure degree, at 2 m or 10 m heights above the ground) associated with topological data like obstacles, roughness and orography. The core idea of physical approaches is to estimate the generation power of the wind tur- bine with outsourcing these obtained data up or down to exact http://dx.doi.org/10.1016/j.ijepes.2014.07.008 0142-0615/Ó 2014 Elsevier Ltd. All rights reserved. ⇑ Corresponding author. Address: Dept. of Electrical Engineering, Faculty of Engineering, University of Guilan, Rasht, P.O. Box 3756, Iran. Tel.: +98 131 6690274; fax: +98 131 6690271. E-mail addresses: Ghadi.mjabbari@gmail.com (M.J. Ghadi), Saeedhakimi@msc. guilan.ac.ir (S.H. Gilani), Ho_afrakhte@guilan.ac.ir (H. Afrakhte), Alfred@guilan.ac.ir (A. Baghramian). Electrical Power and Energy Systems 63 (2014) 962–970 Contents lists available at ScienceDirect Electrical Power and Energy Systems journal homepage: www.elsevier.com/locate/ijepes