Time series analysis of Bahrain's first hybrid renewable energy system Mohamed Bin Shams * , Shaker Haji, Ali Salman, Hussain Abdali, Alaa Alsaffar Department of Chemical Engineering, University of Bahrain, P.O. Box 32038, Isa Town, Bahrain article info Article history: Received 29 January 2015 Received in revised form 16 February 2016 Accepted 23 February 2016 Keywords: Hybrid renewable energy system Time series analysis Forecasting PV (photovoltaic) module temperature Wind speed Solar irradiance abstract The performance of multisource renewable energy system depends strongly on the meteorological pa- rameters pertinent to the energy generating systems. Therefore, a method of modelling and forecasting meteorological and system parameters is necessary for efficient operation of the renewable energy power management system. Bahrain's first hybrid renewable energy system utilizes two renewable energy sources, namely solar irradiance through a 4.0 kW p PV (photovoltaic) panel and wind through a 1.7 kW p wind turbine. The focus of the present work is to investigate the proficiency of the BoxeJenkins based modelling approach in analysing and forecasting the daily averages of wind speed, solar irradiance, ambient air temperature, and the PV module temperature. Different non-seasonal ARIMA (Autore- gressive Integrated Moving Average) models have been constructed. ARIMA(1,0,0), ARIMA(1,0,0), ARIMA(0,1,2), and ARIMA(0,1,1) have been found adequate in capturing the auto-correlative structure of the daily averages of wind speed, solar irradiance, ambient air temperature, and PV module temperature, respectively. In addition, a functional relationship that correlates the diurnal PV module temperature to the ambient air temperature and solar irradiance have been developed. Residual and forecasting analyses have been used to ensure the adequacy of the identified models. © 2016 Elsevier Ltd. All rights reserved. 1. Introduction The prevalent use of renewable energy system is limited by the unpredicted nature of its driving forces [1]. Therefore, different attempts have appeared in literature to forecast meteorological parameters that are pertinent to renewable energy systems. Sug- gested approaches are either deterministic, stochastic, or a com- bination of both and they are aimed to provide short-term forecasting (hours), medium-term forecasting (daily) or long-term forecasting (yearly). Wind speed, solar irradiance, ambient tem- perature, and PV (photovoltaic) module temperature have been identified as important parameters in determining the overall performance of any hybrid renewable energy system that consists of PV panels and wind turbines. Kamal and Jafri investigated several ARMA (Autoregressive Moving Average) models to forecast wind speed in Quetta, Pakistan [2]. A good agreement between the forecasts and actual wind speeds has been shown within the 95% prediction intervals. In addition, several non-Gaussian distributions have also been suggested to model wind speed [2]. Nasir et al. investigated the suitability of Weibull and Rayleigh density func- tions to fit wind speed data in Quetta, Pakistan as well [3]. They found that Weibull density function outperforms the Rayleigh density in fitting actual data. Carta et al. reviewed in details a wide collection of probability density functions which have been pro- posed in literature for wind energy analysis [4]. Once the param- eters of an appropriate density function were identified, Monte Carlo simulation was used to generate forecasted values. A clear shortcoming of the probability density function modelling approach is its ignorance to the autocorrelation structure in the data. Different variations of time series modelling techniques have been used to assess energy-based systems. Javid and Qayyum used STSMs (Structural Time Series Models) to model the electricity demand function in Pakistan [5]. The structural form of the model allows the stochastic nature of the underlying energy demand trend as well as deterministic components, represented by the electricity consumption; to be model simultaneously with the real economy activity. Hu et al. utilized the concept of hybrid prediction to forecast wind speed in north-western China [6]. The hybrid model exploited the advantages of the EWT (Empirical Wavelet Transform), CSA (Coupled Simulated Annealing), and LSSVM (Least Square Support Vector Machine). CSA reduced the number of cost- function evaluations required by other search methods (e.g., neural * Corresponding author. Tel.: þ973 17 876050; fax: þ973 17 680935. E-mail address: mshams@uob.edu.bh (M. Bin Shams). Contents lists available at ScienceDirect Energy journal homepage: www.elsevier.com/locate/energy http://dx.doi.org/10.1016/j.energy.2016.02.136 0360-5442/© 2016 Elsevier Ltd. All rights reserved. Energy 103 (2016) 1e15