Abstract— The increasing numbers of international trading ports around the world are facing significant energy and environmental challenges such as rising energy consumption and greenhouse emissions. To understand the energy demand behaviour of ports or cranes, several simulation studies have been carried out using data from the Port of Felixstowe in the UK. The aim of this paper is to propose a 24-hours active power forecast model and analysis tools for a single electrified RTG crane. This model could be a potential solution to these energy consumption and management problems. The crane data has been collected for 30 days and analysed in terms of the daily demand usage, the number of crane moves and the weight of containers. Two different forecast methods, ARIMAX and Artificial Neural Network have been used to forecast highly stochastic, non-smooth and very volatile active crane power demand. The results indicate that the ANN forecast model is more accurate according to the mean absolute percentage error (MAPE) results during the testing period. Keywords---ARIMAX; Neural Network; load forecasting; Rubber Tyred Gantry Cranes. I. INTRODUCTION The electrical energy demand in ports has been rising as a result of the shift from diesel RTG cranes to electrified RTG cranes, which are connected to low and medium voltage grids, to reduce gas emissions and fuel consumption. According to [1] and [2] the volume of container traffic in the United Kingdom may increase by around 54% in the next 15 years. Furthermore, this may increase the total power consumption and peak demand at port substations. Ports may need a new substation or to upgrade the port infrastructure to cover this rise in demand. However, there is a lack of understating of the energy demand behaviour at port substations and electrified RTG cranes. This understating, is vital for developing strategies and solutions to reduce the environmental effects of emissions and peak demand problems [3] [4]. Electrical load forecasting is key for developing an efficient energy management system and accurate load forecast models are required for energy planning and substation operations. Load forecasting provides the necessary information for making decisions on generating power, load shifting and electrical infrastructure development. Generally, electric load forecasting with lead time can be divided into four categories: Manuscript received November ,6 2016. This work was supported by the Port of Felixstowe and they are to be acknowledged for their aid in this research, particularly for the specific data on the eRTGs and substations mentioned in this paper. F.A authors is with the University of Reading, Reading, UK TABLE I: THE LOAD FORECASTING CATEGORIES Load forecast type Prediction target Very short term Few minutes - one hour Short term One hour - several days Medium term One week - one year Long term One year or more Short-term load forecasting deals with intervals of one hour to several weeks [5]. The accuracy of short-term load forecasting has a significant effect on the operating efficiency of any utility, especially for the interval forecast time from one hour to one week. Some utility decisions such as the scheduling of power generation and scheduling of energy purchases are based on the short-term forecast results [4][5]. A large variety of statistical methods and intelligence techniques have been used in short-term forecasting. The statistical methods developed built on input data have a specific structure such as being based on seasonal trends or autocorrelation patterns. The methods below are commonly used in time series techniques [6]:  ARMA: Autoregressive moving average.  ARIMA: Autoregressive integrated moving average.  ARIMAX: ARIMA with exogenous variables. Fuzzy Inference Systems, expert systems and Artificial Neural Networks (ANN) are intelligence methods that can be used for short-term forecasting. Many techniques have been developed for forecasting electrical energy demand and several research studies in the literature have used ANNs and ARIMAX. However, very few studies are directly related to investigating electrified RTG cranes or substation port loads [5][6][7][9]. As shown in figure 1 the stochastic and volatile nature of the active power demand of cranes creates a significant challenge in predicating or forecasting this demand. A novel short-term load forecast models using ARIMAX and ANN have been developed in this paper. The aim is to generate an hourly active power forecast for 24 hours ahead for a single RTG crane. II. ANALYSIS OF RTG CRANE LOAD DEMAND A. Data In this paper, the electrified RTG crane data was collected for 30 days from the 11 th of April 2016 to the 11 th of May 2016 from the Port of Felixstowe in the UK. The record of data includes hourly three-phase active power; number of crane moves and gross container weights. The consumption of three-phase active Analysis of RTG Crane Load Demand and Short-term Load Forecasting Feras Alasali 1 , Stephen Haben 2 , Victor Becerra 3 , William Holderbaum 4 Int'l Journal of Computing, Communications & Instrumentation Engg. (IJCCIE) Vol. 3, Issue 2 (2016) ISSN 2349-1469 EISSN 2349-1477 https://doi.org/10.15242/IJCCIE.EAP1216012 448