Load and renewable energy forecasting for a microgrid using persistence technique

• Main Authors: Dutta, Shreya, Li, Yanling, Venkataraman, Aditya, Costa, Luis M., Jiang, Tianxiang, Plana, Robert, Tordjman, Philippe, Choo, Fook Hoong, Foo, Chek Fok, Püttgen, Hans Björn
• Other Authors: Energy Research Institute @ NTU (ERI@N)
• Format: Article
• Language: English
• Published: 2018
• Subjects: Load Forecasting; Renewable Energy Forecasting; DRNTU::Engineering::Environmental engineering
• Online Access: https://hdl.handle.net/10356/80406; http://hdl.handle.net/10220/46510
• Institution: Nanyang Technological University

A microgrid system, be it connected to the utility grid or an independent system, usually consists of a mix of generation - renewable and non-renewable; loads - controllable or non-controllable and Energy Storage Systems (ESSs) such as batteries or flywheels. In order to determine how much power is utilized from the controllable resources such as ESS, diesel generators, micro-turbines or gas turbines, we first need to determine how much the demand is or how much the renewable energy sources are generating is which is accomplished using forecasting techniques. Due to the intermittent nature of renewable resources such as wind energy or solar energy, it is difficult to forecast wind power or solar power accurately. These forecasts are highly dependent on weather forecasts. It is evident that forecast of any data based on forecast of other parameters would lead to further inaccuracy, even if the relation between the inputs and output maybe predetermined through regression methods. Therefore, this paper illustrates an approach to use historical power data instead of numerical weather predictions to produce short-term forecast results. The concept is based on persistence method presented in [1]. This method uses the “today equals tomorrow” concept. From [2], we know that persistence technique produces results that are more accurate as compared to other forecasting techniques for a look-ahead time of 4-6 hours. Both [1] and [2] were based on wind power forecasting. In this paper, we investigate persistence method for short-term electrical demand, solar PV (Photovoltaic) power and wind power forecasting. Since the forecasts are dependent on historical averages of the data in the ‘near’ past, the accuracy is inversely proportional to the variation of power between the historical data and the actual data.