Solar Irradiance and Load Demand Forecasting based on Single Exponential Smoothing Method

Abstract

Forecasting of the solar irradiance and load demand are essential for system level control and components coordination in the supervisory controller of an off-grid hybrid energy system. This paper presents the analysis of the predictions for solar irradiance and load demand using two different Single Exponential Smoothing forecasting approaches. Both approaches perform prediction based on hourly basis. The first approach uses the current day data while the other uses the previous day data. Comparison between the two approaches is carried out, and the forecast results show that the Single Exponential Smoothing forecast models utilizing the previous day data achieved higher accuracy as compared to the one using the previous hour data.

Daily solar radiation forecasting using historical data and examining three methods

“…Of course, when talking about seasonal, day by day wouldn't be the option, even month to month comparison wouldn't show a suitable result, but still for this paper month by month and 6 months we calculated. The calculation of seasonal data for each selected month need to be started with the first averaging all the ratios for that particular season [7][8]. Seasonal approach: Second approach was seasonal forecasting, as long as the weather in would be almost the same as pervious years so using seasonal forecasting can be suitable.…”

“…Besides the two categories those were mentioned, there are other factors that will effect on forecasting such as time or duration which are called as time-series, causal, and judgmental [8]. A time-series model basically depends on the pervious recorded data to do the calculations for forecasting [11].…”

“…These techniques include regression models. Judgmental techniques are basically lies on probability and calculations so that all the time it might not going to predict the result with same value for similar situations [8].…”

“…Yet, results were surprising not as accurate as it was expected [7]. In this review paper, three most common used approaches for forecasting irradiation [7][8][9][10][11][12][13][14][15][16][17][18][19][20][21] were examined, which are exponential smoothing, seasonal method and neural network. For all these three same historical database were used and at the end comparison between them were designed.…”

“…That's why, accurate forecasting of radiation is one of the challenges for scientists, not just only to predict, for designing as well. For time being, there are many approaches for forecasting irradiance [5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21]. It has grown from mathematical formula and combination of them up to analyzing satellite images by using artificial intelligence.…”