Polynomial-Fourier series model for analyzing and predicting electricity consumption in buildings

Water and electricity are two essential elements that need to be used effectively and efficiently and should not be wasted. Various methods have been adopted to manage the amount of energy being used everywhere. Accurate prediction of energy consumption in buildings is considered in this research paper. The electrical usages in buildings vary depending on the activity of the user.

Global climate change or ageing of equipment are some factors that affect energy consumption. The construction models find it difficult to accurately predict the consumption of electricity in buildings as energy consumption varies in terms of periodic and irregular activities.

To estimate the electricity consumption and its performance in buildings, DOE-2 based building energy simulation program has been proposed by this team following the earlier researchers. For accessing a building’s thermal response, sensitivity techniques have to be used. This helped energy professionals to detect changes in electrical consumption rapidly. To predict regional energy consumption Artificial Neural Network (ANN) model was proposed which provided accurate forecasts. They also used applied Adaptive Network based Fuzzy Interference System (ANFIS). This also provided them accurate forecasts.

Interestingly Fourier series model was used to analyze structural vibration and to investigate the periodic demand of electric demand series which provided only satisfactory results. To characterize regular electricity demand, Gaussian processes are found to be useful and to predict uniform electricity consumption Fourier transforms is better suited.

Researchers led by Professor Chin-Shiang Chang at National Yunlin University of Science & Technology, Taiwan proposed another application to analyze electricity consumption, a combination of polynomial and Fourier series known as Polynomial Fourier series is developed (P-FS). Here consumption is mostly determined by periodical activities and long term trends (eg.  university libraries). Initially, raw data is collected for analysis. The new findings appear in the journal, Energy and Buildings.

In their study the data set was separated into two parts; one to determine the parameters of the model and the other to determine the number of terms for the polynomial. The real data is compared with the predictions based on Polynomial Fourier series. If more terms are selected in the polynomial the better will be the fitting that appears for the first part of analyzed data. Errors are taken and expressed as Fourier series of sinusoid. The sinusoid term with maximum amplitude is picked but the condition is that it should fit with the data of the second part. The more the sinusoidal terms better will be the fitting. The sinusoids represent the consumption of electricity due to different usage levels.

The main advantage of the Polynomial Fourier Series proposed in this study is better evaluation of energy policy and accurate predictions of electricity consumption even with limited data compared to other models.