Abstract
Forecasting electricity demand and consumption accurately is critical to the optimal and costeffective operation system, providing a competitive advantage to companies. In working with seasonal data and external variables, the traditional time-series forecasting methods cannot be applied to electricity consumption data. In energy planning for a generating company, accurate power forecasting for the electrical consumption prediction, as a technique, to understand and predict the market electricity demand is of paramount importance. Their power production can be adjusted accordingly in a deregulated market. As data type is seasonal, Persistence Models (Naïve Models), Seasonal AutoRegressive Integrated Moving Averages with eXogenous regressors (SARIMAX), and Univariate Long-Short Term Memory Neural Network (LSTM) is used to explicitly deal with seasonality as a class of time-series forecasting models. The main purpose of this project is to perform exploratory data analysis of the Spain power, then use different forecasting models to once-daily predict the next 24 hours of energy demand and daily peak demand. To split the electricity consumption data from 2015 to 2018 into training and test sets, the first three years from 2015 and 2017 were used as the training set, while values from 2018 were used as the test set. The obtained results showed that the machine learning algorithms proposed in the recent literature outperformed the tested algorithms. Models are evaluated using root mean squared error (RMSE) to be directly comparable to energy readings in the data. RMSE has calculated two ways. First to represent the error of predicting each hour at a time (i.e. one error per-hourly slice). Second to represent the models’ overall performance. The results show that electricity demand can be modeled using machine learning algorithms, deploying renewable energy, planning for high/low load days, and reducing wastage from polluting on reserve standby generation, detecting abnormalities in consumption trends, and quantifying energy and cost-saving measures.
Publication Date
12-23-2021
Document Type
Master's Project
Student Type
Graduate
Degree Name
Professional Studies (MS)
Department, Program, or Center
Graduate Programs & Research (Dubai)
Advisor
Sanjay Modak
Advisor/Committee Member
Ioannis Karamitsos
Recommended Citation
Mohammadigohari, Mahdi, "Energy consumption forecasting using machine learning" (2021). Thesis. Rochester Institute of Technology. Accessed from
https://repository.rit.edu/theses/10989
Campus
RIT Dubai
