Time series-based forecasting is one of the most popular machine learning methods due to its effectiveness in estimation of future values based on observations, interpolations, and interpretations of past data. Forecasting the load in power distribution networks is an essential step in energy trading and system operation. Machine learning, specifically time series-based forecasting, can be used in the precise prediction of energy consumption in power networks. Precise predictions have the potential to reduce operating and maintenance expenses, enhance the dependability of power supply and delivery systems, and enable informed decisions for future development endeavors. This chapter employs time series analysis to forecast energy usage in 10-minute intervals specifically for the city of Tétouan in Morocco by applying gradient boosting algorithm. Past and present data trends have been presented along with various accuracy parameters such as mean absolute error (MAE), mean squared error (MSE), and root mean squared error (RMSE).
TopIntroduction
Forecasting electricity demand is crucial for strategic planning and facility expansion within the electricity sector. Traditionally, load forecasting is done using statistical projections. With the advent of soft computing and machine intelligence techniques, new approaches have been suggested for the estimation of load. Specifically, significant advancements have been made in time series forecasting techniques, leading to enhanced accuracy. These algorithms have a wide range of applications due to their improved precision. Time series refers to a set of data points recorded at regular intervals of time. It encompasses different elements, including a trend that represents long-term patterns, cyclical patterns characterized by repeated up and down movements, seasonal variations that occur regularly within specific months or quarters, and irregular components which can’t be explained by known factors (Popeangă & Lungu, 2014). Forecasting involves making predictions about future values based on historical and current data in a time series. It plays a crucial rule in planning and controlling future power system operations. Load forecasting, specifically, is important for anticipating the demand for electricity. Load forecasting can be categorized into three types: short-term load forecasting (STLF), medium-term load forecasting (MTLF), and long-term forecasting (LTLF). STLF focuses on predicting load demand up to one day or at most one week ahead, MTLF covers the forecasting range of one day to several months, while LTLF looks further into the future, forecasting more than a year in advance (Pedregal & Trapero, 2010). STLF is utilized for scheduling electricity generation and transmission, MTLF aids in planning fuel purchases, and LTLF aims to develop and optimize the power supply and delivery system, which includes generation units, transmission networks, and distribution systems (Almeshaiei & Soltan, 2011).
Based upon the 2014 Census, Tétouan is a Moroccan city situated in the northern part of the country, overlooking the Mediterranean Sea. It spans over an area of 11,570 km2 and comprises of a population of approximately 380,000 individuals. The summer weather in Tétouan city is extremely hot and humid. According to a study as provided by the state-owned organization, known as the National Office of Electricity and Drinking Water (ONEE), Morocco’s energy production in 2019 was predominantly sourced from coal (38%), followed by hydroelectricity (16%), fuel oil (8%), natural gas (18%), wind (11%), solar (7%), and other sources (2%) (Morocco-Energy, 2019). Morocco is one of the developing nations that is dedicated to the growth, progress and enhancement of its economy.
Energy is a significant catalyst for economic growth, and one of its key sources is electricity, which is considered a crucial commodity. The improvement of living standards, economic development, and industrial progress are all closely linked to access to electricity. In fact, the presence or absence of electricity has become an economic benchmark that determines the advancement or backwardness of a country. Presently, electricity consumption is one of the most significant indicators for governments seeking development, and accurate long-term forecasting of electricity usage is vital for enhancing energy efficiency and avoiding costly errors (Elsaraiti, Ali, Musbah, Merabet, & Little, 2021). Accurate power forecasting is essential for effective planning of electricity consumption and for the implementation of decision support systems that guide the decision-making process within the power system (Nichiforov et al., 2017). Precise forecasting of electricity consumption also plays a critical role for policymakers in formulating effective electricity supply policies. However, the availability of limited data and variables often hampers the acquisition of valuable information necessary for achieving satisfactory prediction accuracy (Li & Zhang, 2018). Considering the significant reliance on non-renewable sources, accounting for 64% of the energy mix, forecasting energy consumption can be instrumental in assisting stakeholders in effectively managing procurement and inventory. Furthermore, Morocco aims to decrease its dependence on energy imports by expanding the production capacity of renewable sources. However, it is widely acknowledged that renewable sources such as wind and solar may not be consistently available throughout the year. Therefore, gaining a comprehensive understanding of the energy requirements of the country, starting with an intermediate-sized city, could serve as a progressive measure in strategizing the utilization of these resources.