Due to climate change, growing energy prices, and increasing energy consumption, energy efficiency has become a key topic in recent years. Most energy market traders also want to be able to foresee the energy market in the future so that they can take the appropriate actions to optimize their trading profits. As a result, energy market evaluation models are required. Energy markets, on the other hand, are location-dependent, as each market has its auctions and procedures. As a result, specific models for each energy market should be developed. The primary aim of this study is to provide a comprehensive comparison of various machine learning methods in the Turkish electricity market. A comparative analysis is provided on support vector machines (SVM)-based methods, k-nearest neighbors (KNN)-based methods, and ensemble-based method to analyze system marginal price (SMP). According to the accuracy value, the ensemble-based method gives better results.
TopIntroduction
Today's consumers have important rights that depend on electricity, including the ability to heat their homes, light their homes, travel, and communicate (Gul, 2008). Therefore, countries must deliver electricity to end-users in a timely, reliable, and economical manner. Although it helps countries' economies thrive, the availability of energy also imposes significant financial responsibilities. Countries began allowing investors to participate in the process with a win-win strategy since they realized they could not handle these enormous loads alone. Liberalization is changing markets for energy in particular by reducing public spending and increasing private sector investments. Turkey's electrical market has adopted this pattern (Ünal, Onaygil, Acuner, & Cin, 2022). The establishment of a system similar to the stock market for the electricity market provided great convenience for the participants, and trade started to be carried out in a structure similar to the pool system. Thus, countries started to create their energy markets and had the opportunity to trade in these environments (Arslan & Ertuğrul, 2022). To some extent, it behaves like a stock market price; its changes are cyclical, variable, and affected by a wide range of events (Wang & Ramsay, 1997).
In Turkey, there are three different electricity markets (EPIAS, 2022a). These include the day-ahead market, the intra-day market, and the balancing market. Balancing power market provides reserve capacity that can be online within 15 minutes at the latest to sustain real-time balancing for system operators. Frequency control and demand control services are ensured through ancillary services (EPIAS, 2022a). Real-time supply and demand balance is achieved through the balancing power market (Sirin & Yilmaz, 2021). Although the electricity markets were brought to balance one day in advance through the markets operated by EPIAŞ, the demand and supply balance may change as the real-time approaches. Because of these changes, deterioration may occur in the system that was balanced the day before. The price of the bid, which corresponds to the net volume of orders issued in the balancing power market, determines the SMP (Dursun, Eke, & Tezcan, 2020).
In the previous vertically integrated industry, lowering production and operating costs was the main objective of power system planning and management. However, since the introduction of competition in the electricity market, the objective has changed to maximize the profit or return to the market participants. Employing strategies to increase profit may become possible. The SMP is the market price that is set after taking the characteristics of the generators into account during the bidding process, given the demand and condition of the power system (Lee, Park, Shin, & Lee, 2005). SMP prediction is a significant issue for both consumers and producers of electrical power. Making SMP predictions is a difficult challenge since SMP is very volatile and dependent on a wide range of variables, including the price of oil, load demand, and even natural parameters. By offering suitable bidding strategies, a good SMP prediction can assist both the consumer and the electric power company in maximizing their profit (Yudantaka, Kim, & Song, 2019).
In the competitive energy market, SMP prediction is a significant source of information. The significance of electricity price prediction models like SMP has expanded with the development of a competitive energy market. However, compared to other traded commodities, electricity is well known for being more volatile, making it more difficult to predict the SMP. Because the price of primary energy sources is a major determinant of generation costs, the SMP is also influenced by these costs, which change over time (Jufri, Oh, & Jung, 2019).