Gün öncesi piyasasında saatlik ve günlük elektrik fiyatları tahmini

Abstract

Türkiye ekonomisinde, nakit giriş ve çıkışlarının büyük miktarlarda olduğu enerji sektörü, ciddi bir etkiye sahiptir. Yer altı kaynakları bakımından, kendi ihtiyacını karşılamaktan çok uzakta olan Türkiye, petrol ve doğal gaz gibi sanayide ve günlük yaşamda gereklilik haline gelmiş kaynaklara büyük bedeller ödeyerek ithal etmektedir. Bu büyük riski minimize edebilecek politikalar geliştirilmekte, özellikle elektrik üretiminde mevsimler arası farklılık gösteren arz güvenliği konusu, kaynak çeşitliliği ile daha güvenilir hale getirilmeye çalışılmaktadır.Elektrik piyasasında, yeni yapılanmasıyla birlikte düzenleyici ve denetleyici rol üstlenen kamu, enerji sektöründeki yatırımları da özel sektöre bırakmakta, hatta özelleştirmeler yoluyla EÜAŞ'ın pazar payı düşürülmektedir. Güncel haliyle elektrik piyasasında üreticiler, elektrik ticaretinde rekabet edebilmek için, yüksek verimli teknolojilere yönelmekte, tüketiciler ise, en uygun fiyattan elektrik ihtiyaçlarını karşılayabilmektedirler. Katılımcılar için piyasada oluşacak fiyatları tahmin etmek, maksimum karlılık adına stratejilerini ve piyasadaki konumlarını belirlemede büyük önem taşımaktadır.Elektrik piyasasında, ikili anlaşmalarla elektrik satışı yapan şirketler için fiyatların aylık veya dönemlik tahmin edilmesi, yapılacak anlaşmaların ticari değerini belirlerken, gün öncesi piyasada elektrik ticareti yapan piyasa oyuncuları için ise elektrik fiyatlarının tahmini, üretim planlarını belirlemekte, özellikle doğalgaz santralleri için verecekleri teklifler açısından büyük önem arz etmektedir. Hatta, enerji sektöründe santral yatırım planı olan firmaların yaptıkları fizibilite çalışmalarında, gelirlerini teşkil eden elektrik satış rakamları yeteri kadar öngörülerbilir olmak zorundadır. Tüm bunlar, yani gerçekçi fiyat tahminleri, uygulanabilir projelerin ülkeye kazandırılması ve nihai tüketicinin en ucuz elektriği tüketebilmesi anlamına gelmektedir. Ancak şunu da belirtmek gerekir ki, ülke ekonomisinin ve piyasanın dengeli halde olması, dışsal veya içsel sebeplerle dalgalanmaması, fiyatların tahmin edilebilir ve öngörülebilir olmasında önemli etkenlerden biridir. Bu çalışmada, gün öncesi piyasada oluşan elektrik fiyatları saatlik ve günlük ortalama şeklinde tahmin edilmeye çalışılmıştır. Çalışma boyunca 2010, 2011, 2012, 2013 yıllarına ait farklı parametre kümeleri kullanılmış, tahmin metodu olarak ise regresyon analizi ve yapay sinir ağları tercih edilmiştir. Sonuç olarak, yapılan her bir tahmin çalışmasında kullanılan parametrelerin ağırlıkları saptanmış, 2014 yılına ait saatlik ve günlük ortalama elektrik fiyatları tahmin edilmiştir. Oluşturulan tahmin modelleri ile elde edilen fiyatlar, gerçekleşen fiyatlarla kıyaslanmış, hataları ve başarıları ortaya konulmaya çalışmıştır. Gerçekleşen ve tahmin edilen fiyatlar arasındaki sapmalar, MAPE ve RMSE hata ölçütleri ve NASH başarı ölçütü ile değerlendirilmiştir.

Energy sector, which has parallels with the growth of countries, is strategic and dynamic sector in terms of investment volumes and contracts size. In Turkey as a developing country, energy sector has a significant impact on Turkish Economy as large cash inflows and outflows. Turkey, which is far from meeting its own underground sources need, imports petroleum and natural gas by paying high fees. Policies that can minimize this major risk are developed. Especially, security of supply, which varies between the seasons, is being tried to be safer in electricity generation.When we take a glance at change of Turkish Electricity Market, some topics come into prominence as positive developments such as performed deregulation works since 2001, private sector being encouraged to investment with legislations, liberalization process in electricity market, production without production license up to 1 MW. On the other hand, some topics emerges as a risk on security of supply such as the share of natural gas power plants in total production, transmission line constraints, price pressure on producers depending on decrease in demand.Considering historical data, it seems that development of electricity market depends on economy of the country. Fluctuation of electricity consumption in which industry has a large share can explain with growth and constriction of national economy. Especially in recent years, actual electricity consumption is lower than forecasted because of actual economic growth which is lower than expected.Otherwise, when we look at the Turkish Electricity Market figures, by the end of 2014, Turkey's installed power has reached 69.519,8 MW. 41.801,8 MW of them was formed by thermal power plants, 404,9 MW of them was formed by geothermal power plants, 23.643,2 MW of them was formed by hydropower plants, 3.628,7 of them was formed by wind power plant and 40,2 MW of was formed by solar power plants. Also, realized gross electricity demand was 257,2 GWh and peak power demand was 41.002,9 MW by the end of 2014. Furthermore, 252 GWh electricity generation were realized and 8 GWh was imported. Besides, 2,7 GWh of the total electricity supply was exported. Turkish Transmission Network, which is totally 53.408,7 km, comprise of 683 transmission substations, 1550 power transformers, 127.705 MVA transformer's power, 11 interconnection lines with contiguous countries. Voltage level of transmission system is normalized with 400 kV and 154 kV. Moreover, interconnection lines with Georgia and Bulgaria are 220 kV in accordance with the voltage level of these countries.When we look at the final Turkish Electricity Market structure, players divided into two side as private and public. Especially, public side is keeping their important role in the market as it should be despite all liberalization process. For instance, when required, installation and operation of power plants is entrusted to EÜAŞ in addition to power plants which are in operation and possession of EÜAŞ. On the other hand, TETAŞ was established in order to wholesale operations. It is responsible for not only realize power purchase guarantees which are depend on BO, BOT and TOR contracts but also meets their costs. Additionally, TEİAŞ has two major tasks. One of them is that when required, installation and operation of new power transmission lines as well as operation of existing power transmission lines. The other responsibility is that operate to market financial settlement system. Lastly, TEDAŞ is responsible for unlicensed production connection approval. Also, it is owner of distribution infrastructure which is operated by private sector players. Public enterprise, which is taking regulatory and supervisory role with new restructuring in electricity market, is leaving energy investments to private sector so much so that EÜAŞ's market share is being reduced through privatizations. Production companies must keep pace with the high efficient technology in order to compete in energy trading and consumers can provide electricity needs with affordable prices in the current electricity market structure. It is so crucial that market participants can identify their own strategy and market position with forecasting of the day-ahead electricity prices.Day-Ahead Market is an organized market which is operated by market operator and used on the purpose of electricity power trade and balancing in a day before the delivery time of the electricity. By this way, the market provide that the participants make purchases and sales of energy for the next day in addition to their bilateral agreements, the participants ensure compensation of contractual obligations and needs of production and consumption on a day in advance, the reference price of electricity is determined, balanced system is left to the system operator one day before, the system operator can also conduct congestion management for large-scale constraints.Balancing Power Market is an organized market that is operated by the system operator. In an attempt to balancing supply and demand in real time, the participants which are able to implement a power change of 10 MW in 15 minutes trade their reserve capacities in this market. The purpose of this market is to determine the marginal price of electricity and perform real time congestion management.Forecasted electricity prices on monthly and seasonal basis by companies selling electricity through bilateral agreements can valorize the commercial value of the contract. Furthermore, electricity prices which form a basis to revenue must be predictable on feasibility studies done by power plant investors. Realistic estimations mean that putting high-yield projects into practice on behalf of the country, and besides, consumption of best possible price of electricity for end users. However, it should also be noted that the country's economy and the electricity market are supposed to be stable and shall not fluctuate with external and internal reasons.The methods, which are used in electricity price forecasts, are proven with success and used in many commodity price forecasting. There are some kinds of forecasting methods which are collected under main headings such as multi-agent, fundamentals, reduced-form, statistical, computational intelligence, are using in these area. In some cases, a combination of multiple models can be used as a hybrid model. In the thesis, regression model as statistical model and feed-forward neural network model as computational intelligence model were used to forecast of electricity prices in hourly and daily average basis and tested to measure success of these models. Firstly, regardless of the model type, parameters which cause to formation of electricity prices and their weights which determine the effectiveness of the parameters are identified. Then, the electricity price, which is intended to forecast hourly and daily, is forecasted in consideration of parameters which are selected, and their weights which are calculated.In this study, day-ahead electricity market prices were forecasted as both hourly and daily average. During the study, different parameter sets of 2010, 2011, 2012 and 2013 were used and regression analysis and artificial neural networks were preferred to predictions. As a result, weights of parameters which were used for each estimation models were determined then forecasted daily-ahead electricity prices in 2014 as hourly and daily average. Forecasted prices which obtained by the estimation models were compared with the actual prices of 2014. Deviations between actual and forecasted prices were evaluated by using MAP, RMSE and NASE which are success and error criterions.