Regional Electricity Price Forecast In Interconnected Electricity Market

At present,in the major developed countries worldwide,electricity market has already got rid of monopoly and become a competitive market.Therefore,in many countries or regions,electricity is now being traded as a commodity in the market under market rules.Electricity price is the core element in an electricity market,accurate forecast of electricity price is of great significance to all participants in the market.As a commodity,however,electricity has some characteristics that many other commodities do not have.For instance,electricity is both economically and physically difficult to store and there must be a constant and instantaneous balance between the production and consumption in the power system.Moreover,power load and renewable energy production are highly dependent on weather conditions or people's daily activities.These unique characteristics make electricity price much more difficult to forecast accurately.Meanwhile,there are many factors that may affect the electricity price in an area,including historical electricity prices,loads,power reserve capacity,renewable energy generation and so on.More and more modern power systems have been physically interconnected with each other through transmission lines,making it almost impossible to neglect the adjacent regions when forecasting the electricity price of a local region.Information from neighboring regions is also exerting important influence on the local electricity price.This thesis firstly analyzes the basic theory of electricity price in electricity market,the mechanisms of electricity market clearing price as well the basic characteristics of electricity prices.Then,the thesis tests a variety of models for electricity price forecast and compares the performance of these models in predicting the next-day hourly electricity price in the Northern European electricity market.This paper also discusses the impacts of some ensemble methods on the performance of basic machine learning algorithms,including the impacts on prediction accuracy and computation time.In addition,since electricity price is affected by some periodic factors such as power load,this paper analyzes the time series stationarity and periodicity of electricity price by adjusting the input dataset.At the same time,this article discusses the influence of different feature types,regions and time gaps on the accuracy of electricity price forecast,and analyzes the importance of each feature type,region or time gap.Finally,according to the importance of various features for electricity price forecast,an optimal subset of features is selected to minimize the forecast error.