Energy scheduling for a microgrid connected to the main grid based on real time electricity market

Studies on renewable energy sources and their utilization are dramatically increasing in the last couple of years, together with the increasing need for energy. There are two main reasons for that. Firstly, mitigation of greenhouse gases depends onreducing the use of fossil fuels on a large scale and the use of renewable energy sources is the only alternative to prevail. Secondly,energy dependence is one of the most crucial economic and strategic matter for all countries. For example, Turkey is an energy importing country and an important part of its current deficit consists of energy importation costs.For such countries, increasing the national renewable energy generation and using it in the most effective way, is vital. Unfortunately, renewable sources are yet to totally replace fossil fuels. The main problem with renewable sources is the uncertainty they contain which results in discontinuities and fluctuations of power supply. For dealing with these uncertainties and using the limited renewable energy generation in the most effective way, energy storage systems are suggested together with using hybrid energy sources. Energy storage can be briefly explained as saving the produced energy for later usage at desired time and place. Storage systems prevent waste of energy at moments, when the time of energy production does not match the time of energy demand. In the recent years, thanks to improving technology, efficiencies of these systems have increased, while the prices decreased. For that reason, these systems are having much more attention these days. In this study, a microgrid with hybrid renewable energy sources and an integrated storage unit is investigated. Renewable sources consist of wind turbine and photovoltaic panels and the storage unit is chosen to be reused lithium-ion batteries. Lithium batteries that are used in the electrical vehicles are usually replaced at the energy capacity level of 85% due to performance drops. However for storage purposes these levels are quite satisfactory. Therefore reusing these types of batteries is logical both economically and environmentally.The microgrid is also connected to the main grid. So, electricity can be purchased and sold at the same time. The sales of electricity are done to the real time market (gün içi piyasası) with the hourly prices. Hourly energy demand has to be satisfiedat all times through renewable energy generation, main grid or the battery. Main objective of the study is to create a mathematical model for this system and optimize it in order to generate an hourly energy schedule which minimizes the costs. This mathematical model can be used as a decision support system for the energy planning. The result of the optimization model will give the hourly amounts of energy generations for wind turbine and photovoltaic panels, charging and discharging of the battery, the purchases made from the grid and sales made to the electricity market. There are many advantages of the suggested microgrid becoming prevalent. Utilizing the renewable energy sources in the most effective way, environmentally friendly energy usage will increase and the energy dependency of the country will decrease. Energy costs and dependency will be minimized for the stakeholders. By using multiple energy sources and energy storage unit, the stakeholders will be less susceptiple to the electricity black outs. With the use of energy storage unit, waste energy levels and the uncertainties of renewable energy sources will be decreased. Especially at peak times, the amount of energy requested from the main grid will be diminished. This is called peak shaving and has a great influence on main grid reliability and stability. Also by using reused lithium batteries, these imported products will be better utilized before going to waste. This is both enviromentally and economically beneficial. Proposed sytem can also be applied in places like hospitals, university campuses, district areas or even residences. The capacities of the energy generation sources can be scaled down according to the usage area. There are uncertainties at some parameters of the proposed model. These paremeters are the maximum levels of production from each renewable energy sources, energy demand levels and the market prices for the electricity sales to the grid. Case studies were done for an island in Gebze Organized Industrial Zone (Gebze Organize Sanayi Bölgesi – GOSB). Therefore, the electricity purchase prices were used according to the special tariff for the industrial zone. There are three different prices in a day according to the hour. The day is seperated into three categories as day, night and peak load times. As might be expected, the highest price is for the peak load hours and the lowest price is for the night hours. A forecasting algorithm is developed using artificial neural networks (ANN) using R program. This work does not contain any package algorithm modules and is used for the forecasts of two uncertain parameters: Energy demand levels and the electricity market prices. By developing a time series forecasting method with ANN, learning of the neural network is done with the historical data and 24 hours of forecast is given as the output for the desired day. Meteorology forecasts are suggested to be used for the other uncertainties, i.e. wind speed and solar radiation levels. The mathematical optimization models that were developed for energy scheduling are coded in GAMS program. The models consist of linear programming and mixed integer linear programming models.Using multiple models helps us to see the effects of adding and removing each energy supply to the system.The input values are read from excel sheets that were created by R program and energy scheduling output is also printed to excel sheets by GAMS. Robust optimization is also integrated to the model in order to observe the worst case scenario results. Although, in the case studies the whole year of 2016 is investigated, the planning horizon is selected as 24 hours. The first reason for 24 hours of planning horizon is, the longest sale contract in the real market covers the next day. The second reason is that the forecast reliabilities of energy demand, real market prices, wind speed and solar radiation levels tend to decrease after longer periods. The optimization of the mathematical model is done using the forecasted values for the uncertain parameters. Forecast errors are inevitable and need to be taken into account. In the case studies, excess energy levels and energy deficit levels that were due to forecast errors, are also considered. Additional costs due to energy deficit and additional revenues due to excess energy are calculated. The main strategy used at times of energy deficit is to purchase the additional amount from the grid. At times of excess energy, on the other hand, selling the excess energy to the real market is not possible. One needs to make a sale contract at least one hour ago in order to sell to the market. Therefore the strategy is to store the excess energy if the battery is available and sell that amount to the market in the next hour. In order to observe the value of perfect information, the models that are run with forecasted values, are also run with real values for the uncertain parameters. This way, it is possible to see how much better the optimization model results could be without uncertainty within the parameters. Considering all details, the results of the case studies done for the island in GOSB can be listed as: Adding the battery to the system together with the renewable energy sources, reduces the overall costs. Energy demand is the parameter to which overall cost level is most susceptiple. Maximum energy generation level at the photovoltaic panels seem to have the least effect on cost. This is probably due to the relatively high production costs of photovoltaic panels compared to the wind turbine. While making forecast errors, the direction of the error seems to have an effect on the final cost. The opportunity costs of missed sales, which is the result of pessimistic forecasts, is much bigger than the costs of making additional electricity purchases due to energy deficit. In other words, having the forecasts of renewable energy geenration limits and electricity market prices higher than the actual values and having the forecsts of energy demand lower than the actual values is better than vice versa. Therefore, while making the forecasts it might be logical to be a little bit on the side of optimism. In the calculations, sales to the real time market needs a sale contract at least one hour earlier than the transaction. This is the reason for the excess energy that can be observed due to forecast errors, needs to be stored first and sold at the next hour. This causes a serious amount of waste energy that cannot be stored due to battery availability. Doubling the battery capacity has a positive effect in the case studies about the costs and waste energy levels. The optimum energy storage capacity of each system has to be decided according to its energy production and consumption levels. This investment decision is by itself and optimization problem. The proposed model is studied for an island in the GOSB, which is geographically poor in terms of wind speed and solar radiation levels. So it might be interesting to use the model at different geographic locations. Also the case studies are done with the assumption, that the initial investments of renewable energy sources -wind turbine and the photovoltaic panels- were already made. Therefore the unit energy production costs for these sources do not include the first investment cost. Another study which considers the lifetime costs of each energy source might be a good future work alternative.