4 TH RENEWABLE POWER GENERATION CONFERENCE (RPG TM ) – IET CONFERENCE 17 - 18 OCTOBER 2015 | NORTH CHINA ELECTRIC POWER UNIVERSITY, BEIJING, CHINA 1134 PROBABILISTIC UNIT COMMITMENT IN MULTI-AREA GRIDS WITH HIGH RENEWABLE ENERGY PENETRATION BY USING DYNAMIC PROGRAMMING BASED ON NEURAL NETWORK S. S. Kaddah, K. M. Abo-Al-Ez, M. G. Osman, T. F. Megahed Electrical Engineering Department, Faculty of Engineering, Mansoura University, Egypt. Keywords: Unit Commitment, Renewable energy, Storage System, Renewable forecasting. Abstract Nowadays, economic and environmental requirements are increased. Therefore, this paper presents a proposed solution of the unit commitment problem for a multi-area grid, which contains conventional and renewable energy sources and storage units. To ensure optimum and economic operation with the stochastic nature sources, it is essential to develop an efficient forecasting model for renewable power generation. Forecasting model was built by using a hybrid Markov to forecast solar power, while, auto regressive integrated moving average model is used to predict wind power. Unit commitment problem incorporate with forecasting model to develop probabilistic unit commitment. The proposed formulation is subject to multi-constraints. To overcome the variation and error of renewable power forecasting, the reserve coefficient is modified to develop two new reserves; up reserve and down reserve. The optimization algorithm used to solve probabilistic unit commitment is Dynamic programming based on Neural Network. The system under study in this paper is a standard IEEE 30, with wind speed and solar radiation data are based on the available data of the city of Florida in USA. 1 Introduction With the increased electrical energy consumption and the shortage of fossil fuel sources, it becomes essential to increase the dependence on renewable energy sources, mainly wind and solar energy. Due to the uncontrolled nature of renewable energy sources, it is important to install energy storage units to provide power during low levels of generation. The increase of the penetration level of renewable energy sources requires the restructuring of current electrical grids, to ensure optimal, reliable, and economical operation. Many recent researches have discussed the issues of Unit commitment (UC) problems. Some of recent researches considering the conventional power plants, short term operation and planning as introduced in [1,2]. Economic power was treated by using a good optimization tools to find suitable operation planning of the conventional units. These units are used fossil fuels which face many problems. Researches [3-5] used the same objective to minimize the economic operation considering the presence of wind and solar power generation systems and storage units. Although those researches have proposed optimization methods to solve the UC problem considering renewable power plants, they did not tackle the problem of wind and solar power prediction due to variable wind speed profile and solar radiation. But the optimization was solved by treating the intermittent resources as negative loads. There are very limited researches such as in [5] discussed the impact of renewable power forecasting on UC. In [5], an optimization strategy was developed to wind park control level that enabled defining the commitment of wind turbines and their active and reactive power outputs. That optimization strategy is solved by using mixed integer linear programming optimization problem based on minimizing the connection/disconnection changes of the individual wind generators for a given time horizon. Short term wind speed forecasts were expressed as power availability. The problem with renewable energies comes from the fluctuating and stochastic nature of the supply that is inherent in its nature. Also the power generation fluctuates independently from demand. So, increasing importance to find suitable control strategy is able to effectively control and manage the energy production in a flexible and proactive way. So renewable power forecasting is very important task to ensure the best use of the power generated from these sources. Also, a new control strategy is needed when inserting large amount of intermittent energy sources into the grid. On the other hand, most of the renewable energy sources are located far from the loads e.g. (deserts or seas) therefore it is important to study multi-area constraint in UC problem. This paper introduces a UC problem on a multi-area grid, which contains conventional/renewable sources and storage units with considerable penetration level of renewable power. The new formulation is aimed to use largest possible proportion of electrical energy to feed the loads from renewable sources. For forecasting, auto regressive integrated moving average (ARIMA) forecast wind speed and modified Markov chain to forecast sun radiation. The ARIMA method is further enhanced by taking into account the non-stationary characteristic of the wind power speed with only few model parameters. In addition, the proposed model does not rely on quantization and thus does not suffer from quantization errors. The modified Markov method used to forecast solar radiation is based on using a hybrid model of two Markov theories; the first one is a Markov estimation method to predict the new data. The predicted data are analyzed by using the second Markov analysis method to find the amount