Abstarct: It is impossible and too dificult to analytically model many of the real process without differential and algebraic equations. In such casess, we can use time series for define analysis process. The goal of this thesis is to reduce costs by recommending a purchase box from power market. In that respect, the elecrical consumption of REJAL petrochemical is predicted by chaos methods over long term period. Hence, chaos time serise are investigated at first. Moreovere, the way of phase space reconstructure is presented by calculating the chaos time series's index such as delay time and embedded dimension. Then, the interpolation and denoised data of chaos time series are predicted by nearest neighbour method. The evaluation of used method is recognized by observation of simulation's results of known chaos systems. Finally, the cost function is defined baxsed on prediction of hourly electricity consumption data of REJAL petrochemical and statistical frequence distribution for programing to purchase electricity consumption of this company from the bilateral power market. The analysis of results show the suggested purchase box is excellence and beneficial.