Application Of Data Mining In Line Losses Calculation For Electric Power Marketing System

The line loss rate is an important comprehensive technical indicator which is used to assess the quality of power supply enterprise by the country. A reasonable and accurate losses calculation is the foundation of energy computation calculation, statistics, and analysis to guide the loss reduction work in electricity sector. The crucial problem for the power marketing system is how to make full use of large data to get the accurate theoretical line losses. And the line losses error is related to the accuracy of raw data, mathematical models and mathematical methods, but the main error is caused by the inaccuracy of raw data.On the background of the State Grid Corporation research and development projects in 2010(State Power Grid Development (2010) 151), the paper designs the application scheme of applying the data mining technology in line losses calculation to the question of the inaccuracy of raw data, which realizes preprocessing of the raw data participated in the line losses calculation to ensure the accuracy and integrity of the raw data. The scheme uses the idea"Birds of a Feather Flock Together"of the clustering analysis to process the large historical data. The accurate raw data can be got together faster, while the inaccurate data are removed, so it can obtain more accurate results of the line losses calculation.To the question of the imperfect data, the paper proposes the method of replacing the missing value. According to the practical characteristic of the raw data, the paper takes the place of the missing value by the different kinds of data to approach the real data. The two methods mentioned above are simulated by the data mining software SPSS. Then carry the simulation results substitution to the online theoretical line losses calculation system, the results show that the theoretical line losses which are obtained after the processing of data mining technology are closer to the real line losses. The results prove the accuracy of the raw data, and the feasibility, efficiency and superiority of the methods.