Intelligent Data Mining in Power Distribution Company for Commercial Load Forecasting

In a company, the manual checking of load data resulting from measurements is a repetitive process. It takes a long time and is subject to errors, since after one or two hours performing that task, it is difficult for the licensee's technician to identify the existing deviations in the large databases. Another aspect is that the data in those bases are usually distributed through tables on the screen. Such fact renders the task more complicated due to the lack of a better notion of sets such as possible rearrangements, for example, if those same data were displayed in graphics. The aim of the current work is to substitute the load monitoring previously done through manual checking by a computer system specially conceived to meet the necessities of Energias Brasil - Bandeirante. Because of the proposed methodological development and the implemented computer software, the data checking is practically automatic, eliminating errors that could not be identified. The computer software brings a new paradigm to check flaws in the data, making possible to relate them in many dimensions. The software performs a repetitive task based on pattern recognition techniques. Moreover, the program indicates possible flaws in measurement, easing the correction of figures and helping in the correct measurement of the company's load. This paper aims to present the methodology that was developed to automate the load monitoring and its computer implementation.