The Research On The On-line Revising Algorithm For The Industrial Power Proportion Based On Daily Load Profile

With the boost of the interconnection progress of the main grid in China,the scale of the grid is becoming larger and the complexity of it is becoming higher, the dynamic stability and voltage stability problems is becoming more and more outstanding, the influence of load model on the power system digital simulation can't be neglected. Nevertheless, the time variation and territorial difference problems make the work of identifying the aggregate model that could accuracely describe the factual load is becoming harder. From the viewpoint of making the analytical results more reliable and providing precise basis for the design, operation and control of power system, more reasonable load model should be constituted.From the perspective of solving the time variance problem, datum from the supervisory control and data acquisition(SCADA) system and load control system is put into knowledge discovery procedure. On the base of analyzing the characteristics of component-based load modeling, the feasibility of building mapping between the data of SCADA system and load control system is illustrated. According to the real-time characteristic of the datum from SCADA system and load control system, mapping relation between the daily load profile of them is formulated. Firstly, the load rate, minimum load rate, peak to valley rate and the maximum load point of daily load curve are selected as the characteristic parameters of it. Secondly, the whole year is divided into four episodes, such as the winter maximum load episode, the winter minimum load episode, the summer maximum load episode and the summer minimum load episode, the maximum load day of every episode is selected as the classical day of it. Thirdly, the typical daily load curve of every episode is clustered through fuzzy C means clustering algorithm and class center of the daily load profiles of substation and load control system are obtained at last, that is the class center matrix of substation and load control characteristic matrix.Pattern recognition algorithm is used in calculating the typical overall membership via distance function method and fuzzy comprehensive discrimination combined with distance function method that use the datum through the clustering operation before: at first, constitute the typical overall membership relation between each class center of substation and each row vector of load control matrix for the purpose of forming the membership row vector and then the the matrix theory is applied to establishing the mappings between the classic overall membership and the typical industry power proportions which using component-based approach of every substation that in the class, the reliability of the mappings are judged by making use of the historical classical daily load datum. The result illustrates that the method that using fuzzy comprehensive discrimination combined with distance function is more precise, that is to say, this classic overall membership of this kind is more accurate in building mappings. On the premise of selecting suitable mappings, the real-time daily load profile is revised by means of calculating current overall membership and using the mappings formed above.For realizing the revising algorithm in actual surroundings a platform is developed that applying popular integrated developing environment(IDE) VC++.NET and relation database system SQL Server, the ActiveX data object(ADO) application program interface(API) is used in the operation on the datum in SQL Server. The actual substation daily load curve is revising and the results is given. The results show that the revising algorithm is explicable and is valuable in engineering to some extent.