LOAD PROFILING BY SELF ORGANIZATION WITH AFFINITY CONTROL STRATEGY
Ovidiu Ivanov · 2008
This paper describes a modified self-organizing algorithm which addresses the problem of consumer classification in electric distribution networks according to the shape of the load profiles. This algorithm combines the traditional growing self organizing map algorithm with fuzzy implementation and a control strategy based on affinity measures. The way consumers are represented in the general model of the network is largely responsible for the efficient planning and operation of distribution networks. It is generally accepted that during the planning stage the usage of global (e.g. peak load, time of peak load, time of losses) and statistical (e.g. simultaneity coefficients) parameters of the load profiles offers good results. However, during the operation stage, decisions like sectionalizing open-loop networks, reactive power compensation, optimal tap position for transformers in substations, or postfault network reconfiguration can no more be made based on the above methodology. The literature suggests using typical load profiles (TLPs) associated with simple readings in the network as models for end-use consumers. Standard procedures generate TLPs by off-line processing a great amount of load data recorded at consumers with known, simple structure. TLPs are represented using 60, 30 or 15 minutes sampling rates and are expressed in kW or p.u. [4, 5]. Selforganizing algorithms applied to load profile classification offer certain advantages, such as the simultaneous development of the classification and TLP generation processes or the avoidance of human interference during classification. This paper proposes a new technique for consumer classification and TLPs data base generation. A fuzzy implementation of Self Organizing Maps (SOMs) is applied as a basic approach, combined with a weighting procedure to compute