Application of Expectation Maximization Algorithm for Measurement-based Power System Load Modeling

This paper describes a practical application of the Expectation Maximization algorithm for computing the static load model of large electricity consumers. Such a model is necessary for effective and reliable operational control of electric power systems. The possibility of estimating basic states of electrical load using arrays of measured voltage and power data is shown. The criteria for selecting the most appropriate clusters are formulated and a step by step linear load model parameter estimation method is proposed. A practical method application is illustrated by using measurements of voltage and real power obtained at a large industrial facility. As a result, a load model with two basic states and corresponding linear models is obtained.