Automatic Identification of Power System Load Models Based on Field Measurements

Abstract

With an ever growing complexity, the power grids are designed and operated with an increasingly reduced stability margin. Under such circumstances, the adequate modeling of existing and new power system loads, with all its challenges, is receiving renewed attention. To simplify and to large extent automate the task of load modelling, this paper presents a methodology and associated software tool-Automated Load Modelling Tool (ALMT) to automatically (without human intervention) develop load models and derive corresponding model parameters from recorded power system responses. The approach facilitates automatic identification of load models and derivation of corresponding parameters for three different types of load models, i.e. polynomial, dynamic exponential, and composite load model. Several case studies using real life measurements at 11 kV distribution buses are used to test and validate developed methodology and software tool.

A Statistical Method to Establish Voltage Dependency Load Parameters Based on Field Measurements

“…Voltage variations due to OLTC operation is between 0.5 to 2.5% [6]. A 0.2% is selected so as to include minor changes in voltage due to load/equipment switching in the network.…”

“…Step 5: Data extraction in the form of V, P, and Q that corresponds to each time T, T-1 and T+1 as in (5) and (6).…”

“…The importances of using appropriate load model in the power system study are mentioned in [6]. Dynamic load models are required for stability studies whereas static load models are applied to load flow studies.…”

Enhanced ambient signals based load model parameter identification with ensemble learning initialisation

“…In addition, the quantity and components of the load are time‐varying, which also makes the load models time‐varying and uncertain [2]. Therefore, improving the accuracy of load models will significantly help to provide more reliable power system simulation results [3].…”

“…In the previous measurement‐based load modelling approaches, optimisation methodologies have been widely applied to identify the load model parameters from measurement data [1]. Both gradient‐based algorithms [15, 16, 17] and swarm intelligence algorithms [3, 18, 19] have been used in previous load modelling approaches. In the optimisation‐based approaches, how to get the global optimal solution instead of a local optimal solution is one of the most challenging problems.…”