Environment-Dependent Components Identification of Behind-the-Meter Resources via Inverse Optimization

TL;DR

This paper introduces a hybrid physics-inspired and data-driven method to identify and decompose behind-the-meter resources into environment-dependent components using limited measurement data, enhancing load modeling accuracy.

Contribution

It presents a novel hybrid framework combining inverse optimization and physics models for BTM resource component identification under data privacy constraints.

Findings

Accurately decomposes BTM load into environment-dependent components.

Validates modeling accuracy and robustness through numerical tests.

Demonstrates application in reducing system operation costs.

Abstract

With the increasing penetration of behind-the-meter (BTM) resources, it is vital to monitor the components of these resources and deduce their response behavior to external environment. Owing to data privacy, however, the appliance-wise measurement is invisible to the power system operator, which hinders the accurate modeling of load identification. To this end, this paper proposes a hybrid physics-inspired and data-driven framework for decomposing BTM components based on external measurement of total load and environmental factors. The total load is decomposed into different environment-dependent components, namely storage-like component, PV generation component, thermostatically-controlled load component, and periodic component. The overall load identification adopts a double-layer iterative solution framework. A data-driven inverse optimization algorithm is developed to identify parameters of the energy storage-like component. The physics-inspired model is proposed to identify the capacity and response of the rest components. The modeling accuracy and robustness of the proposed method are validated by numerical tests. The application significance of the proposed BTM identification method is also validated in electricity market clearing for reducing system operation costs.