arXiv subjects

Mladen Kezunovic

Publications and source records attributed to Mladen Kezunovic.

5 recordsLinked to original sources

arXiv2026

Evaluating Large Language Models for Forced Outage Risk Prediction: Benefits and Comparison to Machine Learning

This study examines the ability of large language models (LLMs) to predict the risk of weather-related forced outages in the distribution grid in a zero-shot framework, without labeled training data. The problem is formulated as a binary severity classification task across three forecast horizons (3h, 6h, 12h), using six years of outage records and high-resolution weather data for a utility service area in central Texas. Four zero-shot LLMs are benchmarked against two supervised classifiers across two input configurations: one using current weather observations and the other using weather forecast data. Results show that supervised models outperform LLMs on macro-F1 and precision, while newer LLM generations achieve competitive scores. Beyond accuracy, LLMs offer complementary strengths in actionable reasoning and geographic scalability, suggesting that combining them with supervised models may be the best practice.

cs.LG↗

arXiv2025

Multi-layer Optimized Coordination of Smart Building Resources in Active Power Distribution Systems

This paper proposes a multi-actor coordination platform for the optimal utilization of smart buildings resources, including roof top PV generation and battery energy storage system (BESS), in active power distribution systems. The proposed multi-actor coordination includes the Smart Building Coordinator (SBC), Micro-Grid Coordinator (MGC) and Distribution System Coordinator (DSC). The coordinators operate independently and only exchange limited information with each other to reach an optimal solution. In the proposed platform, a hierarchical optimization problem is solved to optimally determine the operating point of all distribution system resources. The proposed platform fully preserves the confidentiality of the behind the meter (BTM) data of the buildings since no information about the status of the PV system, BESS, and load of the building is shared with the owner of the power system. The proposed platform has a flexible and scalable architecture where the computational task of coordinating microgrids and smart buildings with distribution grid is performed locally at the MGC and SBC layers, respectively. Numerical simulations show the efficacy of the proposed platform in coordinating the BTM resources with the rest of the distribution system.

eess.SY↗

arXiv2022

Data-driven State of Risk Prediction and Mitigation in Support of the Net-zero Carbon Electric Grid

An approach for reaching the net-zero carbon electricity grid is to intensify the deployment of distributed renewable generation resources such as photovoltaic (PV) solar and wind generation, complemented with stationary and mobile (electric vehicle) battery energy storage systems (BESS). This paper assumes a scenario where the PV renewable generation and BESS are integrated into a nano-Grid (n-Grid), a prosumer-owned virtual power plant installed at the residential or commercial sites. To be profitable, this distributed energy resource (DER) needs to be managed effectively to support its own load and the wholesale and retail market services. Our paper introduces a risk-based, data-driven approach focused on predicting the State of Risk (SoR) of the utility grid outages. The SoR prediction enables the development of optimal mitigation strategies aiming at reducing the impact of the grid outages by harvesting the n-Grid flexibility. Several data analytics tools to assist the n-Grid operators and aggregators for n-Grid participation in the wholesale market ancillary service products are introduced, and some preliminary implementation results are demonstrated.

eess.SY↗

arXiv2022

Prediction of Solar Radiation Based on Spatial and Temporal Embeddings for Solar Generation Forecast

A novel method for real-time solar generation forecast using weather data, while exploiting both spatial and temporal structural dependencies is proposed. The network observed over time is projected to a lower-dimensional representation where a variety of weather measurements are used to train a structured regression model while weather forecast is used at the inference stage. Experiments were conducted at 288 locations in the San Antonio, TX area on obtained from the National Solar Radiation Database. The model predicts solar irradiance with a good accuracy (R2 0.91 for the summer, 0.85 for the winter, and 0.89 for the global model). The best accuracy was obtained by the Random Forest Regressor. Multiple experiments were conducted to characterize influence of missing data and different time horizons providing evidence that the new algorithm is robust for data missing not only completely at random but also when the mechanism is spatial, and temporal.

cs.LG↗

arXiv2020

Economic Evaluation of Transformer Loss of Life Mitigation by Energy Storage and PV Generation

High penetration of plug-in electric vehicles (PEVs) can potentially put the utility assets such as transformer under overload stress causing decrease in their lifetime. The decrease in PV and battery energy storage system (BESS) prices has made them viable solutions to mitigate this situation. In this paper, the economic aspect of their optimal coordination is studied to assess transformer hottest spot temperature (HST) and loss of life. Monte Carlo simulation is employed to provide synthetic data of PEVs load in a residential complex and model their stochastic behavior. For load, temperature, energy price and PV generation, real data for City of College Station, Texas, USA in 2018 is acquired and a case study is developed for one year. The results illustrate using BESS and PV is economically effective and mitigates distribution transformer loss of life.

eess.SY↗