Users' and Prosumers' Baseline for Local Flexibility Markets: a Comparative Assessment of Traditional and Machine-learning Techniques Applied to Smart Meters Data

The increasing complexity of electrical distribution networks, driven by the integration of renewables and distributed energy resources, requires new mechanisms such as local flexibility markets. These markets support distribution system operators in enhancing operational efficiency and mitigating congestions. In this regard, a key component of is the baseline calculation, crucial for distribution system operators and balancing service providers to estimate expected energy consumption or generation and remunerate the flexibility service accordingly. This study evaluates different baseline methodologies, ranging from traditional persistence-based approaches to advanced machine-learning techniques, employing real-world data from smart meters in Milan and Brescia. The analysis considers three user categories: active and passive domestic users, and non-domestic passive users. The numerical results indicate that advanced models, such as random forest regression, outperform traditional methods in baseline estimation. These findings contribute to optimizing baseline calculations, enhancing reliability in local flexibility markets, and supporting better management of distributed energy resources.