Agent-based analysis of the impact of near real-time data and smart balancing on the frequency stability of power systems

Single imbalance pricing incentivizes balance responsible parties to intentionally introduce power schedule deviations in order to reduce the control area imbalance and receive remuneration through the imbalance settlement mechanism. This practice, known as smart, passive, or implicit balancing, is actively encouraged in the Netherlands and Belgium through the publication of near-real-time imbalance data by the transmission system operator. This paper is the first to assess how different types of near real-time data influence smart balancing and frequency stability when multiple profit-driven participants apply smart balancing with different strategies. A Monte-Carlo simulation of a dynamic multi-agent model analyses these effects for varying agent behaviours and environmental conditions, using historical data from the German control block as reference. When near real-time data is published with high accuracy and short delays, over 75% of simulations show reductions in frequency restoration reserve activation — averaging 25% for positive aFRR, 10% for negative aFRR, and 20% in associated costs. In contrast, using coarse bins and long delays leads to increased costs in more than 50% of cases. Short-term frequency variability increases in more than 90% of the simulations, with the interquartile range consistently between 0 and 10% independent of binning and delays. Overall, adverse impacts on frequency stability are more pronounced when near real-time data are published using large bins and with long delays, which also leads to more extreme peak frequency deviations. These results show that system operators must carefully design how near real-time data is published in order to make smart balancing beneficial. • Novel dynamic agent-based model for smart balancing by balancing responsible parties. • Analysis of effects of near real-time data and smart balancing on frequency stability. • Poor data design worsens frequency stability. • Agent and environmental parameter variations in a Monte-Carlo setting. • Imbalance estimation and decision models based on multivariate Gaussian distributions.