Investigation of the impact of net load forecasting accuracy on the rational reserve level of the powersystem

Abstract

Relevance: the development of modern power systems is characterized by an increasing share of renewable energy sources, whose variable generation increases the uncertainty of power system operating conditions. Under these conditions, accurate net load forecasting, which accounts for variations in electricity consumption and the output of variable renewable energy sources, becomes increasingly important. Net load forecasting errors are one of the main sources of uncertainty in power system operation planning and directly affect the required amount of regulation reserve. An increase in forecasting errors necessitates additional reserve capacity, while maintaining excessive reserves results in inefficient utilization of available regulating resources. Conversely, an insufficient reserve level increases the probability of failing to cover actual net load deviations and the risk of regulation capacity shortages. In existing approaches, the probability of covering forecasting errors is often determined using predefined normative or statistical values. However, changes in forecasting quality, characterized by the magnitude and distribution of forecasting errors, can significantly affect the rational relationship between the probability of error coverage and the required reserve capacity. Therefore, investigating the impact of net load forecasting accuracy on the rational reserve level of a power system is an important scientific and practical task. Establishing this relationship makes it possible to account for the actual forecasting quality when selecting the probability of error coverage and determining the required regulation reserve, thereby ensuring the required level of regulation reliability while promoting the efficient utilization of available energy resources.

Aim: to analyze the impact of power system net load forecasting accuracy on the rational reserve level and to substantiate the probability of forecasting error coverage that ensures the required regulation reliability with a rational amount of reserve resources.

Methods: there were used statistical methods of analysis and forecasting.

Results: at the same error coverage level of 93,5%, the use of the Random Forest model reduces the total regulating reserve from 1441 to 1018 MW, representing a reduction of 423 MW (29,36%) compared with the ARMAX model. The largest reduction is observed in the downward regulating reserve, which decreases from 967 to 555 MW, corresponding to a reduction of 412 MW (42,56%). The upward regulating reserve changes only slightly, decreasing from 474 to 462 MW, or by 11,51 MW (2,43%).