Lukas Baur

Electrical load forecasting has seen many specialized models, yet systematic comparisons across load domains are rare. This paper closes this gap by evaluating twelve mod-els-spanning baselines, linear methods, tree-based, similaritybased, deep learning, and AutoML approaches-on 80 two-year datasets with 15-minute resolution enriched by weather and holiday data. Aggregated loads showed the highest predictability (median sMAPE of 5.01), followed by commercial (17.52, industrial (27.55), and residential (46.08) loads. Notably, commercial loads outperformed industrial ones despite lower aggregation. Beyond aggregation level, load variance, temporal self-similarity, and external factors substantially affected accuracy. This work provides the first systematic, cross-domain empirical analysis of electrical load predictability.

The increasing integration of renewable energy sources has led to greater volatility and unpredictability in electricity generation, posing challenges to grid stability. Ancillary service markets, such as the German control reserve market, allow industrial consumers and producers to offer flexibility in their power consumption or generation, contributing to grid stability while earning additional income. However, many participants use simple bidding strategies that may not maximize their revenues. This paper presents a methodology for forecasting bidding prices in pay-as-bid ancillary service markets, focusing on the German control reserve market. We evaluate various machine learning models, including Support Vector Regression, Decision Trees, and k-Nearest Neighbors, and compare their performance against benchmark models. To address the asymmetry in the revenue function of pay-as-bid markets, we introduce an offset adjustment technique that enhances the practical applicability of the forecasting models. Our analysis demonstrates that the proposed approach improves potential revenues by 27.43 % to 37.31 % compared to baseline models. When analyzing the relationship between the model forecasting errors and the revenue, a negative correlation is measured for three markets; according to the results, a reduction of 1 EUR/MW model price forecasting error (MAE) statistically leads to a yearly revenue increase between 483 EUR/MW and 3,631 EUR/MW. The proposed methodology enables industrial participants to optimize their bidding strategies, leading to increased earnings and contributing to the efficiency and stability of the electrical grid.

The liberalization of the German energy market has created opportunities for end-consumers, including industrial companies, to actively participate in the electricity market. By making their energy loads more flexible, consumers can generate additional income and thus save money. Energy storage systems can be utilized to achieve the required flexibility by temporarily storing excess electrical energy in the form of heat, cold, or electricity for later use. This publication focuses on how the dimensionality of energy storage is influenced by load forecasting. The results show that inaccuracies in load forecasting lead to a direct over-dimensioning and thus, a deterioration of the economics of energy storage technologies. Using two scenario cases, it shows on the one hand how important good forecasts are and on the other hand that buffers must be included in the conceptual design in order to be able to compensate for forecast errors.

In Zeiten volatiler und steigender Energiepreise gewinnt die Lastprognose als Entscheidungsgrundlage zunehmend an Bedeutung. Die fortschreitende Digitalisierung erlaubt zudem ein detailliertes Aufzeichnen von Verbrauchsdaten auf mehreren Abstraktionsebenen. In diesem Beitrag werden zwei Ansätze, die eine Messzählertopologie in die Prognose integrieren, mit dem Standardansatz auf Hauptzählerebene verglichen. Die Ergebnisse auf den Unternehmensdaten motivieren für einen flächendeckenden Einsatz. Increasingly volatile and rising energy prices make load forecasting increasingly important as a basis for decision-making. Moreover, the advancing digitalization allows a detailed recording of consumption data on several abstraction levels. This paper compares two approaches that integrate a metering topology into load forecasting with the standard approach at the main meter level. The results on enterprise data motivate widespread deployment.