This paper presents a novel methodology for short-term load forecasting in the context of significant shifts in the daily load curve due to the rapid and extensive adoption of Distributed Energy Resources (DERs). The proposed solution, built upon the Similar Days Method (SDM) and Artificial Neural Network (ANN), introduces several novelties: (1) selection of similar days based on hidden representations of day data using Autoencoder (AE); (2) enhancement of model generalization by utilizing a broader set of training examples; (3) incorporating the relative importance of training examples derived from the similarity measure during training; and (4) mitigation of the influence of outliers by applying an ensemble of ANN models trained with different data splits. The presented AE configuration and procedure for selecting similar days generated a higher-quality training dataset, which led to more robust predictions by the ANN model for days with unexpected deviations. Experiments were conducted on actual load data from a Serbian electrical power system, and the results were compared to predictions obtained by the field-proven STLF tool. The experiments demonstrated an improved performance of the presented solution on test days when the existing STLF tool had poor predictions over the past year.

The paper introduces a novel method for short-term load forecasting (STLF) when there is insufficient historical data to train a similar-day-based algorithm. The proposed approach involves constructing a virtual historical day using historical hour arrays, thus improving the algorithm's ability to forecast accurately. Relevant similarity factors can be used to identify similar series of hours, although their impact on power consumption can vary depending on different factors such as geographical location, power network infrastructure, season, and population habits. The proposed model starts with a range of factors and calculates the impact of the significant contributing factors on power consumption. The model allows for the optimal selection of similar hours, even in cases where it may be challenging to determine the most relevant factors and their impact on the forecast in advance. To improve the accuracy of the STLF forecast, the model uses a genetic algorithm (GA) to optimize the impact of each similarity factor. Experimental results presented in the paper demonstrate the potential of this method to enhance the accuracy of the STLF forecast.

This paper presents an adaptive short-term intraday load forecasting strategy designed for the operational requirements of transmission and distribution system operators. Standard forecasting approaches often report strong performance on selected periods, yet real utility operations require accurate predictions for every day and every hour of the year. Deviations during the operating day, caused by unexpected changes in consumer behavior, introduce forecasting errors and financial risk. To address this problem, we propose a multi-tiered forecasting model that selects the base method according to the availability of historically similar days. When many similar days exist, the model uses a pretrained artificial neural network, while linear regression is applied under moderate similarity conditions, and an arithmetic mean is used when only a few similar days are available. A real-time delta correction layer is applied in all cases, using recent intraday measurements to forecast short-term error and adjust the baseline output. This approach enables rapid adaptation to atypical days and intraday anomalies. Testing on five years of utility data demonstrates that the method maintains consistently low MAPE across all days and all hours, providing the level of accuracy needed for intraday market operations and system balancing.