This paper presents an analysis of three typical microgrid operation modes. First is the analysis of the grid-connected mode, in which the main goal is to optimize the microgrid operation (in terms of economics, carbon emissions and/or reliability) when it is connected with the main grid. Second is the analysis of the planned island mode, in which the main goal is to provide a stable and smooth transition from grid-connected to island mode while keeping both the frequency and the voltages within the required limits. Finally, the paper presents the analysis of the emergency island mode, which occurs after sudden islanding, in which the main goal is to maintain the energization of critical customers. The objectives and the operation strategies of microgrid management are different in each operation mode. This paper proposes a novel, centralized, multi-agent-based, microgrid controller architecture, which provides the coordination of all three operation modes and enables the easy configuration/combination of optimization goals that are subject to a given set of operational constraints. The microgrid controller consists of a real-time agent (responsible for closed loop control and maintenance of the system's frequency), an operation planning agent [responsible for near-term (15 minutes up to 24 hours) and short-term (1 to 7 days) optimization] and a state agent (responsible for overall monitoring, management and coordination of the other two agents). A microgrid controller with the aforementioned agents has been developed and tested. The simulation results are presented for a typical microgrid test example. Copyright © 2016 John Wiley & Sons, Ltd.

This paper represents comparison of two artificial intelligence based hybrid models for short term load forecasting (STLF). Models have the same input/output architecture and are built on SVM and ANN technologies, respectively. Algorithm consists of two modules connected in a sequence, and output from first module is connected as additional input to second module. First module acts as a predictor of maximal load of forecasting day and second acts as hourly load predictor. Models are part of large STLF solution and in respect to computational and memory limitations simple input space is designed. This architecture enables short training time which is targeted for frequent re-training needs in modern utilities due to frequent change in customer number and behavior.

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.