Integrated platform for "prosumer" energy flow management based on a multi-criteria optimization model

[A. Z. Petrušić]

XI, 188 str.

Biografija: str. 188. Bibliografija: str. 156-171.

Renewable energy sources

With the emergence of prosumers as market participants who simultaneously produce and consume electricity, and with the growing penetration of RES in the grid, the traditional paradigm of assumed unidirectional power flows in the grid, is changing together with circumstances on electricity markets. The major challenge for a prosumer, as autonomous hybrid system that comprises of a solar power plant, a battery energy storage system and an electric vehicle charger, is the constant need to simultaneously account for economic, environmental and technical criteria, which are most often mutually conflicting. This dissertation develops forecasting models based on the LSTM neural network architecture for day-ahead prediction of electricity generation and consumption. For forecasting the output of the solar power plant, two complementary models are applied, the yearly LSTM model and the specialized MLP model for the winter period, while the separate LSTM model is developed for load forecasting. The main contribution of the dissertation is a multi‑criteria optimization model based on the integration of Multi‑Attribute Utility Theory (MAUT) and the Genetic algorithm, which simultaneously optimizes three criteria: minimization of supply costs, maximization of the share of renewable energy in EV charging cycles, and minimization of degradation of the local battery energy storage system. The methodology for tracking energy provenance based on the principle of superposition is also developed, determining exact share of the renewable energy share both in charging processes and when meeting the demand of other loads within the prosumer system. The comparative analysis of the additive and complementary models with the aggregation of MAUT utility functions, on a case of real prosumer system, has shown the superiority of the complementary model under conditions of pronounced conflicts between criteria, achieving a significantly higher share of renewable energy at lower supply costs. The results confirm the necessity of a multi‑criteria approach for optimal management of prosumer systems and are opening paths for further research toward the areas of prosumer aggregation and hybrid approaches to multi‑criteria optimization.

prozjumer, tržište električne energije, obnovljivi izvori energije, predviđanje, rekurentne neuronske mreže, nelinearna optimizacija, MAUT, genetski algoritam

prosumer, electricity market, renewable energy sources, forecasting, recurrent neural networks, nonlinear optimization, MAUT, genetic algorithm

621.311.243:004.899.8]:519.853:519.816(043.3)

T 190

Tekst