ANN-based scenario generation methodology for stochastic variables of electric power systems

• A scenario generation methodology based on artificial neural networks is proposed.
• Electric load, photovoltaic (PV) and wind production scenarios are created.
• Scenario cross-correlation and reduction techniques are applied.
• Comparison with time series-based scenario generation models is presented.
• Test results on real-world power systems prove the effectiveness of the methodology.

In this paper a novel scenario generation methodology based on artificial neural networks (ANNs) is proposed. The methodology is flexible and able to generate scenarios for various stochastic variables that are used as input parameters in the stochastic short-term scheduling models. Appropriate techniques for modeling the cross-correlation of the involved stochastic processes and scenario reduction techniques are also incorporated into the proposed approach. The applicability of the methodology is investigated through the creation of electric load, photovoltaic (PV) and wind production scenarios and the performance of the proposed ANN-based methodology is compared to time series-based scenario generation models. Test results on the real-world insular power system of Crete and mainland Greece present the effectiveness of the proposed ANN-based scenario generation methodology.