Apriori and K-Means algorithms of machine learning for spatio-temporal solar generation balancing

The number of grid-connected large-scale solar photovoltaic (PV) power plants has increased signifi-cantly in the last 10 years, which results in high PV power penetration into the grid. Especially for the wide-area spatially distributed countries, power ramp in one PV plant can be balanced with another PV power plant generation. This has been studied in the literature for short term horizons for high frequency data. In this study, hourly simulation data are analysed by Kendall's correlation coefficient, unsupervised and rule-based machine learning algorithms for spatio-temporal operational balancing constraints. Association rules generated by using the Apriori algorithm provide power ramp direction maps for Spatio-Temporal analysis. The K-means clustering (based on the Hartigan-Wong algorithm) is used for unsupervised learning application for the spatio-temporal relations for solar PV ramp zones.& nbsp; & nbsp; & nbsp; & nbsp; & nbsp;The proposed model can be used as a fast and effective decision-making tool with qualitative results for the system operator with minimal expert knowledge, while they can be also integrated to optimal power flow analysis constraints. This study introduces the first successful application of association rules integrated with the K-means technique for the Solar PV spatio-temporal balancing purpose to the best of our knowledge. (c) 2021 Elsevier Ltd. All rights reserved.

Automated summary

This study analyzed hourly simulation data using machine learning algorithms and association rules to address spatio-temporal operational balancing constraints for solar PV. The proposed model serves as a fast and effective decision-making tool for system operators with minimal expert knowledge and can be integrated into optimal power flow analysis constraints.