This paper proposes the practical implementation of an approach for improvement of automatic generation control performance and discusses its particular importance for power systems with large share of variable energy resources. The approach allows advancement of the functional block responsible for estimation of plant participation factors, which increases flexibility and selectivity of power flow control. Real time optimization model was established to reach different control goals. To meet the performance requirements, the artificial neural network was developed for power flow estimation. To improve performance and reduce computational burden, the Lasso regression method was proposed and tested for selection of the model features relevant for the considered control task. Finally, the software tool was developed to implement the algorithm, based on the proposed approach, and tested on a model of real 60 GW interconnection containing 464 nodes and 742 branches. The results of the software testing confirm its feasibility and easy integration into existing automatic generation control systems.