High penetration of distributed renewable energy sources in distribution networks, if optimized solely for self-economic benefits, can lead to node voltage violations and exacerbate the peak-shaving burden on the power grid. As accurate electrical parameters of distribution networks are often difficult to obtain, data-driven power-flow methods have gained traction in distribution system applications. The data-driven power-flow model proposed in this article utilizes only measurements from existing monitoring systems, thereby avoiding costly grid upgrades and hardware investments. The node voltage regression model explicitly incorporates the influence of the feeder root node—the substation bus voltage. This allows the model to leverage the voltage regulation capability of the upstream substation when the distribution network’s own regulation resources are insufficient. A load-following term is introduced into the objective function to reduce the system’s peak-shaving demand by minimizing the deviation between the distribution network’s net output and a predetermined load set-point. For handling this load deviation, the absolute deviation is used instead of the conventional squared deviation. This approach not only aligns the unit of the load deviation term with electrical energy but also transforms this part of the objective function into a linear expression through the introduction of auxiliary variables. The revenue for the energy storage system (ESS) is defined as the price differential between the purchase and sale, accounting for charging/discharging efficiency. This formulation eliminates the need for auxiliary binary variables to prevent the physically infeasible scenario of simultaneous charging and discharging of the ESS. Through the aforementioned treatments, the proposed optimization model is formulated as a linear programming (LP) model. This avoids the long computation times and convergence difficulties associated with solving quadratic programming (QP) or mixed-integer linear programming (MILP) models for large-scale systems. The proposed algorithm has been applied to four different real-world distribution networks, the largest of which comprises 114 distribution transformers. The effectiveness of the proposed data-driven power flow algorithm is validated using a full year of historical data from this large network as samples. Furthermore, simulations across five different case studies demonstrate that incorporating voltage security constraints ensures the safe operation of active distribution networks and that considering the load-following objective effectively alleviates system’s peak-shaving pressure.
Zhang et al. (Tue,) studied this question.