Data Analytics and Stochastic Simulation Methods for Risk-Controlled Network Planning: Validation Case Study

Abstract

EDPD is developing initiatives taking advantage of the continued technological investments being made in AMI. Themain focusof these initiatives is on how data analytics can be used to enhance the current simulation methods in order to allow a risk-controlled network planning approach.Thus, we have developedspecific data analytics methodologiesto explore the large volume of metering data aiming at clustering customer profiles into typical load/generation profiles.Within each cluster, load has been modelled by a discrete-time non-stationary Markov process that realistically reproduces high resolution daily load volatility and time-dependency. This process was then integrated and used inDPlan, the decision-making support tool used by EDPD, to perform thousands of power-flows and estimate the current and voltage distributions of each branch and node.As this novel approach to network planning is significantly different from the traditional methods and involves a series of steps of data analysis, it is exceedingly important to validate the results of the simulations by comparing them to real data. This paper intends to assess those results, by presenting a comparison between the calculated synthetic profiles and the metered profiles, for a specific case study.An in-depth analysis of load distribution will allow us to validate the developed methodologies to facilitate and improve the performance of probabilistic solutions, assuring to planners that the resulting errors are not significant and so, the solution framework can be applied to a realistic smart grid context and prospect its applicabilityfor real situational awareness.