Sambaran Bandyopadhyay

Voltage Correlations in Smart Meter Data

The connectivity model of a power distribution network can easily become outdated due to system changes occurring in the field. Maintaining and sustaining an accurate connectivity model is a key challenge for distribution utilities worldwide. This work shows that voltage time series measurements collected from customer smart meters exhibit correlations that are consistent with the hierarchical structure of the distribution network. These correlations may be leveraged to cluster customers based on common ancestry and help verify and correct an existing connectivity model. Additionally, customers may be clustered in combination with voltage data from circuit metering points, spatial data from the geographical information system, and any existing but partially accurate connectivity model to infer customer to transformer and phase connectivity relationships with high accuracy. We report analysis and validation results based on data collected from multiple feeders of a large electric distribution network in North America. To the best of our knowledge, this is the first large scale measurement study of customer voltage data and its use in inferring network connectivity information.

Individual and Aggregate Electrical Load Forecasting: One for All and All for One

Electrical load forecasting is an important task for utility companies, in order to plan future production and to increase the efficiency of the distribution network. Although load forecasting at the aggregate level has been extensively studied in existing literature, forecasts for individual consumers have been shown to be prone to errors. This paper deals with the problem of electrical load forecasting at multiple scales, from individual consumers to the network as a whole. We use smart meter data from carefully selected sets of consumers for this purpose. First, we consider the problem of forecasting the load for individual consumers at the outermost nodes of the distribution network. We propose an algorithm which considers external available information like calendar or weather contexts along with the energy consumption profiles of different consumers for accurate mid-term and short-term load forecasting. Multiple aggregation approaches are considered for utility level forecasting, in order to characterize their error properties. We show that careful clustering of consumers for aggregation can result in smaller errors. We experiment with two public data sets for demonstrating the advantages of the proposed method over the state-of-the-art approaches.

Axioms to characterize efficient incremental clustering

Although clustering is one of the central tasks in machine learning for the last few decades, analysis of clustering irrespective of any particular algorithm was not undertaken for a long time. In the recent literature, axiomatic frameworks have been proposed for clustering and its quality. But none of the proposed frameworks has concentrated on the computational aspects of clustering, which is essential in current big data analytics. In this paper, we propose an axiomatic framework for clustering which considers both the quality and the computational complexity of clustering algorithms. The axioms proposed by us necessarily associate the problem of clustering with the important concept of incremental learning and divide and conquer learning. We also propose an order independent incremental clustering algorithm which satisfies all of these axioms in some constrained manner.

Machine learning for inferring phase connectivity in distribution networks

The connectivity model of a power distribution network can easily become outdated due to system changes occurring in the field. Maintaining and sustaining an accurate connectivity model is a key challenge for distribution utilities worldwide. This work focuses on inferring customer to phase connectivity using machine learning techniques. Using voltage time series measurements collected from customer smart meters as the feature set for training classifiers, we study the performance of supervised, semi-supervised and unsupervised techniques. We report analysis and field validation results based on real smart meter measurements collected from three feeder circuits of a large distribution network in North America.

Algorithm to Control Power Production from Solar Panels

With the cost of renewable energy sources like photo-voltaic solar progressively declining, it is expected that contribution from these sources in overall power generation will increase. However, these newer sources have their own set of drawbacks. First and foremost is the variability in generation, then, controlling the power output from solar is also not trivial. Furthermore, there is a mismatch between the demand profile from the consumer and the supply profile of renewable generation. This increased penetration of renewable sources may lead to grid related problems such as over voltage, power back flow and instability. In this paper we propose a curtailment based solution to ameliorate the problem of over production. We provide an algorithm which controls the power output from solar panels. When demand is less than the maximum production capacity of the panels, the algorithm curtails the output to match the demand. And when demand is more, the algorithm does a best effort to produce maximum possible power.