Recent broad-based deployment of Advanced Metering Infrastructure (AMI) enables the opportunity for broader adoption of time-based rates, and the benefits that result have been sizable contributors to making the investments cost effective. However, some stakeholders have raised concerns about the assumptions underlying the benefits assessments in AMI business cases. Such concerns are especially acute for certain subpopulations of residential customers. Low income, elderly and chronically ill (i.e., vulnerable) customers are believed to have less load that can be shifted or reduced to capture bill savings, lack the know-how or wherewithal with which to curtail usage, likely have more limited financial resources which may compel them to avoid high priced periods by reducing electricity for essential usage potentially causing them physical harm, and more generally may be more adversely affected by higher bills, which might possibly result from certain forms of time-based rates. There is very limited existing literature that addresses these questions specifically with regard to vulnerable subpopulations. This paper, based on a larger report, extends the existing empirical literature on the experiences of low-income customers exposed to critical peak pricing, and provides the first glimpses into the experiences of the elderly and those who reported being chronically ill.

Prior research has shown load reduction estimates from residential event-driven demand response programs (e.g., Critical Peak Pricing)using X of the highest Y days with a weather adjustment method are the best performing within the class of currently used baseline methods. However, they are still biased relative to estimates produced from randomized control trials (RCTs), the unbiased “gold standard” evaluation method. In this paper we identify underlying factors that cause some of the bias found in one commonly used baseline method. Rather than simply quantifying bias, our research provides a deeper understanding of what causes the bias and thus can be used to develop more accurate methods that are not subject to these underlying factors. Previous studies have compared various baseline methods relative to each other; however, because all baseline methods are biased, it is impossible to determine the true bias that exists in them. We have access to a unique RCT dataset: the Sacramento Municipal Utility District's study of critical peak pricing. Our analysis of 23 event days over two summers allows us to identify the true bias on load reduction estimates by using the RCT estimates as the unbiased gold standard against which we compare the estimates from the baseline methods. We found that spillover of energy reductions, from hours targeted by a program onto other hours, is one underlying factor that is a major cause of bias in baseline methods. We discuss alternative baseline methods that may not be subject to this same bias.

The goal of this work is to investigate the impact of missing values in clustering joint categorical social sequences. Identifying patterns in sociodemographic longitudinal data is important in a number of social science settings. However, performing analytical operations, such as clustering on life course trajectories, is challenging due to the categorical and multidimensional nature of the data, their mixed data types, and corruption by missing and inconsistent values. Data quality issues were investigated previously on single variable sequences. To understand their effects on multivariate sequence analysis, we employ a dataset of mixed data types and missing values, a dissimilarity measure designed for joint categorical sequence data, together with dimensionality reduction methodologies in a systematic design of sequence clustering experiments. Given the categorical nature of our data, we employ an “edit” distance using optimal matching. Because each data record has multiple variables of different types, we investigate the impact of mixing these variables in a single dissimilarity measure. Between variables with binary values and those with multiple nominal values, we find that the ability to overcome missing data problems is more difficult in the nominal domain than in the binary domain. Additionally, alignment of leading missing values can result in systematic biases in dissimilarity matrices and subsequently introduce both artificial clusters and unrealistic interpretations of associated data domains. We demonstrate the usage of t-distributed stochastic neighborhood embedding to visually guide mitigation of such biases by tuning the missing value substitution cost parameter or determining an optimal sequence span.

The goal of this paper is to investigate the impact of missing values in categorical time series sequences on common data analysis tasks. Being able to more effectively identify patterns in socio-demographic longitudinal data is an important component in a number of social science settings. However, performing fundamental analytical operations, such as clustering for grouping these data based on similarity patterns, is challenging due to the categorical and multi-dimensional nature of the data, and their corruption by missing and inconsistent values. To study these data quality issues, we employ longitudinal sequence data representations, a similarity measure designed for categorical and longitudinal data, together with state-of-the art clustering methodologies reliant on hierarchical algorithms. The key to quantifying the similarity and difference among data records is a distance metric. Given the categorical nature of our data, we employ an 'edit' type distance using Optimal Matching (OM). Because each data record has multiple variables of different types, we investigate the impact of mixing these variables in a single similarity measure. Between variables with binary values and those with multiple nominal values, we find that the ability to overcome missing data problems is harder in the nominal domain versus the binary domain. Additionally, artificial clusters introduced by the alignment of leading missing values can be resolved by tuning the missing value substitution cost parameter.