Benoit Delcroix

Residential demand response (DR) is an important flexibility resource for supporting power system reliability and deferring costly infrastructure investments, yet observed peak reductions remain highly heterogeneous across households and events due to differences in program design, automation, and household context, limiting predictability for utility planning and program design. This study evaluates residential DR performance across two dynamic pricing designs and an automated direct load control (DLC) program in Quebec, using linked survey data and 15-minute smart meter data from 492 households during the 2024–2025 winter season. Peak curtailment was quantified to compare reduction magnitude and consistency across programs and smart thermostat ownership groups and to identify key drivers of event-level peak reductions. Mean event-level peak reductions were 0.5 kW (standard deviation [SD] = 0.3 kW) and 1.3 kW (SD = 0.5 kW) for the two dynamic pricing programs, compared with 3.4 kW (SD = 0.6 kW) for the automated DLC program, where SD reflects event-to-event variability. Within the dynamic pricing programs, smart thermostat ownership was associated with substantially higher mean event-level peak reductions than non-smart thermostat ownership under the penalty-based pricing design (1.7 vs 0.3 kW) and the reward-based pricing design (3.7 vs 0.5 kW). Feature screening and interpretable decision-tree analysis indicated that program design and smart thermostat ownership were the dominant predictors of event-level peak reductions, with timing effects emerging conditionally within specific program–technology contexts. Overall, the results suggest that residential DR is stronger and more consistent when automation and enabling technology reduce reliance on real-time manual response.

Objective - This study investigates the variability in electricity consumption among tenants residing in the same dwelling in Montreal and Quebec City. The primary goals are to assess the extent to which occupant behavior and lifestyle influence electricity usage—particularly for heating—and to quantify the magnitude of variation in energy use between tenants. Methodology - The analysis proceeded in three phases: data extraction and preparation, data processing and calculation of key indicators, and results analysis. Four data sources were utilized: - A survey of residential electricity use - A registry of tenants with move-in and move-out dates - Monthly electricity consumption records - Historical weather data Key indicators calculated for each dwelling-tenant pair included monthly baseload, weather-adjusted heating energy use, and weather-adjusted total energy use. Outlier data were removed, resulting in a final sample of 652 dwellings. For each dwelling, statistical measures—mean, standard deviation, and coefficient of variation—were computed to evaluate the degree of energy use variation among tenants. Results - The findings reveal substantial variability in energy consumption among tenants occupying the same dwelling. Average variations for total energy use, baseload, and heating were 22%, 33%, and 34%, respectively, with distributions that are both skewed and dispersed.

Smart thermostats have become a promising device to control electric baseboard heaters' energy consumption while considering their flexibility in the demand response (DR) context. This article applies a shrinked-space search method to identify the tuning parameters of line voltage communicating thermostats (LVCTs). The proposed approach based on Bayesian optimization (BO) algorithm takes account of a reference model to drastically shrink the search space while enforcing the identification of a single set of parameters compatible with all arising dynamics of the controller and helping to establish an interpretable model. Furthermore, a subsequent integral tracking strategy has been adopted to convexify the objective function (for identification purposes) while considering the logical constraints governing the thermostat dynamics. This helps to recover the updating logic of the integral part of the controller model. The experimental validation results of eight LVCTs operating in an inhabited house show the effectiveness of the proposed method since it leads to establishing digital twins for the studied controllers. In addition, a case study is presented to demonstrate the usefulness of the reconstructed model in a DR framework.

Direct load control (DLC) is considered a viable solution to promote demand-side energy management, in which the utility provider adjusts consumers’ temperature setpoints via smart thermostats. Users commonly have the option to interrupt DLC and override them by adjusting their thermostat setpoints. However, the occurrence of overrides can have a detrimental impact on the overall efficacy of DLC. The user discomfort and the fact that an override may increase the load unexpectedly on the grid highlight the importance of understanding override mechanisms during DLC and the uncertainty related to occupants’ responses. This study examined user interactions with smart thermostats during DLC events using real-world data from the Ecobee Donate Your Data (DYD) program. According to the results, 35% of DLC was overridden by users, resulting in higher energy consumption during peak periods. A comprehensive analysis of four types of variables was conducted. A decision tree algorithm was used to classify smart thermostat users into two categories: “compliant users,” who rarely override DLC, and “non-compliant users,” who frequently override DLC, based on general information about their behavior and preferences and without any prior DLC experience. Moreover, three distinct types of DLC participants, characterized by their preferences and behaviors, were identified using a clustering algorithm. Classification results provide utilities with insight into where resources and efforts should be allocated to users who are more likely to comply with DLC. Clustering users into different typologies will enable utilities to design targeted and less disruptive DLC better aligned with the needs of DLC participants.

For many years, energy monitoring at the most disaggregate level has been mainly sought through the idea of Non-Intrusive Load Monitoring (NILM). Developing a practical application of this concept in the residential sector can be impeded by the technical characteristics of case studies. Accordingly, several databases, mainly from Europe and the US, have been publicly released to enable basic research to address NILM issues raised by their challenging features. Nevertheless, the resultant enhancements are limited to the properties of these datasets. Such a restriction has caused NILM studies to overlook residential scenarios related to geographically-specific regions and existent practices to face unexplored situations. This paper presents applied research on NILM in Quebec residences to reveal its barriers to feasible implementations. It commences with a concise discussion about a successful NILM idea to highlight its essential requirements. Afterward, it provides a comparative statistical analysis to represent the specificity of the case study by exploiting real data. Subsequently, this study proposes a combinatory approach to load identification that utilizes the promise of sub-meter smart technologies and integrates the intrusive aspect of load monitoring with the non-intrusive one to alleviate NILM difficulties in Quebec residences. A load disaggregation technique is suggested to manifest these complications based on supervised and unsupervised machine learning designs. The former is aimed at extracting overall heating demand from the aggregate one while the latter is designed for disaggregating the residual load. The results demonstrate that geographically-dependent cases create electricity consumption scenarios that can deteriorate the performance of existing NILM methods. From a realistic standpoint, this research elaborates on critical remarks to realize viable NILM systems, particularly in Quebec houses.

For many years, energy monitoring at the most disaggregate level has been mainly sought through the idea of Non-Intrusive Load Monitoring (NILM). Nevertheless, a practical application of this concept in the residential sector should address the underlying concerns raised by the technical specifications of case studies. From one side, such an operation must handle common matters related to the essence of any NILM system. Although this aspect has been thoroughly investigated by basic research, it is limited to the properties of public datasets. On the other side, it must deal with specific concerns corresponding to uncommon situations. These circumstances impose further restrictions on existent NILM schemes, however, they have been overlooked due to the lack of pertinent databases to scrutinize. Accordingly, this paper presents applied research on a potential solution to NILM for Quebec residences. It carries out a relevant investigation into the multi-faceted nature of this problem in order to reveal barriers to feasible implementations in the context of Quebec. This work commences with a concise discussion about the NILM idea to highlight its essential requirements for a fruitful practice. Afterward, it provides a comparative statistical analysis to represent the specificity and potential challenges of the case study in accordance with NILM necessities. For this purpose, the examination exploits data from real-world measurement systems in the same and European regions. Subsequently, this study focuses on a load identification exercise by proposing a combinatory approach that utilizes the promise of sub-meter smart technologies to integrate the intrusive aspect of load monitoring with the non-intrusive one. The former is aimed at extracting overall heating demand from the aggregate one by a supervised procedure based on Deep Learning (DL) models while the latter is designed for disaggregating the residual load through an unsupervised process on the basis of clustering techniques. The results demonstrate that geographically-dependent cases create electricity consumption scenarios under which existing NILM methods can be questioned. From a realistic standpoint, this research elaborates on critical remarks to realize viable NILM systems, particularly, in Quebec houses.