Topic:"Energy Efficiency Resource Assessment Model" in M12932

Conclusion The Year Four Report ( Appendix A ) provides evaluated results for the 2024/25 Season, including sustained residential load shifting, statistically significant commercial load shifting, and initial results for the MURB TOU Tarif...

AI summary The Year Four Report evaluates the 2024/25 Season, highlighting residential and commercial load shifting results and initial MURB TOU Tariff outcomes. Econoler's review supports NS Power's methodology. The report includes survey data, marketing activities, and stakeholder session materials. NS Power plans to resume TVP Pilot rates for the 2026/27 Season, per the Board-approved framework.

Definitions The hourly net system requirement (MW) less all wind generation (MW) Cohorts are numbered based on the Phase they enrolled in the TVP program. For example, Cohort 4 refers to participants recruited in Phase 4 (i.e. the fourth W...

AI summary The text provides definitions related to energy consumption, rate classes, and program phases, including details about the TVP program and income brackets. It outlines cohort numbering, rate classifications, and metrics used in evaluating energy efficiency programs.

Environmental This evaluation uses Environment and Climate Change Canada temperature data[14](#page-126-2) to control for the effect of heating degree days in the regression model and to determine CPP reference days. The temperature data i...

AI summary The environmental evaluation uses temperature data from Environment and Climate Change Canada to control for heating degree days in a regression model. Steps are taken to clean and average weather data from multiple sources to ensure continuity and accuracy.

Where: i are datapoints corresponding the 24 hours in both seasons (Winter/non-Winter) and day types (weekend/weekday) for a total of 96 different combinations. $Tavg_i$ is the average load of the treatment customer during datapoint i. $Ca...

AI summary The text describes a method for analyzing customer load data across different seasons, days, and customer subgroups. It outlines the calculation of average and peak loads for treatment and control customers, and the process of balancing the control group based on various factors to account for E1 program participation.

Treatment and control customer participation in E1 programs (Home Energy Assessment, Green Heat, Appliance Retirement, Efficient Product Installation, Small Business Energy Solutions, Business Energy Rebates, and Custom) are balanced using...

AI summary The text discusses the process for balancing treatment and control customer participation in E1 programs to mitigate confounding effects on TVP results. It mentions the use of energy savings data from 2021 to 2024 and the allocation of program savings using a specific formula. The text also references specific reports and matter numbers related to the efficiency evaluations.

The savings correction factor logic is to account for customers participating in an E1 program part way through either the pre-pilot or pilot period. For instance, continuing with the example above, a Cohort 2 TVP participant's pre-pilot p...

AI summary The text explains the methodology for adjusting savings correction factors in E1 programs, accounting for customer participation timelines and balancing energy savings between treatment and control groups. It details a process to ensure accurate comparisons by selecting control groups based on similarity metrics and achieving energy savings balance.

bing the correlation between being a treatment customer and the average load. β 2 is the coefficient describing the correlation between being during the pilot period and the average load. β 3 is the coefficient describing the correlation b...

AI summary This section discusses a regression model analyzing the impact of load and usage, focusing on coefficients related to treatment customers, pilot periods, and HDD. It explains how the model evaluates the effect of the Thermal Value Program (TVP) on average load using regression analysis.

age daily bill. The regression models for daily price elasticity and inter-period substitution price elasticity are presented in equations ([10)](#page-140-1) and ([11)](#page-142-0), respectively. $$ln(Daily\ Avg\ Usage\ kWh)_{it} = \\ =...

AI summary The text presents regression models for daily price elasticity and inter-period substitution price elasticity, including equations and variables such as temperature, energy prices, and treatment effects. These models aim to analyze customer behavior in response to pricing and temperature changes.

IV.2.2 Phase 2 To account for customers who've modified their heating source from their original rate code classification, the heating classification methodology was modified to use AMI and nearby weather station temperature data for indiv...

AI summary The heating classification methodology was updated in Phase 2 to use AMI and weather station data for individual premises, focusing on days with average temperatures below 10°C. Customers are classified as primary or secondary electric based on consumption correlation with temperature and a mean consumption threshold during top 20 consumption days.

IV.2.4 Phase 4 Phase 4 (this Phase) of the evaluation uses a similar methodology as Phase 3, with two key distinctions that improved the accuracy of the model. - 1. Heating classification is used in addition to the neighborhood criteria du...

AI summary Phase 4 of the evaluation improves upon Phase 3 by incorporating heating classification in control group selection and restricting AMI data to overnight hours for heating classification. These changes enhance the accuracy of load impact analysis, particularly for customers who have adjusted their space heating behavior in response to price signals.

IV.3.1 Temperature Temperature influences customer consumption, as such the regression model used in the first Phase of the evaluation was modified to include heating degree days, refer to equations (19) and (20) taken directly from the Ph...

AI summary The text discusses the influence of temperature on customer energy consumption, explaining how regression models in different phases of an evaluation account for heating degree days (HDD). Equations (19) and (20) are provided, with the latter including HDD as a variable to improve model accuracy.

Attachment V: Validation of Mixed Effects Regression The NS Power commercial TVP evaluations (Phase 1 – 3)[25](#page-152-1) has historically provided statistically insignificant results. Findings such as "statistically insignificant" can b...

AI summary The NS Power commercial TVP evaluations have historically yielded statistically insignificant results due to low statistical power and methodological choices. The use of a fixed effects regression model for commercial customers, similar to residential, introduces Type II errors. In Phase 4, a mixed effects regression model was introduced to address the heterogeneity of commercial customer load profiles, improving the evaluation's accuracy.

g (a, f) AM-Peak Load Reduction, (b, g) PM-Peak Load Reduction, (c, h) AM-Snapback Effect, (d, i) PM-Snapback Effect, and (e, j) Bill-Saving during Winter for TOU-MURBs The RE autocorrelation (r) was evaluated based on the residual serial...

AI summary The text discusses a comparison between Fixed-Effect and Mixed-Effect models in analyzing residual errors and autocorrelation. It highlights that the Mixed-Effect model produces smaller residual errors, less autocorrelation, and a more normal distribution compared to the Fixed-Effect model, which shows negatively skewed residuals and higher autocorrelation.

Robust Margin of Error When TVP load impact was estimated using the DiD analysis, the residual errors can be autocorrelated. If residual autocorrelation is not adjusted for, the standard error will be underestimated. As such, it is necessa...

AI summary The text discusses the adjustment of the margin of error (MoE) in estimating TVP load impact using DiD analysis, addressing residual autocorrelation and heteroskedasticity. It provides formulas for adjusted and robust MoE calculations, including the use of standard error, autocorrelation coefficient, and robust covariance matrix.

Coincidence of Critical Peak Events and System Margin In its Decision and Order on the Year Three (2023/24) Evaluation Report (Year Three Report) in M11823, the Board provided: NS Power agreed with Synapse that system margin forecasts may...

AI summary The document discusses the relationship between system margin forecasts and Critical Peak Pricing (CPP) events, noting that system margin during CPP events was significantly lower than during other hours. The Board directed NS Power to analyze this relationship and refine methods for estimating load reductions through TVP tariffs.

- › A good proportion of that decrease in performance is due to the difference in space heating scenario mix: when adjusting for the same space heating scenario mix as Year 1, Year 3 results are only approximately 25% lower

AI summary The decrease in performance is largely attributed to differences in the space heating scenario mix. When adjusted to match Year 1's mix, Year 3 results show a 25% reduction.

Definitions Adjusted Net Load The hourly net system requirement (MW) less all wind generation (MW) Figure 49: Eco Shift Visualization of Binary Variables 111 Figure 50: (a) Seasonality Segmentation of All Cohorts of TVP Commercial Particip...

AI summary The document provides definitions and includes figures related to energy load management, segmentation, and modeling approaches for commercial participants under Time-Varying Pricing (TVP) and Critical Peak Pricing (CPP) programs. It discusses metrics like Seasonal Load Ratio (SLR), Residual Autocorrelation, and modeling techniques such as Fixed Effect and Mixed Effect approaches.

Nova Scotia Power (NS Power, Company) launched the Time-Varying Pricing (TVP) Tariff Pilot on November 1, 2021, with the purpose of encouraging customers to shift load from Winter peak periods to Winter off-peak periods. These tariffs pres...

AI summary Nova Scotia Power's Time-Varying Pricing (TVP) Tariff Pilot, launched in 2021, encourages load shifting to reduce peak energy demand and system costs. The fourth annual report highlights continued customer enrollment growth, load reduction effects, and improved evaluation methods. The pilot demonstrates significant on-peak load reductions, particularly during high system load periods, and provides insights for future demand response strategies.

Environmental This evaluation uses Environment and Climate Change Canada temperature data13F [14](#page-111-2) to control for the effect of heating degree days in the regression model and to determine CPP reference days. The temperature da...

AI summary The evaluation uses temperature data from Environment and Climate Change Canada to account for heating degree days in a regression model and determine Critical Peak Pricing (CPP) reference days. Data cleaning steps are applied to address gaps in weather data from eight stations.

Treatment and control customer participation in E1 programs (Home Energy Assessment, Green Heat, Appliance Retirement, Efficient Product Installation, Small Business Energy Solutions, Business Energy Rebates, and Custom) are balanced using...

AI summary The text discusses the process of balancing treatment and control customer participation in E1 programs to mitigate confounding effects on TVP results. It outlines methods for calculating energy savings using average per participant savings adjusted for line loss factors and allocates these savings based on program year and TVP cohort.

_6. \ln\left(\textit{Avg Energy Price} \frac{$}{kWh}\right) \\ &+ \beta_7. \left(\textit{Avg Energy Price} \frac{$}{kWh}\right). \ln(\textit{Temp_K})_{it} + \varepsilon_{it} \end{split}$$ $$(10)$$ Where: In(Daily Avg Usage kWh) is the natu...

AI summary The text presents two regression models analyzing the impact of temperature, energy prices, and TVP on electricity usage and load ratios. The models include coefficients for variables such as temperature, energy price, and their interactions, reflecting price responsivity and elasticity.

III.3.2.2 Economic Impact Regression Model $$\begin{aligned} \textit{DailyBill}_{it} &= (\beta_0 + u_i) + \beta_1. Treatment_i + \beta_2. \textit{PilotPeriod}_t \\ &+ \beta_3. (\textit{Pilot} \times \textit{Treatment})_{it} + \beta_4. \tex...

AI summary The text presents three regression models used to analyze the economic impact of energy programs. The models include variables such as treatment, pilot period, heating degree days, and energy prices. The equations are used to assess the effects of demand-side management and time-varying pricing on daily billing, energy usage, and load patterns. The treatment variable is excluded for commercial time-of-use and critical peak pricing analyses due to the lack of control groups.

Where: is the linearized monthly consumption for the treatment customer for month 'i' is the linearized monthly consumption for the control customer for month 'i' is the number of months compared between treatment and control, n = 12 in th...

AI summary The document outlines a method for identifying control customers by comparing their linearized monthly consumption data to that of treatment customers. A similarity metric based on the mean absolute difference over 12 months is used, with a threshold of 0.2. If fewer than 15 matches are found, the selection process expands to secondary and tertiary neighbourhood criteria.

IV.2.3 Phase 3 To account for customers changing heating types between pre-pilot and pilot period, Phase 3 included a pre-pilot and pilot period evaluation of heating classification using the same methodology as in Phase 2, with the distin...

AI summary Phase 3 of the proceeding evaluates changes in customer heating types between the pre-pilot and pilot periods by defining four new heating classifications based on the difference between classifications in each period, using a methodology similar to Phase 2.

Attachment V: Validation of Mixed Effects Regression The NS Power commercial TVP evaluations (Phase 1 – 3)24F [25](#page-137-1) has historically provided statistically insignificant results. Findings such as "statistically insignificant" c...

AI summary This document discusses the limitations of previous TVP evaluations by NS Power, noting that statistically insignificant results may stem from low statistical power or methodological issues. It highlights the difference between residential and commercial customer load profiles and explains how the use of a fixed effects regression model for commercial customers may have introduced Type II errors. In Phase 4, a mixed effects regression model was implemented to better account for commercial load heterogeneity.

Baseload Heterogeneity Analysis The baseline segmentation analysis indicates substantial heterogeneity in electricity consumption patterns across participating MURBs, particularly with respect to Baseload levels. Based on these characteris...

AI summary The baseline segmentation analysis reveals significant differences in electricity consumption patterns among MURB customers, classifying them into Low, Medium, and High baseload classes. These classes show distinct baseload levels and load profiles, indicating substantial heterogeneity with implications for program impact estimation and model specification.

Definitions Adjusted Net Load The hourly net system requirement (MW) less all wind generation (MW) Other Board Directives and Consensus Agreement Commitments 142 List of Tables Table 1: Summary of Peak Load Reductions by Residential TVP Ta...

AI summary The document provides definitions and outlines tables summarizing load reductions, participation metrics, and energy consumption data related to time-varying pricing (TVP) and other tariff programs. It includes references to adjusted net load and other technical terms used in the analysis of energy efficiency and demand response initiatives.

The methodology is primarily composed of two components: control group selection and regression modelling. Control group selection includes matching treatment customers with control customers based on a similar location (i.e. neighbourhood...

AI summary The methodology involves control group selection based on location, heating type, and program participation, and uses regression models like DiD and DDD to evaluate program impacts. Commercial TOU and CPP tariffs use a semi-DiD approach due to control group limitations. Attachments provide further details.

IV.2.3 Phase 3 To account for customers changing heating types between pre-pilot and pilot period, Phase 3 included a pre-pilot and pilot period evaluation of heating classification using the same methodology as in Phase 2, with the distin...

AI summary Phase 3 of the proceeding evaluates changes in customer heating types between the pre-pilot and pilot periods by defining four new heating classifications based on differences in classification, using the same methodology as Phase 2.

IV.2.4 Phase 4 Phase 4 (this Phase) of the evaluation uses a similar methodology as Phase 3, with two key distinctions that improved the accuracy of the model. - 1. Heating classification is used in addition to the neighborhood criteria du...

AI summary Phase 4 of the evaluation improves accuracy by using heating classification alongside neighborhood criteria for control group selection and restricting AMI data to overnight hours for heating classification. This ensures more accurate load impact analysis and captures the effect of TVP price signals on space heating behavior.

IV.3.1 Temperature Temperature influences customer consumption, as such the regression model used in the first Phase of the evaluation was modified to include heating degree days, refer to equations (19) and (20) taken directly from the Ph...

AI summary The document discusses the modification of a regression model in Phase 1 of an evaluation to include heating degree days (HDD) to account for temperature's influence on customer consumption. Equations (19) and (20) are provided, showing the model's structure for different phases, with HDD incorporated in later phases to improve accuracy.

Attachment V: Validation of Mixed Effects Regression [Control Group Selection)](#page-22-1) relative to residential customers. As such, the methodological choice to use a fixed effects regression model for commercial customers, like reside...

AI summary This section discusses the validation of a mixed effects regression model used in the Phase 4 evaluation to account for the heterogeneous nature of commercial load and baseloads. It contrasts this approach with the fixed effects model, which was deemed unsuitable for commercial customers due to the risk of significant Type II error.

Table 3: Impact Evaluation Metric Summary Category Metric MURB TOU Change in load during peak periods (kW) • Change in load (kW) during peak, mid-peak, and overall impact during Winter and non-Winter. • Change in load by month. Snapback (k...

AI summary Table 3 summarizes impact evaluation metrics for MURB TOU, including load changes during peak periods, snapback effects, load and usage impacts, and economic changes such as electricity bill variations and price elasticity.

1 Methodology The methodology is primarily composed of two components: control group selection and regression modelling of load and economic impacts. This methodology in detail was filed as Attachments III to V to the Phase 4 TVP EM&V Repo...

AI summary The methodology for evaluating Time-Varying Pricing (TVP) involves control group selection and regression modeling. Control groups are matched based on location and load profiles, and a mixed-effects regression model using difference-in-differences (DiD) is applied. Data quality for Phase 5 is noted to be lower than in Phase 4, which may affect results.

2.1.5 Change in Demand Demand for this evaluation is defined as the max 15-minute electricity consumption in terms of power (kW) calculated at AMI interval data frequency (i.e. every 15-minutes). Figure 5 illustrates a comparison of relati...

AI summary The evaluation defines demand as the maximum 15-minute electricity consumption during winter months and compares changes in monthly demand for MURB TOU participants. Although some indications of demand changes were observed, they were not statistically significant. Data quality issues during November, December, and January excluded these months from the analysis.

to correct for this mismatch between control and treatment data, all treatment data without the corresponding control hourly usage data was removed from the analysis, refer to [Figure 9b](#page-35-0). Figure 9: Hourly Data Completeness in...

AI summary The analysis discusses data completeness for the Phase 5 MURB TOU EM&V analysis, highlighting that 60.2% of data was usable after removing unpaired treatment data. NS Power used a DiD regression approach similar to Phase 4 for reliability, despite incomplete control group AMI data.

TVP Objectives for Winter 6 (2026-2027) - Restore all Pilot TVP rates to operational status for November 1, 2026. - Focus on the customer experience of existing TVP customers to retain their trust in the stability of the rates. - Continue...

AI summary The document outlines objectives for the Time-Varying Pricing (TVP) program for Winter 6 (2026-2027), including restoring pilot TVP rates to operational status, improving customer experience, continuing communication efforts, applying learnings from prior evaluation reports, and advancing the Areas of Focus Work Plan from stakeholder engagements.

- 2. Multi unit residential building (MURB TOU) Tariff EM&V. MURB TOU scope includes existing commercial TOU scope and additional scope specific to MURB TOU. Table 4: Impact Evaluation Metric Summary Category Metric TOU, CPP, MURB TOU (all...

AI summary The document discusses the evaluation metrics for the Multi Unit Residential Building (MURB) Time-Varying Pricing (TVP) Tariff, including load changes during peak and non-peak periods, regional differences, income classification impacts, and economic effects such as changes in electricity bills and price elasticity.

Fixed and Mixed Effects Regression Modelling $$Load_{it} = \beta_0 + \beta_1. Treatment_i + \beta_2. PilotPeriod_t + \beta_3. (Pilot \times Treatment)_{it} + \beta_4. HDD_{it} + \beta_5. (Pilot \times Treatment)_{it}. HDD_{it} + \varepsilo...

AI summary The text discusses the use of Fixed and Mixed Effects Regression Modelling to analyze load data, considering factors like treatment, pilot periods, heating degree days (HDD), and Eco Shift participation. Mixed Effect modelling was chosen due to small sample sizes and heterogeneity among commercial TVP, MURB TOU customers, and Eco Shift participants.

Energy Usage by Space Heating Classification - The greatest energy reduction was achieved by steady electric heating classification as high as 1.96 kWh per day (3.5%) during non-holiday weekdays. - Electrified participants increased energy...

AI summary The text discusses energy usage differences across space heating classifications, noting significant reductions in energy consumption during non-Winter periods and increased usage by electrified participants during Winter weekends and holidays.