Presentation on theme: "SEEM Calibration: Revisited Revising the regression to use continuous heat loss variable Regional Technical Forum December 17, 2013."— Presentation transcript:
SEEM Calibration: Revisited Revising the regression to use continuous heat loss variable Regional Technical Forum December 17, 2013
Background 2 SF RBSA Pie: 1404 Homes Adjustment factors converted to calibrated thermostat settings Approved by the RTF on May 21, SEEM Calibration “Phase I” Compared SEEM (68°F, day and night) heating energy estimates to billing data estimates. Restricted to 429 RBSA homes with well-known characteristics, no non-utility fuels, and clear heating signatures in billing data. Regression used to determine adjustment factors that align SEEM (68°F) with billing data estimates of total heating energy.
Background SEEM Calibration “Phase II” Independent of Phase I; adjustments apply on top of Phase I adjustments. Based on billing (VBDD) heating kWh estimates--does not use SEEM estimates. Identifies variables that drive patterns in electric heating energy among “program-like” RBSA homes. Variables related to: - Non-utility heat sources, - Gas heat sources, and - Phase I filters. 3 SF RBSA Pie: 1404 Homes Approved by the RTF on September 17, Today’s work applies only to Phase I. It does not affect Phase II.
Phase I Review (1) Intended to limit complication in future UES workbooks by choosing variables that correspond with RTF measures. Wanted to limit to variables well-known through RBSA (e.g., no infiltration). Regression variables (and adjustment factors) coded as indicator functions. Adjustments for: – Heating equipment, – “Poor” insulation in walls or ceiling, – Uninsulated crawlspace, – Climate Zone. 4
Phase I Review (2) Regression yields adjustment factors, which are converted to calibrated T-stat values. Factors converted to calibrated T-stat values using SEEM T-stat sensitivity curves… 5
Why are we revisiting this? Applications more diverse than appreciated in May. Basic proposal is to trade in some simplicity for realism. Regression Variable (Main proposed change). Replace insulation step functions with continuous heat loss function. New function treats heat loss from different sources equally: – Magnitude of heat loss matters but path does not; – Includes loss via infiltration (imputed for homes w/o blower door test); – Small changes yield small calibration adjustments (no threshold effects). T-stat Role (Secondary proposal). Apply adjustment factors directly, rather than converting to thermostat adjustments. – Concern is that thermostat “calibration knob” might bias results; – Adjustment factors would be relative to SEEM (69°F day / 64°F night) rather than SEEM (68°F day / 68°F night). 7
Changing Role of T-Stat (1) Current calibration begins with SEEM Input = 68°F day/night – This arbitrary value didn’t affect the results much since adjustment factors were converted to t-stat settings. Proposal would begin with SEEM Input = 69°F day, 64°F night – Values based on survey results from RBSA (not arbitrary). – Not much difference by heating system type, so the same rounded number used for all. – Values would become standard SEEM input (adjustment factors would be applied to output). 8
Changing Role of T-Stat (2) What if we calculate adjustments relative to SEEM (68/68) and SEEM (69/64) and then convert adjustments into t-stat values? Little difference in the end results. 9
Regression Revision (1) Main work is in developing heat loss variable. Infiltration loss based on CFM-Natural; – CFM-Nat is a SEEM input, derived from blower-door test data; – Blower door tests for about 1/3 of RBSA houses; – Regression-based “averages” for homes w/o blower door tests; – Calculations and regression based on RTF guidance. Convert infiltration loss to same units as conductive heat loss; add heat loss rates together; normalize by surface area. Result is called “Uo-Both”. 10
Regression Revision (2) Developing the heat loss variable... Ran preliminary regressions to see if any additional transform is needed. 11 Effect very pronounced in the low range of U-values, but going from fairly high heat loss to very high heat loss has little effect. Final proposed heat loss function equals Uo-Both up to a point, but stays constant beyond that point. Cut-off value is 0.20 in Z1, in Z2, 0.15 in Z3.
Regression Revision (3) Proposed Regression VariableCoefficientP-value Intercept Climate Zone Climate Zone Electric Resistance Uo Both (cut) Current Regression VariableCoefficientP-value Intercept uninsulated crawl poor wall or ceiling insulation Climate Zone Climate Zone Electric Resistance
Regression Revision (4) Example: Home in Zone 1 with Electric- Resistance heat and moderately insulated walls and floors. 13 Ceiling Insulation Uninsulated Wall/Ceiling Current Regression Adjustment Uo - Both Proposed Heat Loss Variable Proposed Regression Adjustment R5162% % R30084% %
14 Comparing Regression Results Z1 Elec. Resistance Current and Proposed Adjustments (Example) R30 – Current R5 - Current R5 (Proposed) R30 (Proposed)
15 Comparing Regression Results Z1 Elec. Resistance Current (unins. wall-ceiling/ unins. crawl) and Proposed Current 0/0 Current 0/1 Current 1/0 Current 1/1 Proposed
16 Comparing Regression Results Elec. Resistance Current (R0) and Proposed (R3) All Heating Zones
Regression Revision (5) Effect on UES Calculations Insulation – Current: Different pre/post adjustments for only the cases where Uninsulated Insulated – Proposed: Different pre/post adjustments for nearly all the cases, even new construction Windows, Air Sealing – Current: No change in pre/post adjustments. – Proposed: Different pre/post adjustments for nearly all the cases, even new construction Duct Sealing, Heat Pump Upgrades, and Heat Pump CC&S – Current: No change in pre/post adjustments. – Proposed: No change in pre/post adjustments. (Central) Heat Pump Conversions – Current: Different pre/post adjustments. – Proposed: Different pre/post adjustments. DHPs (not a part of this analysis) Measure Interactivity – Old Method: Adjustment factors vary only when components are uninsulated. – Proposed Method: Adjustment factors are different for each “characteristic scenario”. 17
Bottom Line… Regression / Heat Loss Variable Proposal. Staff sees benefits in the new heat loss function: – Heat loss due to infiltration treated the same as conductive loss; – All forms of conductive loss treated the same; – Small changes yield small calibration adjustments (no threshold effects). Drawbacks are added complication and overhead related to making a change. T-stat Proposal. Staff is neutral on this one. What do you believe is really driving differences between SEEM and billing data? – If it’s really T-stat settings, then it’s best to implement adjustments via thermostat calibration; – If it’s something else, then adjustment factors are probably better— thermostat calibration could bias some results. 18
Decisions “I ______ move that the RTF, in its single family calibration method: (choose one) a)Switch to using a function based on continuous U o, as presented. b)Continue using the existing step-functions.” “I ______ move that the RTF, in its single family calibration method: (choose one) a)Switch to using adjustment factors directly, along with pre-assigned thermostat setting inputs of 69F day and 64F night. b)Continue using ‘calibrated’ thermostat settings.” 19
Additional Slides… 20
21 Comparing Regression Results Elec. Resistance Current (R0) and Proposed (R3) All Heating Zones
22 Comparing Models in T-stat terms Elec. Resistance Current (R0) and Proposed (R3) All Heating Zones
23 Comparing Regression Results Gas/Heat Pump Current (R0) and Proposed (R3) All Heating Zones
24 Comparing Models in T-stat terms Gas/Heat Pump Current (R0) and Proposed (R3) All Heating Zones