Merge pull request #51 from NREL/ccaradon/update_hvac_distribution_do…

…cumentation

Ccaradon/update hvac distribution documentation

documentation/reference_doc/4_9_hvac.tex

@@ -7,9 +7,18 @@ \subsection{HVAC System Heating Fuel Type}
 
 The probability distributions are informed by two data sources. First, there are the CBECS 2012 microdata, which include data on heating fuel(s), building type, and census division for the surveyed buildings. This data can be used to produce probability distributions for heating fuel by building type at the census division level. However, several data sources suggest notable variation within census divisions, which indicates that increased granularity may be needed (beyond what CBECS can provide). The heating fuel type probability distributions used in ResStock---which provides data for residential buildings at the county level---were used to add granularity. However, initial comparisons showed discrepancies between the ResStock data and the CBECS data, which is likely due to inherent differences between residential and commercial buildings. This indicated that the ResStock data should not be used directly. To rectify this, the county-level ResStock data were scaled to align with the CBECS data. This preserved the county-level variation in fuel type prevalence provided by the ResStock data, while also preserving the census division totals provided by the CBECS commercial data. District heating values were not available in the ResStock data, so the per-building-type CBECS values were used for all counties in a given census division.
 
-In some cases, filtering down to a specific region and building type in the CBECS data yields very few samples. This can lead to unreliable conclusions for a region. To mitigate this, we took a blended approach, where some fraction of the CBECS region fuel type percentage comes from the regional samples only, and some fraction comes from the national sample for the building type. If more than 70 samples exist for a given building type and region, then 100\% of the fuel type prevalence comes from that specific region. (70 was selected as a conservative value using engineering judgment.) If there are fewer than 70 samples, the number of samples divided by 70 will be the fraction used for the region, and the remainder will use the national numbers. For example, if a region has 35 office samples, 50\% (35\//70) of the effective CBECS regional value will come from the CBECS region, and the other 50\% will come from the national CBECS value for the building type. This will cause region/building type combinations with lower sample sizes to have a stronger inheritance of the national characteristics than the regional characteristics when we lack sufficient evidence to support this level of detail.
+In some cases, filtering down to a specific region and building type in the CBECS data yields very few samples. This can lead to unreliable conclusions for a region. To mitigate this, we took a blended approach, where some fraction of the CBECS region fuel type percentage comes from the regional samples only, and some fraction comes from the national sample for the building type. If more than 15 samples exist for a given building type and region, then 100\% of the fuel type prevalence comes from that specific region. (The threshold of 15 samples was selected baced on engineering judgment to balance process reliability and regional variability.) If there are fewer than 15 samples, the number of samples divided by 15 will be the fraction used for the region, and the remainder will use the national numbers. For example, if a region has only 12 office samples, 80\% (12\//15) of the effective CBECS regional value will come from the CBECS region, and the other 20\% will come from the national CBECS value for the building type. This will cause region/building type combinations with lower sample sizes to have a stronger inheritance of the national characteristics than the regional characteristics when we lack sufficient evidence to support this level of detail.
 
-Some commercial building HVAC systems use multiple fuel types. For example, a VAV system with a gas furnace in the air handling unit and electric resistance coils in the reheat boxes, or a gas furnace DOAS with variable refrigerant flow (VRF) heat pumps serving the zones. This can complicate the categorization of these systems into a single primary fuel type. To address this, we make the assumption that HVAC systems can be grouped into one of two categories: those that use combustion fuel in some capacity, and those that are all-electric, with no combustion present. We used this categorization methodology because most buildings that use all-electric systems do so because they do not have natural gas service. Consequently, some models categorized with a combustion fuel type, such as propane or natural gas, may be assigned a mixed-fuel-type HVAC system type that still contains a large electric heating component. A full list of ComStock HVAC systems and their fuel type categories are shown in Table~\ref{tab:hvac_system_heating_fuel_categories}. Further detail on model HVAC system assignment methodology can be found in Section~\ref{sec:HVAC_System_Type}.
+Some commercial building HVAC systems use multiple fuel types. For example, a VAV system with a gas furnace in the air handling unit and electric resistance coils in the reheat boxes, or a gas furnace DOAS with variable refrigerant flow (VRF) heat pumps serving the zones. This can complicate the categorization of these systems into a single primary fuel type. To address this, we determine the primary heating fuel type for the mixed fuel systems. The primary heating fuel is the heating fuel expected to carry the majority of the heating load. For example, the previously mentioned example of a VAV system with gas heat at the air handler and electric reheat would be classified as an electric-heated system, since the majority of heating for multizone VAV systems usually comes from the reheat. A full list of ComStock HVAC systems and their fuel type categories are shown in Table~\ref{tab:hvac_system_heating_fuel_categories}. Further detail on model HVAC system assignment methodology can be found in Section~\ref{sec:HVAC_System_Type}.
+
+Figure~\ref{fig:fuel_cbecs_v_cstock} compares the prevalence of heating fuel type by stock floor area for CBECS 2012 and ComStock, by building type. In most cases, ComStock closely aligns to the CBECS 2012 values. However, there are some differences between the two sources due to randomness in the sampling process and from the use of other data sources to achieve county-level granularity in fuel type prevalence. The largest difference is in small hotels where ComStock shows 87\% of the floor area using electric heating while CBECS suggest 74\%, an absolute difference of 12\%.
+
+\begin{figure}
+  \centering
+  \includegraphics[width=1.10\textwidth]{figures/cbecs_comstock_fuel_type_comparison.png}
+  \caption[Comparison of heating fuel type prevalence by floor area between CBECS 2012 and ComStock.]{Comparison of heating fuel type prevalence by floor area between CBECS 2012 and ComStock.}
+  \label{fig:fuel_cbecs_v_cstock}
+\end{figure}
 
 The county-level prevalences of different heating fuel types are shown in Figure~\ref{fig:map_naturalgas} (natural gas), Figure~\ref{fig:map_electricity} (electricity), Figure~\ref{fig:map_fueloil} (fuel oil), Figure~\ref{fig:map_propane} (propane), and Figure~\ref{fig:map_district} (district heating).
 
@@ -46,9 +55,9 @@ \subsection{HVAC System Heating Fuel Type}
   \includegraphics[width=0.8\textwidth]{figures/map_districtheating.png}
   \caption[Fraction of ComStock models using district heating/water heating per county]{Fraction of ComStock models using district heating per county.}
   \label{fig:map_district}
-\end{figure}
+\end{figure}   
+
 
-Because ComStock samples by building count rather than building area, discrepancies in the fuel type prevalence can occur when comparing CBECS to ComStock on a floor area basis. This is illustrated by building type in Figure~\ref{fig:cbecs_comstock_fuel_comparison_full_service_restaurant} (full service restaurant), Figure~\ref{fig:cbecs_comstock_fuel_comparison_hospital} (hospital), Figure~\ref{fig:cbecs_comstock_fuel_comparison_large_hotel} (large hotel), Figure~\ref{fig:cbecs_comstock_fuel_comparison_large_office} (large office), Figure~\ref{fig:cbecs_comstock_fuel_comparison_medium_office} (medium office), Figure~\ref{fig:cbecs_comstock_fuel_comparison_outpatient} (outpatient), Figure~\ref{fig:cbecs_comstock_fuel_comparison_primary_school} (primary school), Figure~\ref{fig:cbecs_comstock_fuel_comparison_quick_service_restaurant} (quick service restaurant), Figure~\ref{fig:cbecs_comstock_fuel_comparison_retail} (retail), Figure~\ref{fig:cbecs_comstock_fuel_comparison_secondary_school} (secondary school), Figure~\ref{fig:cbecs_comstock_fuel_comparison_small_hotel} (small hotel), Figure~\ref{fig:cbecs_comstock_fuel_comparison_small_office} (small office), Figure~\ref{fig:cbecs_comstock_fuel_comparison_strip_mall} (strip mall), and Figure~\ref{fig:cbecs_comstock_fuel_comparison_warehouse} (warehouse). Note that differences in fuel type prevalence between ComStock and CBECS on an area basis result in differences in the amount of floor area served by HVAC equipment for that fuel type in ComStock models compared to CBECS.
 
 %figures moved to appendix
 
@@ -215,6 +224,9 @@ \subsection{Unoccupied Air Handling Unit Operation}
 An industry-provided BAS data set of over 5,700 AHUs was used to inform the prevalence of three unoccupied AHU operation modes. The data set includes time series (hourly) BAS variables for ``Occupied Status'' (describes whether the AHU was in an occupied mode for that hour), ``Fan Status'' (describes whether the fan was used for that hour), and ``Ventilation Status'' (describes whether outdoor ventilation air was used for that hour). Counts of AHUs and buildings by building type in the data set are shown in Table~\ref{tab:unnoc_ahu_data_counts}, and the three unoccupied AHU shutdown control schemes are summarized in Table~\ref{tab:unnoc_ahu_schemes}.
 
 The data set suggests that 27\% of AHUs use scheme 1 (least efficient), 50\% of AHUs use scheme 2 (more efficient), and 23\% of AHUs use scheme 3 (most efficient; ASHRAE-90.1 required). The prevalence of the AHU unoccupied control schemes by building type is shown in Table~\ref{tab:unnoc_ahu_scheme_prev}. These probability distributions are used in ComStock sampling to set the fraction of buildings utilizing the discussed control schemes, by building type, for models that use AHU-based HVAC systems. Non-AHU HVAC system types are not applicable to this methodology, nor are building types not listed in Table~\ref{tab:unnoc_ahu_scheme_prev}. Note that building types with less than 25 buildings in the BAS data set (Table~\ref{tab:unnoc_ahu_data_counts}) use the ``All Types'' distribution of the data set at large, as fewer than 25 samples cannot reliably be used to represent a population.
+
+The following building types are not included in the unnocupied air handling unit operation workflow, and utilize default scheduling only: small hotels, large hotels, outpatient, hospitals, primary schools, and secondary schools. The building types may be integrated into this workflow in the future as more data becomes available.
+
 \input{tables/unnoc_ahu_data_counts}
 \input{tables/unnoc_ahu_schemes}