EN 17888-2:2024

SLOVENSKI STANDARD
01-julij-2024
Toplotne značilnosti stavb - Preskušanje gradbenih preskusnih struktur na mestu
vgradnje - 2. del: Analiza podatkov v stanju dinamičnega ravnovesja za preskus
skupnih toplotnih izgub
Thermal performance of buildings - In situ testing of building test structures - Part 2:
Steady-state data analysis for aggregate heat loss test
Thermische Leistung von Gebäuden - In-situ-Tests von Gebäudeteststrukturen - Teil 2:
Steady-State-Datenanalyse für Gesamtwärmeverlusttest
Performance thermique des bâtiments - Mesurage in-situ de bâtiment d’essai - Partie 2 :
Analyse des données en régime permanent pour le test de perte de chaleur globale
Ta slovenski standard je istoveten z: EN 17888-2:2024
ICS:
91.120.10 Toplotna izolacija stavb Thermal insulation of
buildings
2003-01.Slovenski inštitut za standardizacijo. Razmnoževanje celote ali delov tega standarda ni dovoljeno.

EN 17888-2
EUROPEAN STANDARD
NORME EUROPÉENNE
May 2024
EUROPÄISCHE NORM
ICS 91.120.10
English Version
Thermal performance of buildings - In situ testing of
building test structures - Part 2: Steady-state data analysis
for aggregate heat loss test
Performance thermique des bâtiments - Essais in situ Wärmetechnisches Verhalten von Gebäuden - In-situ-
des structures de bâtiments d'essai - Partie 2 : Analyse Messung an Bauwerksprüfkörpern - Teil 2:
des données en régime stationnaire pour l'essai de Auswertung stationärer Daten für die Prüfung des
déperdition thermique globale Gesamtwärmeverlustes
This European Standard was approved by CEN on 27 February 2024.

CEN members are bound to comply with the CEN/CENELEC Internal Regulations which stipulate the conditions for giving this
European Standard the status of a national standard without any alteration. Up-to-date lists and bibliographical references
concerning such national standards may be obtained on application to the CEN-CENELEC Management Centre or to any CEN
member.
This European Standard exists in three official versions (English, French, German). A version in any other language made by
translation under the responsibility of a CEN member into its own language and notified to the CEN-CENELEC Management
Centre has the same status as the official versions.

CEN members are the national standards bodies of Austria, Belgium, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia,
Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Norway,
Poland, Portugal, Republic of North Macedonia, Romania, Serbia, Slovakia, Slovenia, Spain, Sweden, Switzerland, Türkiye and
United Kingdom.
EUROPEAN COMMITTEE FOR STANDARDIZATION
COMITÉ EUROPÉEN DE NORMALISATION

EUROPÄISCHES KOMITEE FÜR NORMUNG

CEN-CENELEC Management Centre: Rue de la Science 23, B-1040 Brussels
© 2024 CEN All rights of exploitation in any form and by any means reserved Ref. No. EN 17888-2:2024 E
worldwide for CEN national Members.

Contents Page
European foreword . 3
Introduction . 4
1 Scope . 5
2 Normative references . 5
3 Terms, definitions and symbols . 5
3.1 Terms and definitions . 5
3.2 Symbols . 7
4 Principle . 9
5 Uncertainty. 9
6 Input data. 10
6.1 Raw data . 10
6.2 Irregularities and gaps in the data . 10
6.3 Cleaning data . 11
6.4 Filtering (averaging) . 11
6.5 Checking averaged data . 12
7 Data analysis . 12
7.1 General. 12
7.2 Simple linear regression using the Siviour method . 13
7.3 Multiple linear regression (MLR) techniques. 14
7.3.1 General. 14
7.4 Validation: residuals analysis . 14
7.5 Normality of residuals . 14
7.6 Autocorrelation test . 15
8 Test report . 16
8.1 General. 16
8.2 Data on the measured building/ structure . 17
8.3 Description of the experimental set-up . 17
8.4 Conditions during measurement . 17
8.5 Data pre-processing . 18
8.6 Aggregate heat transfer coefficient and associated uncertainties estimation . 18
8.7 Supplementary and supporting measurements . 19
Annex A (normative) Limitations and sources of errors . 20
Annex B (normative) Process for estimating experimental uncertainty . 24
Annex C (normative)  Data analysis methods . 30
Annex D (informative) Example of building heat loss test data analysis. 39
Annex E (informative) Practical recommendations . 49
Bibliography . 51

European foreword
This document (EN 17888-2:2024) has been prepared by Technical Committee CEN/TC 89 “Thermal
performance of buildings and building components”, the secretariat of which is held by SIS.
This European Standard shall be given the status of a national standard, either by publication of an
identical text or by endorsement, at the latest by November 2024, and conflicting national standards shall
be withdrawn at the latest by November 2024.
Attention is drawn to the possibility that some of the elements of this document may be the subject of
patent rights. CEN shall not be held responsible for identifying any or all such patent rights.
Any feedback and questions on this document should be directed to the users’ national standards body.
A complete listing of these bodies can be found on the CEN website.
According to the CEN-CENELEC Internal Regulations, the national standards organisations of the
following countries are bound to implement this European Standard: Austria, Belgium, Bulgaria, Croatia,
Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland,
Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Norway, Poland, Portugal, Republic of North
Macedonia, Romania, Serbia, Slovakia, Slovenia, Spain, Sweden, Switzerland, Türkiye and the United
Kingdom.
Introduction
EN 17888-1 describes a test methodology that enables the actual in situ building test structure aggregate
heat loss (building heat transfer coefficient) to be quantified. This test method is termed the aggregate
heat loss test method. This document principally covers numerical methods based on steady-state linear
regression techniques. The results obtained using these methods are only valid under the assumption
that, in first approximation, the data can be described by these mathematical and physical laws. Statistical
tests to check the validity of these assumptions are therefore given. It also results in the determination of
an aggregate building heat transfer coefficient for the tested building structure, along with the
uncertainty associated with this coefficient. Both the aggregate building test structure heat transfer
coefficient and its uncertainty can be calculated as an output of this document. The reporting format
relating to the test data and the resulting analysis is also described.
This document is highly linked with EN 17888-1 to which it applies exclusively. It is also complimentary
to EN 17887-1 which deals exclusively with completed buildings.
In first instance, real building co-heating tests and associated data analysis help on determining the global
performance of the building that usually take advantage of free solar gains through well oriented glazing
surfaces. In this case these solar gains are welcome and contribute in reducing the energy demand for
heating of the building. Aggregate heat loss coefficient extracted from data analysis on such real building
is then minimized by associated solar gains and may help to better understand the real energy demand
of the studied building structure regarding weather conditions.
In second instance it is from interest (as example) to concentrate on aggregate thermal performance of
opaque building structures, for more precisely undertake the analysis of the thermal response of the
building structure linked to these same climatic patterns. For that purpose, direct solar gains through
glazing surfaces are excluded of the study by testing preferatelly opaque structures. In most of the cases,
in winter periods, solar gain through opaque insulated surfaces remain poor so that the energy demand
for heating become mostly dependent on temperature difference between internal and external
environments. Nevertheless, this also offers the opportunity to undertake in situ testing with the aim to
evaluate efficiency of passive solar systems that would contribute to minimize the energy demand of the
tested building structure.
This document describes the input data required to undertake the analysis, various statistical methods
that can be used to analyse the data, the uncertainty associated with the measurements, and the reporting
format.
Detailed requirements concerning the test procedure and the data recording are specified in EN 17888-1.
1 Scope
This document specifies the steady-state data analysis methods to evaluate the data from ‘the aggregate
heat loss test method’. These analysis methods enable the actual in situ aggregate heat loss (building heat
transfer coefficient) to be estimated.
NOTE The aggregate heat loss method is specified in EN 17888-1:2024.
2 Normative references
The following documents are referred to in the text in such a way that some or all of their content
constitutes requirements of this document. For dated references, only the edition cited applies. For
undated references, the latest edition of the referenced document (including any amendments) applies.
ISO 5479, Statistical interpretation of data — Tests for departure from the normal distribution
EN 17888-1:2024, Thermal performance of buildings — In situ testing of building test structure — Part 1:
Data collection for aggregate heat loss test
3 Terms, definitions and symbols
For the purposes of this document, the following terms and definitions apply.
ISO and IEC maintain terminological databases for use in standardization at the following addresses:

— IEC Electropedia: available at https://www.electropedia.org/
— ISO Online browsing platform: available at https://www.iso.org/obp
3.1 Terms and definitions
3.1.1
aggregate heat transfer coefficient
sum of the transmission and infiltration component of the ventilation heat transfer coefficient based upon
measurement according to this test standard
3.1.2
external (internal) air temperature
temperature of the external (internal) air measured by external (internal) air temperature sensor
3.1.3
heat transfer coefficient
heat flow rate divided by temperature difference between two environments; specifically used for heat
transfer coefficient by transmission or ventilation
[SOURCE: EN ISO 13789:2017, 3.5]
3.1.4
internal room temperature
air temperature measured at the geometric centre of the room
3.1.5
internal whole building temperature
mean air temperature of all of the measured internal room temperatures
3.1.6
solar heat gain
heat provided by solar radiation entering, directly or indirectly (after absorption in building elements,)
into the building through windows, opaque walls and roofs, or passive solar devices such as sunspaces,
transparent insulation and solar walls
Note 1 to entry: Active solar devices such as solar collectors are considered part of the technical building system.
[SOURCE: ISO 52000-1:2017, 3.6.10]
3.1.7
global solar irradiance
measured or calculated solar irradiance related to the south vertical wall or façade that is expected to
receive the highest proportion of solar gains
Note 1 to entry: Longwave and shortwave radiation exchanges to the sky are not included
3.1.8
temperature difference
difference between the internal whole building temperature and external air temperature
3.1.9
transmission heat transfer coefficient
heat flow rate due to thermal transmission through the fabric of a building, divided by the difference
between the environment temperatures on either side of the construction
Note 1 to entry: By convention, if the heat is transferred between a conditioned space and the external environment,
the sign is positive if the heat flow is from the space to outside (heat loss).
[SOURCE: EN ISO 13789:2017, 3.6]
3.1.10
ventilation heat transfer coefficient
heat flow rate due to air entering a conditioned space by infiltration or ventilation, divided by the
temperature difference between the internal air and the supply air temperature
Note 1 to entry: The supply temperature for infiltration is equal to the external temperature.
Note 2 to entry: In this analysis, the intended ventilation component of the ventilation heat transfer coefficient is
typically omitted and only the infiltration component is included in the heat transfer coefficient, as intended
ventilation routes are sealed during the test.
[SOURCE: EN ISO 13789:2017, 3.7, modified – Note 2 to entry has been added]
3.2 Symbols
Table 1 summaries the symbols and units referred to within this standard.
Table 1 - Symbols and units
Symbol Description Unit
Inputs for the regression analyses
P Electrical heating power W
h
P
Electrical heating power W
h
Global solar irradiance (measured or calculated solar irradiance related to the south W/m
q
sw
vertical wall or façade that is expected to receive the highest proportion of solar gains)
T Internal air temperature °C
i
T External air temperature °C
e
ΔT Temperature difference between the internal whole building and external air K
Parameters
H Heat transfer coefficient (global, including losses by transmission and ventilation or air W/K
infiltration)
H Aggregate heat transfer coefficient W/K
agg
H Transmission heat transfer coefficient W/K
tr
H Ventilation heat transfer coefficient W/K
v
H Internal gain coefficient W/K
i
H External gain coefficient W/K
e
B Bias term W/K
A Equivalent solar aperture m
sw
θ Vector of the parameters of the model -
X Explicative matrix of the model -
Y Vector of the explicated values of the model -
Intermediary quantities
B’ Intercept for the inverted linear regression W/m
B’ Slope for the inverted linear regression m
β* Dimensionless slope -
A, H, d, Internal parameter -
λ ,λ
1 2
SF Scale factor -
λ Lagrange weight coefficient -
Symbol Description Unit
Common statistical notation
r Pearson correlation coefficient -
r Coefficient of determination -
ε Residuals of the model W or
W/K
n−1
Student coefficient for a bilateral Student law with (n-2) freedom degrees -
t
α
1−
Fisher coefficient for v = 1 and α = 5 %, with v = (n-2) freedom degrees -
1 2
F
α;vv;
( )
n Number of points used for the regression -
n Number of sensors deployed -
sensors
n Number of party walls
party
wall
Var(θ) Variance-covariance matrix for the vector of parameters of the model -
ACF Autocorrelation function -
Other index and notation
B Bolted: table, vector or matrix -
τ Transposed B matrix -
B
-1 Inverted B matrix -
B
x* Dimensionless -
ˆ
Best estimate of the random variable X -
X
Average of the random variable X -
X
Var(X) Standard variance of the random variable X -
Cov(X,Y) Covariance between random variables X and Y -
s( ) Experimental standard deviation of the random variable X -
X
u( ) Standard uncertainty of the X quantity -
X
U( ) Enlarged uncertainty of the X quantity (with a 95 % bilateral confident interval) -
X
k Coverage factor used for expanded uncertainty (k = 2 for 95 % confidence interval) -
4 Principle
The building aggregate heat loss analysis shall be calculated by the following energy balance equation,
assuming that the heating power input ( P ) is balanced by thermal losses and solar radiation transfer:
h
P H⋅−TT− A⋅ q
( )
h agg i e sw sw
where
P is the electrical energy supplied by heaters and dissipated by fans [W];
h
H is the aggregate heat transfer coefficient, which combines the transmission
agg
heat transfer coefficient and the infiltration component of the ventilation heat
transfer coefficient [W/K];
T - T = ΔT is the temperature difference between inside and outside air [K];
i e
A 2
sw is the solar aperture [m ];
q 2
sw is the measured or calculated solar irradiance [W/m ].
The aggregate heat transfer coefficient H and the solar aperture A express the relationship between
agg sw
the electrical heating power P , q and ΔT variables in the regression analysis.
h sw
The measured solar irradiance should be obtained in accordance with EN 17888-1:2024 6.2.1.
The solar aperture is a parameter that describes the solar gains incident the building as a whole, including
diffuse gains, direct gains through glazing of all orientations and those from solar radiation incident upon
opaque elements. It can be read as a global solar aperture, which includes the influence of non-
perpendicular incidence of solar radiation, geometry and orientation of the building structure (including
possible shading), solar absorption at opaque surfaces, solar energy transmittance factor and glazing
surface of the building envelope. A can be estimated experimentally from the whole building aggregate
sw
heat loss test data by regression analysis.
The aggregate heat transfer coefficient H can be identified as a mix between heat losses by
agg
transmission and, because ventilation openings are sealed during the test, the infiltration component of
the ventilation through the envelope of the tested building, H .
v
H =HH+=H HH+
agg tr v tr v
The aggregate heat transfer coefficient can be deduced from quasi-steady-state measurements using the
energy balance equation. Assuming this energy balance model holds, several identification techniques,
all based on linear regression methods, can be used to estimate the parameters of interest. Specifically,
simple or multiple linear regression techniques can be applied on whole building aggregated heat loss
test measurement data (time averaged data points for P , q and ΔT).
h sw
5 Uncertainty
The accuracy, reproducibility and interpretation of heat transfer coefficient estimates are limited by
several factors related to both experimental and statistical uncertainties. The margin of uncertainty in
the building aggregated heat loss test results is associated with a number of limitations and sources of
error listed in Annex A Attention shall be given to these limitations and sources of error listed in Annex A
within both data collection and data analysis. Remaining uncertainty within estimates of the heat transfer
coefficient shall then be estimated and stated alongside any results.
=
Two methods of estimating uncertainty are presented in this document:
a) experimental uncertainty shall be estimated based upon the GUM method (JCGM 100:2008) and can
incorporate both Type A (statistical analysis of observations) and Type B (non-statistical analysis,
e.g. prior knowledge, previous experiments etc.), providing a stated uncertainty for an estimate of
the heat transfer coefficient. An example of this analysis is given in Annex B;
b) statistical uncertainty shall be estimated based on the residuals between best estimates and
measured data points. Statistical uncertainty is defined in Annex B and shall be used to determine
the most appropriate regression method. Statistical uncertainty can be combined with experimental
uncertainty to provide an increased error estimate (8.6).
The aggregate heat transfer coefficient, H [W/K], the solar aperture A [m ], and associated
agg
sw
uncertainties, shall be indicated in the report according to 8.6.
6 Input data
6.1 Raw data
The raw data set requires a minimum of 15 days of continuous recordings without gaps. Measurements
in the raw data set shall be non-biased and meet these requirements:
— sensors shall be calibrated in order to correct constant error (non-biased sensors);
— measurements are representative of the physical parameter. For instance, internal temperature
measurements shall be averaged from several non-biased temperature sensors in order to represent
the spatial dispersion due to air stratification.
NOTE Requirements to reach these conditions are described in EN 17888-1.
6.2 Irregularities and gaps in the data
An analysis of the raw data are key to reduce difficulties in the subsequent modelling and provide
meaningful results. This analysis shall aim at pointing out unusual phenomena, so-called irregularities,
as well as outliers, measurement errors and missing data. These issues are often introduced either in the
experiment set-up, by the measuring apparatus, or the data handling. A list of often-encountered
phenomena found in the raw data set that can introduce problems, such as nonlinearity and outliers in
the modelling and estimation step, is given in Annex A.
It is firstly required to plot the raw data set (for all input variables) as a function of the time step as
recorded in EN 17888-1, in order to check for outliers, measurement errors, missing data and any
potential irregularities.
NOTE Scatter plots of pairs of measurements like energy consumption vs temperature difference can also be
insightful to show patterns and correlations between variables.
All these irregularities shall be removed from the raw data set. In the end, the quality of the remaining
data set obtained (cleaned data set) shall be tested so that:
— in total, all irregularities shall not represent more than 10 % of the global amount of data;
— irregularities for each data input (i.e. T , T , q and P ) shall not represent more than 15 % each (i.e.
i e sw h
no more than 15 % irregularities for T data set, even if there are no other irregularities for T , q
i e sw
and P );
h
— time without data (or with irregularities) shall not exceed 50 min (based upon a 24-h aggregation
period) or 4 % for the P measurement, as this will be measured as accumulated flow using an energy
h
meter;
— time without data (or with irregularities) shall not exceed 3 h for T , T , and q measurements. If
i e sw
any of the above irregularities are exceeded, then the test is deemed to be invalid.
Any irregularity removed shall be recorded in the report, along with a brief justification of the reasons
for their removal (see 8.5).
6.3 Cleaning data
Often, the data recorded from the beginning of the test will need to be removed, as the dwelling will be in
the process of being heated-up and the thermal mass charged. Once the thermal mass of the building has
charged, the electrical energy input into the building should stabilize and only be influenced by external
environmental conditions.
Any experimental overheating periods due to uncontrolled heat input from solar gains shall be excluded
to maintain a constant mean internal temperature during the building heat loss test.
Irregularities on T , T , and q shall be corrected by linear interpolations and recorded in the report,
i e sw
along with a brief justification of the reasons for correction and description of the interpolation method
(see 8.4).
6.4 Filtering (averaging)
The effects of dynamic behaviour shall be minimized by low pass filtering the time series;, by resampling
the time series - i.e. aggregating the measured data points into longer time intervals (e.g. 24 h). The
appropriate resampling interval depends on how fast the building fabric responds to external
environmental conditions: for insulated buildings, one-day (24-h) averages are usually appropriate,
whereas for high performance (well-insulated or heavyweight) buildings a higher resampling interval
may be needed.
The following procedure shall be undertaken to select an appropriate resampling interval:
a) start with a short resampling interval (i.e. 24-h averages), check averaged data (see 6.5) and apply
statistical analysis (see Clause 7);
b) analyse residuals for autocorrelation and check that the cross-correlation to the inputs, especially to
solar radiation, is not significant (see model validation step in Clause 7);
c) if residuals are not autocorrelated and there is no significant cross-correlation, stop the resampling
and provide the results (see Clause 8);
d) if residuals are autocorrelated or there is significant cross-correlation, increase the resampling time
by 24 h until white noise residuals are obtained (see 7.6);
e) the same averaging interval shall be used for all the signals.
It is important that an integer of 24 h is used to minimize bias from dynamic effects. Additionally,
although a daily average using the time period 00:00 to 23:59 is common, a time period of 06:00 to
05:59 should be considered and is often more appropriate, as it allows more time for solar gains to
remerge from the thermal mass of a tested building within the same aggregation period. Dawn-to-dawn
or a similar interval can also be considered.
A simple check of raw 24-h total power input against average 24-h temperature difference across the
valid test period can be useful as a check on the regression process itself, particularly when there is not a
wide spread in data points.
If resampling is applied, the resampling interval shall be recorded in the report (see Clause 8).
6.5 Checking averaged data
Linear regression shall be applied only for “regular” groups of points after averaging over integer 24-h
periods (i.e. not applied to data sets at shorter time intervals – e.g. 1-h data), meaning that outliers can be
removed. Checking outliers is simpler by using Siviour analysis (even if not used to determine the
parameters of interest in Clause 7). It is then required to plot, for each averaged data, Y as a function of X
where:
q P
 
sw,1 h1,
 
∆T ∆T
1 1
 
 
X= Y=
and
q P
sw,n h,n
 
∆T ∆T
n n
 
 
All outliers identified shall lead to the removal of the corresponding averaged data from the data set used
for statistical analysis. The outlier removed shall be recorded in the report, along with a brief justification
of the reasons for their removal (see Clause 8).
7 Data analysis
7.1 General
The data analysis techniques within Annex C shall be used.
The parameters of interest shall be determined by applying simple or multiple linear regression
techniques according to Table 2, based on the energy balance equation, such as:
Table 2 — Regression techniques
Linear regression
Description
technics
This simple method of linear regression is driven by normalized
measurement uncertainties on variables Y and X (see 7.2).
Simple linear
There are three cases (see Annex C) where residuals are used to apply
regression using the
the least square method between experimental points and estimations:
Siviour method
— the vertical distance;
(see 7.2)
— the horizontal distance;
— the orthogonal distance.
Multiple regression is a more complex broader class of regression that
encompasses linear and nonlinear regressions with multiple
explanatory variables.
Multilinear regression
Two options are proposed (see 7.3.1, Annex C):
(see 7.3)
— biased energy balance model;
— unbiased energy balance model.
Linear regression
Description
technics
Although a unbiased energy balance model (e.g. zero intercept) can
represent a simplification of real world processes, it is typically adopted
over biased (non-zero intercept) models.
NOTE Methods to calculate and correct for solar gains based upon measured solar radiation and assumed
building characteristics (e.g. glazed areas, g-values) are omitted from this document due to the significant
uncertainty introduced by assumptions regarding building and glazing properties, alongside increased significance
for accurate and complete solar radiation measurements.
7.2 Simple linear regression using the Siviour method
The Siviour analysis of building heat loss test data shall be used to account for the effect of solar gains in
the estimation of the aggregate heat transfer coefficient. The method consists of undertaking a linear
regression of the electrical heating power, P , divided by the daily mean air temperature difference
h
between the indoor and the external ambient ΔT (dependent variable), plotted on the y-axis, against the
daily mean global solar irradiance q divided by the daily mean air temperature difference between the
sw,
indoor and the external ambient ΔT (independent variable), plotted on the x-axis. From this analysis, the
y-intercept of the regression line represents the aggregate heat transfer coefficient of the building, H
[W/K], whilst its gradient is the equivalent solar aperture, A [m ], see Figure 1:
sw
P q
h sw
= HA−⋅
agg sw
∆∆T T
Key
data points
linear regression through data points

designed heat loss coefficients

Figure 1 — Example of linear regression analysis
7.3 Multiple linear regression (MLR) techniques
7.3.1 General
MLR can be carried out between the average electrical heating power P [W] and two independent
h
variables, namely the temperature difference ΔT [K] and the solar irradiance q [W/m ]. This allows
sw
the aggregate heat loss coefficient H [W] and the equivalent solar aperture A [m ] to be estimated
agg sw
through regression.
The energy balance equation (see Clause 4) is based on the hypothesis that the daily energy balance is
not biased. This means that when daily internal and external air temperatures are the same (T = T ) and
i e
when there is no solar radiation (q = 0), the daily heat power is assumed to be exactly zero.
sw
It is important to note that the energy balance equation represents a simplified energy balance model
and not all real world effects are captured (see Annex A). Such effects can be represented by the
introduction of a statistical bias or intercept term ( B ). In most cases, a bias term is unlikely to yield
improved results and a un-biased, zero intercept model shall be used.
Biased, non-zero intercept model:
P= B+ H ⋅∆T− A⋅ q
H 0 agg sw sw
Un-biased ( B = 0 ), zero intercept model:
P H⋅∆T− A⋅ q
H agg sw sw
7.4 Validation: residuals analysis
Once the data analysis is completed, validation checks shall be carried out to estimate the quality of the
regression and the appropriateness of the averaging interval used. The results of validation tests shall be
stated in the report (see 8.5).
Time series plots of residuals shall be inspected to see if any clear patterns are present. If patterns occur,
it can indicate that the model is oversimplified and that additional physical mechanisms shall be included.
Specifically, tests concerning Gaussian residual hypothesis and the lack of autocorrelation shall be
undertaken.
7.5 Normality of residuals
The variability of the residuals shall be Gaussian. To validate this hypothesis a Shapiro-Wilk test will be
applied on the residuals according to ISO 5479.
It is a good practice to check if the scatter plot is bivariate normal, meaning that distribution in both x
and y directions is approximately Gaussian. An analysis of the scatter plot is generally enough to validate
the bivariate hypothesis. Although, it is difficult to validate in practice because the number of points is
quite limited, it is required to check whether the scatter plot is not split into distinct group of points (see
Figure 2).
=
a) “ideal” case (perfectly b) “acceptable” case c) “Pathological” case
bivariate)
Key
data points
q / ΔT [W/(m K)]
sw
P /ΔT [W/K]
h
Figure 2 — Bivariate scatter plot
Linear regression technics formally imply that the residuals are independent of time (homoscedasticity
hypothesis). So, it is a good practice to check if the residuals are homoscedastic. This test could be carried
out by plotting residuals as a function of time. The absolute value of residuals shall not vary with time, as
described in the following examples (see Figure 3):

a) Homoscedasticity b) Heteroscedasticity c) Heteroscedasticity
(constant variance) (non-constant variance) (nonlinear data/wrong
model)
Key
data points
Observation order
Residuals
Figure 3 — Residuals as function of time
7.6 Autocorrelation test
The autocorrelation test shall be carried out on the timeseries of the residuals to ensure that they are
white, i.e. not correlated at different time intervals (‘lags’), and therefore the model used is adequate. The
autocorrelation function (ACF) is defined in the following formula:
lag = correlation xx,
( )
k t tk+
For non-zero lags, the ACF shall be insignificant or, more specifically, not be significantly different from
that of white noise. This implies that there shall be no systematic pattern in the ACF, and hence not more
than 5 % to 10 % of the lag correlations shall be above the 95 % confidence bands for white noise (Figure
4). This is calculated as follows:
n−1
Ut= Var ACF
( )
ACF
α
1−
or
1 2
U =−±
ACF
n
n
where
n is the number of data points used in the calculation of the ACF.
EXAMPLE
Key
Lag
Autocorrelation function
Figure 4 — Autocorrelation function (Y) for each lag (X) and 95 % confidence interval
If this condition is not fulfilled, the regression process shall be restarted by increasing the sampling
interval (i.e. from 1-day average to 2-days average). If a greater sampling interval does not change the
conclusion, the report shall indicate the result at the previous one and state that the validation test failed.
See Annex D for an example of a whole building heat loss test dtata analysis.
Annex E includes some useful practical recommendations.
8 Test report
8.1 General
The test report shall include the information listed in 8.2 to 8.8.
The test report will also include the following information:
the European and International Standard used (including its year of publication)
any unusual features observed.
8.2 Data on the measured building/ structure
— Building location, form, construction type, and number of storeys;
— floor plans and orientation of the building;
— approximate building age;
— building characteristics: floor area, total envelope area, glazed area, party wall/party floor areas;
— main construction materials used (if known) or assumed;
— results of the pre- and post-test fan pressurization test results.
8.3 Description of the experimental set-up
— Specifications (e.g. make, model, accuracy) of sensors used;
— locations of internal sensors, electrical heaters and mixing fans;
— locations of external sensors (including orientation of solar radiation measurement);
— interval between records in the raw data and number of measurements averaged or sampled.
— the magnitude of any estimated unrecorded energy inputs.
8.4 Conditions during measurement
— Information on the test protocol as described in Part 1 of this document;
— state if data were collected in full accordance with Part 1 of this document and explain any deviation;
— date and time at the beginning and end of measurement period (entire data collection period);
— data presented as a timeseries or tabular form, at a maximum of hourly intervals for:
— internal conditions: air temperatures in all zones, electrical heating power, relative humidity, air
temperatures in any unheated adjoining spaces (e.g. attics, basement, voids);
— external conditions: air temperature, global solar irradiance and any additional measurements
(e.g. relative humidity, wind speed, wind direction);
— details of the building set-up during the measurement period:
— intended ventilation routes sealed/closed;
— curtains, blinds or shutters opened/closed;
— drains and water traps filled/sealed/unsealed;
— internal doors held open.
8.5 Data pre-processing
— Period of data used for analysis (including data omitted from analysis);
— any irregularities and gaps in the data (see Clause 6.2);
— any data cleaning (see Clause 6.3);
— clarify which data has been removed and the justification;
— the process used for any manipulation of the measured solar data (e.g. to transform global horizontal
to vertical);
— averaging period used for data analysis, e.g. 24 h, 48 h (see Clause 6.4);
— averaging time interval used, e.g. 06:00 – 06:00 or 24:00 – 24:00;
— a table of the averaged input data set, i.e. averaged heat power P , average internal and external
h
temperatures T and T , and average sun irradiation q .
i e sw
8.6 Aggregate heat transfer coefficient and associated uncertainties estimation
— A graph showing the scatter of P against ΔT prior to any linear regression;
h
— a graph showing the scatter and the fitted line, y = f(x), after the linear regression;
— as a minimum, all test results shall present the estimated value of H and A coefficients
agg sw
calculated according to the Siviour method described in 7.2. Normalized standard deviations on x
and y variables, and consequently what linear regression method has been used (see Clause 8.8);
— results of the Shapiro-Wilk and autocorrelation function test;
— the aggregate heat transfer coefficient, H [W/K] and associated uncertainties, shall be indicated
agg
in the report, using the following conventions:

HU±
agg H agg
for example: H = 150,3 ± 12,9 W/K
agg
where H is the best estimate of the aggregate heat transfer coefficient and U is the expanded
agg
H
uncertainty at 95 % confidence intervals. It shall be estimated using the measurement, u
mes
(Annex C) and model uncertainty u (Annex C);
model
— measurement uncertainties on each input (P , T , T and q ) quantified according to (Annex B) of
h i e sw
this document.
8.7 Supplementary and supporting measurements
Details of any supporting or supplementary measurements (e.g. air permeability, infiltration rate via
tracer gas decay or continuous concentration methods, in situ U-value measurements, thermography,
humidity etc.) for any methods include:
— methodology and reference standard used to carry out the test;
— a full copy of the testing report, in line with the methodology used;
— uncertainty in measurement.
3 2
— Airtightness test result in m /(h.m ), with associated pressure difference of the test measurement
e.g. at 50Pa;
— minimum, maximum, and average external wind speed, external air temperature, and temperature
difference.
Annex A
(normative)
Limitations and sources of errors
A.1 General
The accuracy, reproducibility and interpretation of the aggregate heat transfer coefficient estimates are
limited by several factors related to both experimental and model uncertainties.
This Annex details uncertainties related to the experimental data collection required for an aggregate
heat transfer coefficient estimate. These can be reduced or better characterized through adoption of the
testing protocol set out within EN 17888-1. Operative experience and the potential for operative error
shall be reported according to Clause 8.
A.2 Limitations and errors due to experimental uncertainties
A.2.1 Temperature measurements
Uncertainties related to temper
...