Merge 793e2b6 into 6e5854d

docs/index.rst

@@ -16,6 +16,15 @@ Welcome to oemof.thermal's documentation!
    examples
 
 
+.. toctree::
+   :maxdepth: 1
+   :caption: Model validation
+
+   validation_compression_heat_pumps_and_chillers
+   validation_concentrating_solar_power
+   validation_solar_thermal_collector
+   validation_stratified_thermal_storage
+
 .. toctree::
    :maxdepth: 1
    :caption: User's guide

docs/validation_compression_heat_pumps_and_chillers.rst

@@ -0,0 +1,246 @@
+.. _validation_compression_heat_pumps_label:
+
+Compression Heat Pumps and Chillers
+===================================
+
+Scope
+_____
+
+The validation of the compression heat pump and chiller has been conducted within the
+`GRECO project <https://github.com/greco-project>`_.
+Monitored data of the two components in combination with PV without storage has been provided by Universidad
+Politécnica de Madrid (UPM).
+The set of data contains amongst others external and internal temperatures of the components and a calculated
+COP / EER value.
+** TODO: what technology - air to air? **
+
+Method
+_______
+
+In order to calculate the COP and EER using oemof.thermal the temperature of the heat source, the temperature of the
+heat sink and the quality grade are required. The quality grade describes the relation between the actual
+coefficient of performance and the coefficient of performance of the Carnot process. Please see the
+`USER'S GUIDE <https://oemof-thermal.readthedocs.io/en/latest/compression_heat_pumps_and_chillers.html>`_
+on compression heat pumps and chillers for further information.
+
+The monitored coefficients from UPM are compared with the coefficients calculated using different quality grades
+to evaluate which quality grade fits best the examined chiller and heat pump. For the heat pump, the temperature of the
+external input into the evaporator is the heat source and the temperature of the internal output from the condenser
+is the heat sink. In case of the chiller the heat source is the external temperature input into the condenser
+and the heat sink the internal temperature output from the evaporator. The monitored coefficients are calculated as the
+ratio between the thermal capacity and the electrical capacity.
+
+The data set contains data points where the solar modules provide an electrical power less than 100 watts and where the
+compressor of the installation is turned off. These data points are excluded from calculations since they differ from
+the ones under operational behavior of the component. For this purpose the data is preprocessed in order to attain
+only data points with electrical power greater or equal to 100 watts and with the integral fan turned on.
+In comparison to the chiller's monitored data, the data of the heat pump contains a higher number of excluded data
+points. It further has multiple downtimes, which lead to reversed temperatures of heat sink and heat source.
+
+The functionalities for calculating the COP and EER of oemof.thermal are made for a stationary process, while the data
+provided by UPM includes mostly data from non-stationary periods as different control modes are explored. To balance the
+fluctuating values we decided to analyze average hourly values.
+
+Various types of charts are used for the validation of the calculated COP and EER. For the validation the residual,
+which corresponds to the difference of the monitored and calculated coefficients is used. For both, chiller and heat
+pump, correlations with the residual are shown in various types of charts: the histograms, the correlation between
+calculated and monitored coefficients, the root mean square error (RMSE), the relation between the residuals and
+temperature hub as well as the relation between residuals and monitored coefficients.
+
+Results of the chiller
+______________________
+
+Typical EER of chillers used for cooling are around 4 to 5 [1]. By adhering to these reference values we conclude
+that EERs with quality grades ranging from 0.25 to 0.4 give fitting results.
+
+The RMSE for the validated quality grades presents the standard deviation of the residuals. Among the range of quality
+grades for the chiller, the RMSE is smallest for the quality grade 0.30 with 0.573 and larger for 0.25 with 0.758 and
+for 0.35 with 1.181 (cf. Tab.1).
+
+============================= =============================
+    Quality grade                   RSME
+============================= =============================
+    0.05                            3.831
+    0.10                            3.030
+    0.15                            2.236
+    0.20                            1.460
+    0.25                            0.758
+    0.30                            0.573
+    0.35                            1.181
+    0.40                            1.943
+    0.45                            2.732
+    0.50                            3.531
+============================= =============================
+Tab.1: Root square mean error for different quality grades of the chiller
+
+The correlation with quality grades below 0.30 show an overestimation of the coefficients. In contrast,
+calculated EER at quality grades above 0.30 indicate an underestimation. Fig.1 shows the estimated correlation of EERs
+in the middle together with over- and underestimation to the left and right:
+
+.. figure:: _pics/Correlation_EERs.png
+    :width: 100 %
+    :alt: Correlation_EERs.png
+    :align: center
+    :figclass: align-center
+
+    Fig.1: Correlation between monitored and calculated EER with overestimation showing a quality grade of 0.25 (left),
+    quality grade of 0.30 with least error (middle) and with underestimation connected to a quality grade of 0.35
+    (right).
+
+Fig.2 shows the residual over monitored EER for quality grades of 0.25, 0.30 and 0.35. In Fig.3 the residual is plotted
+over the temperature hub for the three quality grades. From both graphs can be derived that the residual is minimal
+for a quality grade of 0.30. Furthermore they indicate a dependence of the residuals to both parameters. Smaller
+temperature hubs cause larger residuals, while larger temperature differences lead to smaller residuals. In general,
+residuals decrease with rising quality grades.
+
+.. figure:: _pics/residual_eer.png
+    :width: 100 %
+    :alt: residual_eer.png
+    :align: center
+    :figclass: align-center
+
+    Fig.2: Correlation between residual and monitored EER with overestimation showing a quality grade of 0.25 (left),
+    quality grade of 0.30 with least error (middle) and with underestimation connected to a quality grade of 0.35
+    (right).
+
+.. figure:: _pics/temphub_eer.png
+    :width: 100 %
+    :alt: temphub_eer.png
+    :align: center
+    :figclass: align-center
+
+    Fig.3: Correlation between temperature hub and monitored EER with overestimation showing a quality grade of 0.25
+    (left), quality grade of 0.30 with least error (middle) and with underestimation connected to a quality grade of
+    0.35 (right).
+
+The histogram in Fig.4 depicts that most of the calculated coefficients have small deviations with the quality grade of
+0.30 (middle). Based on the left graph it gets clear that the average calculated EER decreases with lower quality grades
+due to the shift to the right of the histogram. As seen in the right graph of Fig.4 the average calculated EER
+increases with higher quality grades due to the shift to the left of the histogram.
+
+.. figure:: _pics/Histogram_eer.png
+    :width: 100 %
+    :alt: Histogram_eer.png
+    :align: center
+    :figclass: align-center
+
+    Fig.4: Histogram of residuals with overestimation showing a quality grade of 0.25 (left), quality grade of 0.30
+    with least error (middle) and with underestimation connected to a quality grade of 0.35 (right).
+
+The outliers in the monitored data could be due to the start-up and shutdown of the prototypes’ compressor.
+
+An examination of the complete data set of the chiller shows a linear dependence of the residuals to the monitored EER.
+In Fig.5 this linearity can be seen for a quality grad of 0.05 (left graph) and a quality grade of 0.50 (right graph).
+It is striking that the linearity dependence is higher for smaller quality grades such as 0.05 (cf. left graph in
+Fig.5). The dispersion of residuals in areas of lower as well as higher monitored EER increases with larger quality
+grades.
+
+.. figure:: _pics/Correlation_EER_whole.png
+    :width: 75 %
+    :alt: Correlation_EER_whole.png
+    :align: center
+    :figclass: align-center
+
+    Fig.5: Correlation between residual and monitored EER of the complete data set with a quality grade of 0.05 (left)
+    and a quality grade of 0.50 (right).
+
+
+Results of the heat pump
+________________________
+
+The RMSE calculated using the heat pump's monitored data is smallest for the quality grade 0.35 with 0.991 and larger
+for 0.30 with 1.123 and for 0.40 with 1.206 (cf. Tab.1).
+
+============================= =============================
+    Quality grade                   RSME
+============================= =============================
+    0.05                            3.726
+    0.10                            3.140
+    0.15                            2.566
+    0.20                            2.015
+    0.25                            1.512
+    0.30                            1.123
+    0.35                            0.991
+    0.40                            1.206
+    0.45                            1.635
+    0.50                            2.155
+============================= =============================
+Tab.2: Root square mean error for different quality grades of the heat pump
+
+The comparison of the smallest RSME of both chiller and heat pump indicates that the monitored data of the heat pump
+contains higher deviations.
+
+.. figure:: _pics/Correlation_COPs.png
+    :width: 100 %
+    :alt: Correlation_COPSs.png
+    :align: center
+    :figclass: align-center
+
+    Fig.6: Correlation between monitored and calculated COP with overestimation showing a quality grade of 0.30 (left),
+    quality grade of 0.35 with least error (middle) and with underestimation connected to a quality grade of 0.40
+    (right).
+
+Just as with the chiller, the correlations indicate an overestimation at lower quality grades and an underestimation at
+larger quality grades.
+
+In Fig.7 shows the residual over monitored COP for quality grades of 0.30, 0.35 and 0.40. In Fig.8 the residual is
+plotted over the temperature hub for the three quality grades. From both graphs can be derived that the residual is
+minimal for a quality grade of 0.35. As in the cooler's results the dependency of residuals and both parameter is
+observable: Residuals decrease with rising quality grades.
+
+.. figure:: _pics/residual_cop.png
+    :width: 100 %
+    :alt: residual_cop.png
+    :align: center
+    :figclass: align-center
+
+    Fig.7: Correlation between residual and monitored COP with overestimation showing a quality grade of 0.30 (left),
+    quality grade of 0.35 with least error (middle) and with underestimation connected to a quality grade of 0.40
+    (right).
+
+.. figure:: _pics/temphub_cop.png
+    :width: 100 %
+    :alt: temphub_cop.png
+    :align: center
+    :figclass: align-center
+
+    Fig.8: Correlation between temperature hub and monitored COP with overestimation showing a quality grade of 0.30
+    (left), quality grade of 0.35 with least error (middle) and with underestimation connected to a quality grade of
+    0.40 (right).
+
+
+In Fig.9 the histograms of the heat pump are shown. The peak of the histograms shifts to the right with smaller quality
+grades (cf. left graph in Fig.9) and to the left with larger quality grades (cf. right graph in Fig.9).
+The values of the coefficients fluctuate more compared to the chiller.
+
+.. figure:: _pics/histogram_hp.png
+    :width: 100 %
+    :alt: Histogram_hp.png
+    :align: center
+    :figclass: align-center
+
+    Fig.9: Histogram of residuals with overestimation showing a quality grade of 0.30 (left), quality grade of 0.35
+    with least error (middle) and with underestimation connected to a quality grade of 0.40 (right).
+
+.. figure:: _pics/Correlation_COP_whole.png
+    :width: 75 %
+    :alt: Correlation_COP_whole.png
+    :align: center
+    :figclass: align-center
+
+    Fig.10: Correlation between residual and monitored COP of the complete data set with a quality grade of 0.05 (left)
+    and a quality grade of 0.50 (right).
+
+Looking at the whole preprocessed, monitored data, a linear dependence of the residuals to monitored COP values can
+be identified. The linear dependency for two quality grades 0.05 (left) and 0.5 (right) is depicted in Fig.10. Just as
+with the chiller the linearity dependence is higher for smaller quality grades such as 0.05 (cf. left graph in Fig.10).
+The dispersion of residuals in areas of lower as well as higher monitored COP increases with larger quality grades.
+
+
+References
+__________
+
+[1] Ziegler, D.-I. F. (1997).
+Sorptionswärmepumpen.
+Erding: Forschungsberichte des Deutschen Kälte- und Klimatechnischen Vereins Nr. 57
+

docs/validation_concentrating_solar_power.rst

@@ -0,0 +1,6 @@
+.. _validation_csp_label:
+
+~~~~~~~~~~~~~~~~~~~~~~~~~
+Concentrating solar power
+~~~~~~~~~~~~~~~~~~~~~~~~~
+

docs/validation_solar_thermal_collector.rst

@@ -0,0 +1,8 @@
+.. _validation_solar_thermal_collector_label:
+
+~~~~~~~~~~~~~~~~~~~~~~~
+Solar thermal collector
+~~~~~~~~~~~~~~~~~~~~~~~
+
+
+

docs/validation_stratified_thermal_storage.rst

@@ -0,0 +1,6 @@
+.. _validation_stratified_thermal_storage_label:
+
+~~~~~~~~~~~~~~~~~~~~~~~~~~
+Stratified thermal storage
+~~~~~~~~~~~~~~~~~~~~~~~~~~
+