Liens de la barre de menu commune

2.7 Validation

Numerous experts have contributed to the development, testing and validation of the RETScreen CHP Project Model. They include CHP heating modeling experts, cost engineering experts, greenhouse gas modeling specialists, financial analysis professionals, and ground station and satellite weather database scientists.

Validation of parts of the RETScreen CHP Project Model was done against other models used in the industry. The validation focused on three areas: calculation of the load duration curve (section ‎2.7.1), calculation of the heating value of biomass (section ‎2.7.2), and heating network pipe sizing (section ‎2.7.3). A more global validation is shown in section ‎2.8, Validation by an Independent Company‎.

2.7.1 Validation of load duration curve

To validate the load duration curve generated by RETScreen (see section ‎2.1.4), a comparison was made with a computer model developed by Mr. Ingvar Larsson at FVB District Energy Consultants in Sweden. Mr. Larsson’s model, hereafter named “DD-IL,” was developed using extensive records from two large and closely monitored district heating systems (St. Paul, MN, USA and Uppsala, Sweden). The RETScreen model was tested against DD-IL with data for four different cities: Edmonton, Alberta (Canada), Toronto, Ontario (Canada), St. Paul, Minnesota (USA), and Stockholm (Sweden). For all cities, degree-days data from DD-IL were used in RETScreen (rather than degree-days from the RETScreen Climate database) to eliminate artificial differences that could result from using weather data from different sources in the two programs. The only exception is for Edmonton where data from the climate database of RETScreen were used in DD-IL. Load duration curves were generated for the four cities using a 2.74 °C-d/d (1,000 degree-days annually) equivalent degree-days for domestic hot water heating, except for Uppsala where a value of 2.88 °C-d/d (1,050 degree-days annually) was used.

Table 11 compares the equivalent full load durations calculated by the two programs for the four locations. The results are very similar (less than 1% difference). Figure 18 shows the load duration curves calculated by the two programs. Again the differences are minute. For Toronto and Uppsala the two programs generate exactly the same curves. For Edmonton and Saint Paul the generated curves are very close.

2.7.2 Validation of heating value algorithm

To validate the heating value algorithm used by RETScreen (see section ‎2.6.4), its predictions were compared to findings reported in the Summer Meeting of the Technical Section, Canadian Pulp and Paper Association, Quebec, Quebec, Canada, June 6 to 8, 1955. In the paper called Determination of Bark Volumes and Fuel Properties, data was collected from thirty mills by the Forest Products Laboratories of Canada and the Federal Department of Mines and Technical Surveys. The chemical analyses (proximate and ultimate) from the samples were all performed by one laboratory. The heating values were statistically analyzed by the Forest Products Laboratories with the following results:

Age: no correlation between heating value and the age of the tree was noticeable.

Geographical area: analyses of tests did not reveal any significant differences among heating values due to area.

Species: the tests show a significant difference in the heating value among the various species in the following order (highest first): 1 – Balsam, 2 – Jack Pine, 3 – Poplar, 4 – Spruce.

Table 12 summarizes the heating values measured in the test. These values should be compared to those proposed by RETScreen for the heating value of wood waste, which range from a low of 17,723 MJ/t to a high of 19,760 MJ/t with an average of 18,673 MJ/t. The variation according to this test is +/- 3% for Jack Pine and up to –5% for Black Spruce. The estimate given by RETScreen is amply sufficient at the pre-feasibility stage of a project.

The higher heating value algorithm of RETScreen (equation (63)) was also tested against 55 samples measured by the US National Renewable Energy Laboratory (NREL) under Subcontract TZ-2-11226-1 in February 1996. Figure 19 compares measured values against values predicted by RETScreen. The average difference between the laboratory tests and the RETScreen model is 3.41% with a standard deviation of 3.75%. The difference between the results is again quite acceptable; one has to keep in mind, for example, that the typical variation in moisture content over a year for a biomass fuel can be more than 15%.

2.7.3 Validation of district heating network design

The district heating network design algorithms of RETScreen (see section ‎2.6.5) were validated with the help of ABB’s R22 computer program. The R22 computer program developed by ABB Atomic Division for sizing pipe distribution systems has been used extensively in the Scandinavian countries for design of district heating networks.

Table 13 shows pipe sizes calculated by the RETScreen model and values calculated by the R22 program. The values calculated by the two programs compare well. The RETScreen model tends to be more conservative than the R22 model. This is intentional, as the R22 model is a tool for detailed design, whereas the RETScreen is a pre-feasibility tool. The selected pipe size is also a function of how much money can be spent on the project. If money is restricted the designer typically allows for higher friction losses. The sizing is still very safe with respect to sound and erosion problems.

Theoretically the main distribution pipes should be sized with low friction losses and allow higher losses in the secondary distribution pipes to minimize required pumping load and investment costs. However, in reality it is common that space is limited and capital costs needs to be controlled resulting in a small main line. As for the secondary line it is typically oversized, as the customers heating load is not well defined and to avoid noise problems.