12th International Symposium on District Heating and Cooling

To assure that the temperatures measured in thesubstation corresponded to the average temperaturelevels in the various risers in the radiator circuits,temperature sensors were installed in two of thehouses. This enabled measurement errors ordisturbances in the radiator circuit to be identified. Theindoor temperature could be monitored thanks to sixwireless sensors installed in each house in the area.Modifications in the substationsAfter some initial tests, the circulation pumps werefound to be generally oversized to such an extent thatthe flow rate could not be decreased as much asdesired. There exists a predetermined minimumrotational speed for this type of pump, implying that thespeed could be reduced by 60–70%. Discussions withthe manufacturer revealed that the lowest pump speedcould not be changed in this model, for which reasonthe decision was made to throttle the flow with anexisting shut-off valve located after the pump, whichshifted the pump‘s operating range. The throttling wasconducted in order for the pump to give half the flowrate at 100% rotational speed. The control curve wasmodified accordingly, leading to the temperature dropin the radiator circuit becoming doubled and the heatsupply remaining unaltered.We were unable to receive a comprehensive reply fromthe pump manufacturer with respect to the possiblemeasures regarding the regulation of the pump. Adiscussion with another manufacturer implied that therewere no technical limitations for how far down thepump speed could be controlled. However, such anextension of the manoeuvrable range has so far notbeen requested. After a simple modification of thepump‘s frequency converter, the working range couldbe extended from today‘s 30–100% to, in an extremecase, 2–100%.Existing control of the radiator circuitsAlthough the radiator circuits within the area weredesigned by the same consultant, there is today a largespread in the choice of control curve and resultanttemperature drop (10–30 °C). It is likely that the curveshave been gradually adapted to the circuits‘ hydraulicproperties and balancing, and one can assume that thisis a common situation.When older houses are renovated and their radiatorcircuits are modernised, there are no guarantees thatoversizing is taken into consideration. For example, theradiator HEX in a substation that was installed in 2005in one of the houses was dimensioned for 185 kW heatoutput at DOT with temperatures corresponding to80/60 °C at a flow of 2.25 l/s. However, whenexamining data for this substation, it turned out that thesubstation delivered less than 40 kW at an outdoorThe <strong>12th</strong> <strong>International</strong> <strong>Symposium</strong> on District Heating and Cooling,September 5 th to September 7 th , 2010, Tallinn, Estonia209temperature around 0 °C, which corresponded to aload of approximately 50%. The actual flow rate wasabout 1.1 l/s and the temperatures corresponded to60/40 °C, thus representing an oversizing around100%.ADAPTIVE OPTIMISATION - METHODIn the theoretical example, the system was assumed tobe 100 % oversized, while in an arbitrary system onecannot be sure of the degree of oversizing. It is alsodesirable to have a robust and adaptive controlalgorithm. The method found to function the best isdescribed below. This approach consists in graduallymodifying, by automatically performed tests, the controlcurve and determining the associated flow rate.Online testingBy locking the control valve (CV), one can assume tohave approximately the same primary flow through theradiator HEX, and since the variations in the cooling ofprimary water is relatively small, the heat supply is alsoapproximately constant. If the secondary flow isreduced while the CV is maintained locked, thetemperature of the secondary flow leaving the HEX willrise. When a new flow and its associated supplytemperature are tested, the current level of the primaryreturn temperature is compared to the level before theexperiment. In this way, the new combination of flowand supply temperature can be either accepted orrejected. This method renders it possible to implementthe adaptive algorithm in any arbitrary system, leadingto the control curve becoming gradually modified. Thismethod we suggested in [7].One problem associated with this kind of optimisation isthat the method is sensitive to disturbances. If theprimary supply temperature, primary differentialpressure or the outdoor temperature changes duringthe test, one cannot be sure that the heat supply isconstant. In that case, a reduced return temperaturecould be the result of a heat supply that is too low.Such tests have to be rejected.In order to render the tests less sensitive todisturbances, the CV is locked only briefly, in order forthe HEX to stabilise. Subsequently, we return toautomatic control, but instead of using the controlcurve, the control aims at maintaining a constanttemperature drop in the radiator system. If this issuccessful, the heat supply is also kept constant. Onecan assume that the secondary flow is relativelyconstant: as long as tests are conducted at night, nosolar radiation is present and internally generated heatis likely to be at a relatively steady level. If, for instance,the primary supply temperature or differential pressurerises during the course of a test, the CV will closesomewhat causing the secondary supply temperature