【正文】 low rate meter 0–2 kg/min ^%RTD 100 4 wires 2100 to 500 8C ^ 8CPiezoelectric absolute pressure gauge 1–10 bar。 ^% .C. Aprea et al. / International Journal of Refrigeration 27 (2020) 639–648642major in parison with the time necessary when thepressor works at a 50 Hz nominal frequency. So, it mayhappen sometimes that even when the pressor works forfrequencies lower than that nominal ones the energy savingexepected may be partially obtained because the pressorhas worked at lower frequencies indeed but for too muchtime. So, in order to regulate the working time of thepressor at lower frequencies it is important that, whenthe fuzzy algorithm input and output variables membershipfunctions are to be defined, the choice of the subset numberand of its wideness has to be proper and guided by theexperimental knowledge. As for the the choice of the rules itis necessary to do similar consideration. For this reason thealgorithm membership functions and rules suggested fromthe authors have been experimentally verified. Table 2shows the rules fixed to set the algorithm and the five fuzzysubsets used to characterize the input and output linguisticvariables marked with the following labels: very low (VL),low (L), medium size (MS), high (H) and very high (VH).As to tune the membership function is much easier than totune the control rules, the attention is focused here on theformer, in order to realize a robust fuzzy controller for thepressor speed control. After some experimental considerations to understand the control characteristics of thealternative pressor, the membership functions in Figs.3–5 have been defined for the temperature differencebetween the setpoint temperature and the real temperatureof the air in the cold store, the derivative of this temperaturedifference with the time and the frequency of the pressormotor supply current. The triangular membership function,with one center and two limits, has been adopted here. Asfor the temperature difference the range covered is locatedbetween 0 and 13 8C(Fig. 3). In order to increase thesensitivity of the fuzzy controller as the cold store airtemperature approaches the setpoint, the membershipfunction is tuned more narrowly to the VL, L and MStemperature differences. As for the derivative with the timeof the temperature difference (Fig. 4), a range prisedbetween and K/s has been covered. It has beenconsidered as a variable input with the derivative also takinginto account mainly the fast variations when the coolingload varies suddenly。=dt ms l l ms h vhh msmsh hvhFig. 3. Membership function of the temperature difference betweenthe setpoint temperature and the real temperature of the air in thecold store.C. Aprea et al. / International Journal of Refrigeration 27 (2020) 639–648 6435. Test results and discussionSeveral experimental tests have been conducted toexplain the energy saving obtainable with the fuzzyalgorithm in parison with the classical thermostaticcontrol, that determines the on/off cycles of the pressorthat works at a nominal frequency of 50 Hz. To simulatebetter the real working conditions of the cold store, varioustypes of cooling loads have been considered. In particular, inthe experimental tests either the electric heaters or the fruitsand vegetables have been adopted as cooling load. Moreover, a further load results to be both due to the periodicopening of the cold store door and also due to the inevitableheat exchanges with outdoor air when the cold store door isclosed. In Fig. 6 a parison in terms of electric energyconsumption, measured by means of a proper electricenergy meter, between the control on–off realized by theclassical thermostat and the pressor speed continuouscontrol obtained by the fuzzy algorithm, is reported whenthe cooling load is due only to the periodic opening of thecold store door. The experimental tests have been realizedfor cold store air temperatures fixed at 25, 0 and 5 8C andfor a constant cooling load obtained opening the cold storedoor every 20 min for about 5 min with an outdoor airtemperature of about 18 8C. One can clearly observe that theenergy consumption increases when the cold store airtemperature decreases. This is due to the fact that a constantcooling load has been considered for all the cold store airtemperatures and so the time necessary to reach thetemperature of 25 8C will be greater and will determine ahigher electric consumption. Moreover, it is possible toobserve that the energy saving obtained with the algorithmwith respect to the thermostat is on an average of about 10%,even if it clearly diminishes slowly when the cold store airtemperature decreases because under this circumstance theworking time of the pressor increases.In Fig. 7 the electric energy consumption of thepressor obtainable with two control systems, related tothe R507 and R407C, is reported both for the summer andfor the winter, when the cooling load is due both to theperiodic opening of the cold store door and to the presenceof the electric heaters. In these experimental tests related tothe electric heaters it has been considered an electric powerconstant of about 200 W. It has been observed that the bestperformances are related to the R407C that allows, with acontinuous control of the pressor speed, an mediumenergy saving of about 13% in parison with thethermostatic control for both the outdoor air temperaturesconsidered. In particular, the absolute electric energyconsumption in the summer season is about 5% higherthan that of the winter season even if the energy saving inthe two seasons is practically the same. In Fig. 8 the energyconsumption related to a real cooling load, represented by200 kg of fruit and vegetables and by the periodic opening ofthe cold store door, is considered when both the fuzzycontrol and the thermostatic control