SIMULATION

This study compares the performance of a single gas- fired conventional domestic space heater with that of the same conventional heater in combination with a smaller auxiliary heater. The auxiliary heater could be a solar heater or could obtain its energy from more conventional sources. The auxiliary heater is presumed to have a shorter time constant than the main furnace. As room temperature falls below the present temperature, the auxiliary heater turns on before the main furnace, but turns off when the main furnace comes on. In the study the following param eters were varied:

Hysteresis of the main furnace's thermostat Hysteresis of the auxiliary heater's thermostat

Time constants of the controller, main furnace, and auxiliary heater

The size (heat output capacity) of the auxiliary heater as a percentage of the size of the main furnace.

The outputs recorded were temperatures variations of the room air and the cycling time of the system. In each case results are plotted as a function of the relative size of the auxiliary heater for 3 values of one of the time constants, the other two time con stants and the hystereses of the thermostats being held constant. Operating conditions were a set point temperature 70°F (21°C) for room air and an outdoors air temperature of 0°F (-18°C). In all cases examined the system with the auxiliary heater produced more uniform room temperatures. Shortening the time constants of the controller and the auxiliary heater reduced variations in room temperature.

This paper describes ACSL, a continuous system simu lation language designed to help the engineer or scientist with a mathematical model analyze the behavior of his system. The computer calculates dynamic responses and interactively provides pictures (plots) and tabular display of selected variables. The user has a choice of two Runge-Kutta integration algorithms and two variable stepsize algorithms which includes Gear's implicit method for stiff sys tems. Any other programmable algorithm can be readily incorporated. The main features of the lan guage are illustrated by a detailed example.

Two systems are simulated using the same random- number stream. To compare their means one can use standard statistical techniques, e.g., the t-test for paired comparisons. To compare their variances, however, the standard F-test fails. Three approaches are suggested. Approach 1 uses the traditional F- statistic and is conservative. Approach 2 splits the observations into two groups. For each group the F- statistic is exact. It is shown how to combine the two F-statistics in a very simple way. This approach is also conservative. Approach 3 was developed in 1946 by Wilks. It is slightly more complicated but exact.