riso.belief_nets.TemporalBeliefNetwork electrical-demand
{
	riso.belief_nets.BeliefNetwork electrical-demand
	{
		accept-remote-child-evidence false % don't bother asking downstream slices for lambda messages

		% -------------------------- PRELIMINARIES -------------------------------
		riso.belief_nets.Variable t { distribution riso.distributions.Uniform { a 0 b 768 } }
		riso.belief_nets.Variable outdoor-temp
		{
			parents { t }
			distribution riso.distributions.RegressionDensity
			{
				regression-model riso.regression.HarmonicModel
				{
					ncomponents 2 offset 50
					components
					{
						{ amplitude 25 period 8760 phase-shift -2.7418 } % t=0 corresponds to Jan 1
						{ amplitude 7 period 24 phase-shift -2.455 }
					}
				}
				noise-model riso.distributions.Gaussian { mean 0 std-deviation 10 }
			}
		}

		riso.belief_nets.Variable total-insolation
		{
			parents { t }
			distribution rates.TotalInsolation { }
		}

		riso.belief_nets.Variable sol-air 
		{
			parents { outdoor-temp total-insolation }
			distribution rates.SolAirCalc { alpha 0.8 ho 5 epsilon-delta-R-by-ho 7 }
		}

		% ------------------------ ENVELOPE HEAT TRANSFER -----------------------------
		% corresponds to first term in Eq. 7-21 in McQuiston+Parker

		riso.belief_nets.Variable inward-heat-transfer-per-area
		{
			parent-prior prev[sol-air] riso.distributions.Gaussian { mean 50 std-deviation 20 }
			parent-prior prev[prev[sol-air]] riso.distributions.Gaussian { mean 50 std-deviation 20 }
			parent-prior prev[prev[prev[sol-air]]] riso.distributions.Gaussian { mean 50 std-deviation 20 }
			parent-prior prev[prev[prev[prev[sol-air]]]] riso.distributions.Gaussian { mean 50 std-deviation 20 }

			parents
			{
				sol-air prev[sol-air] prev[prev[sol-air]] prev[prev[prev[sol-air]]] prev[prev[prev[prev[sol-air]]]]
			}
			% constants b_n from Table 7-24 in McQuiston+Parker.
			% "4in or 100mm heavyweight concrete".
			distribution riso.distributions.LinearCombination { 0.0078 0.0705 0.0355 0.0011 0.0 }
		}
		
		% corresponds to second term in Eq. 7-21 in McQuiston+Parker
		% constants from ``Transfer function coefficients for exterior walls,'' Table 7-24 in McQuiston+Parker.
		riso.belief_nets.Variable heat-transfer-storage
		{
			parent-prior prev[net-inward-heat-transfer] riso.distributions.Gaussian { mean 50 std-deviation 20 }
			parent-prior prev[prev[net-inward-heat-transfer]] riso.distributions.Gaussian { mean 50 std-deviation 20 }
			parent-prior prev[prev[prev[net-inward-heat-transfer]]] riso.distributions.Gaussian { mean 50 std-deviation 20 }
			parent-prior prev[prev[prev[prev[net-inward-heat-transfer]]]] riso.distributions.Gaussian { mean 50 std-deviation 20 }

			parents
			{
				prev[net-inward-heat-transfer] prev[prev[net-inward-heat-transfer]] prev[prev[prev[net-inward-heat-transfer]]] prev[prev[prev[prev[net-inward-heat-transfer]]]]
			}
			% constants d_n from Table 7-24 in McQuiston+Parker.
			% "4in or 100mm heavyweight concrete".
			% roots([1 -d.^2]) yields 0.7797, ... OK. VARIANCES DO NOT BLOW UP !!!
			distribution riso.distributions.LinearCombination { -0.8789 0.0753 -0.0001 0.0 }
		}

		% corresponds to third term in Eq. 7-21 in McQuiston+Parker
		riso.belief_nets.Variable outward-heat-transfer-per-area
		{
			parents { heat-transfer-consts.indoor-temp }
			% constant \sum_n c_n from Table 7-24 in McQuiston+Parker.
			% "4in or 100mm heavyweight concrete".
			distribution riso.distributions.LinearCombination { 0.1149 }
		}

		riso.belief_nets.Variable inward-heat-transfer
		{
			parents { heat-transfer-consts.area inward-heat-transfer-per-area }
			distribution riso.distributions.Product { }
		}

		riso.belief_nets.Variable outward-heat-transfer
		{
			parents { heat-transfer-consts.area outward-heat-transfer-per-area }
			distribution riso.distributions.Product { }
		}

		% rhs of Eq. 7-21 in McQuiston+Parker
		riso.belief_nets.Variable net-inward-heat-transfer
		{
			parents { inward-heat-transfer heat-transfer-storage outward-heat-transfer }
			distribution riso.distributions.LinearCombination { 1 -1 -1 }
		}

		% ---------------------- NON-ENVELOPE HEAT TRANSFER ---------------------------
		riso.belief_nets.Variable occupancy-heat-transfer
		{
			parents { t }
			distribution riso.distributions.RegressionDensity
			{
				regression-model rates.ScheduleModel
				{
					0, 7 | 2000 

					7, 20 | 8500    % figured as 340 Btu/h per person, times 25 people
					20, 31 | 2000 

					31, 44 | 8500
					44, 55 | 2000 

					55, 68 | 8500
					68, 79 | 2000 

					79, 92 | 8500
					92, 103 | 2000 

					103, 116 | 8500
					116, 127 | 2000 

					127, 140 | 2500   % smaller peak on Saturday
					140, 151 | 2000 

					151, 164 | 1500   % smaller peak on Sunday
					164, 168 | 2000 
				}
				% noise should be smaller when occupancy loads are smaller -- oh, well !!!
				noise-model riso.distributions.Gaussian { mean 0 std-deviation 2000 }
			}
		}

		riso.belief_nets.Variable ltg+equip-heat-transfer
		{
			parents { t }
			distribution riso.distributions.RegressionDensity
			{
				regression-model rates.ScheduleModel
				{
					0, 7 | 30000
					7, 20 | 50000   % figured as twice as much as the occupancy load
					20, 24 | 30000
				}
				noise-model riso.distributions.Gaussian { mean 0 std-deviation 10000 }
			}
		}

		riso.belief_nets.Variable vent-heat-transfer
		{
			parents { outdoor-temp heat-transfer-consts.indoor-temp }
			% We need to calculate vent-heat-transfer = (factor)(outdoor-temp - indoor-temp).
			% Factor in combination is (volume/min)(60 min/h)(Btu/lbm)(lbm/ft^3)
			% Note that for air, (60 min/h)(Btu/lbm)(lbm/ft^3) == (60)(0.062)(0.24) == 0.89.
			% Assume airflow == 1000 cfm.
			distribution riso.distributions.LinearCombination { 890 -890 } 
		}

		% -------------------------- TOTAL HEAT TRANSFER ------------------------------
		riso.belief_nets.Variable total-heat-transfer
		{
			parents { occupancy-heat-transfer ltg+equip-heat-transfer vent-heat-transfer net-inward-heat-transfer }
			distribution riso.distributions.LinearCombination { 1 1 1 1 }
		}

		% --------------------- ELECTRICAL DEMAND FOR COOLING -------------------------
		riso.belief_nets.Variable electrical-demand
		{
			parents { total-heat-transfer }
			% Convert Btu/h to kW: 1 kW == 3413 Btu/h. ASSUME CONSTANT COP=3 !!! 
			distribution riso.distributions.LinearCombination { .00009766578767457759 }
		}
	}
}