摘要 |
PURPOSE:To accurately predict the air conditioning heat load by sequentially correcting a predictive model by using the learning function of a neural network based on the comparison error of the air conditioning heat load predicted value calculated in the previous day based on past observation data and tentative predictive model with the heat load performance value of an appointed day. CONSTITUTION:At the time of starting service, a tentative model forming module is executed, and an air conditioning heat load simulation is conducted for an object building. Then, the output value of a heat load predicting neural network (N.N.) is compared with an air conditioning heat load simulation signal, the N.N. is tuned in response to the error, and an N.N. tentative model for predicting a heat load is formed. An air conditioning load predicted value of next day is calculated from appointed day environmental factor data from a sensor 2 and next day weather factor data from a weather forecast data collector 4, compared with next day air conditioning heat load performance value, the comparison error is corrected, and the corrected value makes use of air conditioning heat load prediction. |