Heating and cooling degree day prediction within the London urban heat island area

摘要

This paper describes the London Site Specific Air Temperature prediction model, which comprises of a suite of artificial neural network (ANN) models to predict site-specific hourly air temperature within the Greater London Area (GLA). The model was developed using a back-propagation ANN model based on hourly air temperature measurements at 77 fixed temperature stations (FTS) and hourly meteorological data (off-site variables) from Heathrow; it also includes six on-site variables calculated for each FTS. The temporal and spatial validity of the model was tested using data measured 7 years later from the original dataset, which include new FTS locations. It was found that site-specific hourly air temperature prediction is within accepted range and improves considerably for average daily and monthly values. Therefore, the model can be used with confidence to predict daily and seasonal variations of air temperature within the GLA and in particular for the calculation of monthly and annual heating degree days (HDD) and cooling degree hours (CDH). It was found that as expected HDD increase and CDH decrease with distance from the urban heat island centre point; however, all variations cannot be explained with distance and six key on-site variables namely aspect ratio, surface albedo, plan density ratio, green density ratio, fabric density ratio and thermal mass have been identified to explain the remaining variation.