| HVAC engineers must design reasonable, high efficient and optimum HVAC systems to meat the trend of sustainable development. Only on the basis of annual energy analysis, would optimum design be archived. Energy analysis begins with cooling/heating loads calculation, which require climatic data as inputs. However, there isn't sufficient climatic data for energy analysis in China. So research work as to this aspect is emergent.Climatic data may be constructed to form TRY/TMY by statistic methods or synthesized by stochastic models. Stochastic models are chosen because they have the advantages of conciseness and less subjective, and are more suitable for the original data in China. Analysis is carried out on the data base of Beijing, Xi'an; Chongqing and Chengdu. Part data of Kunming and Fuzhou is also employed. Three climatic factors including solar radiation, temperature and humidity are considered, to which building energy analysis is most sensitive. Conditioned by the fact that daily records are sufficient while hourly records are insufficient, two-stage models are supposed. That means that models do not generate hourly variables directly. Daily variables would be simulated at first and then hourly variables would be simulated on basis of daily data.Daily variables including daily total solar radiation, daily mean and range of temperature, daily mean and range of water-vapor pressure are modeled. Daily climatic variables consist in deterministic and stochastic components, both of which have important impacts on energy analysis. Thus combination models :deterministic models + stochastic models are established. Deterministic models describe the periodic variation and are Fourier series with only significant cycles. Significant cycles are identified by Fourier transfer and spectrum analysis. The most significant frequencies of daily climatic variables are 0,1,2,3,4 cycles/day.Stochastic components formed by subtracting deterministic components from the original data are non-normal, cross-related series with stationary means and time-varying variances. The series become weak stationary after standardized with monthly variance. Normal transfers are performed on the standard series, which ensure that simulated datahave the same distributions of the original data. The normal series are described by multivariate ARX model, of which input is solar variable and outputs are temperature and water-vapor pressure variables. The orders are determined by correlation analysis and FPE criteria, and parameters are estimated by least square mean technique. The stochastic solar variables are separately modeled by univariate AR(2). The supposed multivariate ARX models have the advantage of keeping the same cross-correlation among daily variables over the common multivariate AR model. The reason is that the solar variable of the n'th day comes into effect when simulating temperature and water-vapor pressure variables of the n'th day by ARX. The form of Fourier model plus ARX model is employed first time in China, and have the best performance of keeping cross-correlation and distribution. Daily diffuse radiation wasn't included in combination model, because stochastic model can not ensure that daily diffuse radiation is not more than daily total radiation. Polynomial models are built to estimate daily diffuse radiation from daily total radiation. A probability density model has been supposed also.Models for hourly solar radiation are deterministic ones. Hourly total radiation models are Fourier series models with significant frequencies determined by Fourier transfer,which include 1,2,3,4 cycles/day and 4 bandsides of every frequency above and frequencies of 0,1,2,3,4 cycles/year . The Fourier models should be adjusted by daily values generated by combination models. Polynomial models separating hourly diffuse radiation from hourly total are built under different air mass. Polynomial fits under different air mass are more accurate than general model for the whole day.Fourier models with freq...
|
PDF Full Text Request