在地学、农学等领域，日总辐射量是一个很重要的参数，但相比常规气象观测站，全球辐射观测站的数量却很少，因此，太阳辐射的估算一直受到重视。本文在分析比较前人建立的太阳总辐射模型的基础上，选取较常见且有代表性的5个模型：Angstrom-Prescott模型（I）、Ogelman模型（II）、Bahel模型（III）、日照百分率和气温日较差综合模型（IV）和刘可群等模型（V）。用华北地区6个代表站点2001-2010年的逐日太阳总辐射观测数据，按照整个分析期（10a）、不同季节、不同天气3种情况，对5个模型模拟的效果进行对比与分析。结果表明：（1）5个模型在各代表站点模拟的逐日太阳总辐射值与该站实测值间具有极显著的正相关关系（P≤0.01），相关系数（R）均在0.93以上，平均绝对误差（MAPE）均在9.68%～17.56%，归一化均方根误差（NRMSE）在12.47%～23.12%，模拟结果大多为“好”，个别为“可以接受”，说明5个模型及相应的系数适合估算华北地区日太阳总辐射。（2）分析期内，5个模型在6个站点的MAPE平均值分别为14.28%、14.93%、12.78%、12.27%、13.01%，相应的NRMSE平均值分别为18.80%、19.71%、17.09%、16.27%、17.24%，模拟结果为“好”，其中模型IV误差最小，但与模型III和V无显著差异。（3）模型I-V的MAPE平均值在春季依次为11.97%、12.19%、11.17%、10.86%、11.24%，相应的NRMSE平均值依次为15.46%、15.75%、14.27%、13.95%、14.27%；夏季MAPE平均值依次为14.46%、15.47%、13.32%、12.45%、13.36%，相应的NRMSE平均值依次为18.89%、20.21%、17.21%、16.22%、17.05%；秋季MAPE平均值依次为14.81%、15.65%、12.67%、12.19%、12.20%，相应的NRMSE平均值依次为18.94%、20.00%、16.66%、15.94%、15.95%；冬季MAPE平均值依次为18.08%、18.56%、15.19%、14.99%、14.11%，相应的NRMSE平均值依次为22.52%、23.28%、19.42%、19.06%、18.31%。总体上，四个季节的模拟结果大多为“好”，个别为“可以接受”；春、夏、秋季，模型IV误差最小；在冬季模型V最小，模型IV次之。（4）模型I-V的MAPE平均值，在有日照时，依次为11.23%、12.03%、9.52%、9.32%、9.94%，相应的 NRMSE 平均值依次为14.92%、15.92%、12.75%、12.44%、13.13%。模拟结果均为“好”，其中模型 IV误差最小；无日照时，MAPE 平均值依次为49.25%、47.92%、49.71%、46.03%、45.51%，相应的 NRMSE平均值依次为61.92%、62.09%、61.89%、58.02%、55.70%，模拟结果均为“不好”。综合分析可知，5个模型均可用于估算华北地区日太阳总辐射，但模型IV即日照百分率和气温日较差综合模型精度最高。%Daily solar radiation is a very important parameter in earth science, agriculture science, and other fields. However, compared with the conventional meteorological observatory, there were few observatories of the global radiation. Therefore, the estimation of solar radiation had become a focus. Five representative models, such as Angstrom-Prescott model, Ogelman model, Bahel model, the comprehensive model of sunshine duration and diurnal temperature range, and Liu’s model (named model I to V), were used to estimate solar radiation in this paper, under the analysis and comparison previous solar radiation estimation models. Daily observation solar radiation data (from 2001 to 2010) of six representative stations in Northern China Plain were used to compare the effect of five models, according to the whole analysis period (about 10 years), different seasons, and weather conditions. The results showed that: (1) in each representative station, simulated values of five models had an extremely significant positive correlation between the measured values (P≤0.01), the correlation coefficient (R) were above 0.93, the mean absolute percentage error (MAPE) were within 9.68%-17.56%, and the normalized root mean square error (NRMSE) were within 12.47%-23.12%. The simulation results mostly were "good", individual as an "acceptable". Five models and the corresponding coefficient showed high accuracy in estimating the solar radiation in Northern China. (2) During the whole analysis period, the average value of MAPE (AMAPE) of five models were 14.28%, 14.93%, 12.78%, 12.27%, and 13.01%, respectively, corresponding to the average value of NRMSE(ANRMSE) were 18.80%, 19.71%, 17.09%, 16.27%, and 17.24%, respectively, indicating all of the simulation results were "good". Model IV was the best, following by model III and V. (3) In spring, AMAPE of five models were 11.97%, 12.19%, 11.17%, 10.86%, and 11.24%, respectively, corresponding to ANRMSE were 15.46%, 15.75%, 14.27%, 13.95%, and 14.27%, respectively. In summer, AMAPE of five models were 14.46%, 15.47%, 13.32%, 12.45%, and 13.36%, corresponding to ANRMSE were 18.89%, 20.21%, 17.21%, 16.22%, and 17.05%, respectively. In autumn, AMAPE of five models were 14.81%, 15.65%, 12.67%, 12.19%, and 12.20%, corresponding to ANRMSE were 18.94%, 20.00%, 16.66%, 15.94%, and 15.95%, respectively. In winter, AMAPE of five models were 18.08%, 18.56%, 15.19%, 14.99%, and 14.11%, corresponding to ANRMSE were 22.52%, 23.28%, 19.42%, 19.06%, and 18.31%, respectively. In general, most simulation results were "good", individual was "acceptable", in the four seasons. Model IV showed best in spring, summer, and autumn, while model V showed best in winter, following by model IV. (4) Under the condition that sunshine duration existed, AMAPE of five models were 11.23%, 12.03%, 9.52%, 9.32%, and 9.94%, respectively, corresponding to ANRMSE were 14.92%, 15.92%, 12.75%, 12.44%, and 13.13%, respectively. All of the simulation results were "good", model IV showed best. In contrast, without sunshine duration, AMAPE of five models were 49.25%, 47.92%, 49.71%, 46.03%, and 45.51%, respectively, corresponding to ANRMSE were 61.92%, 62.09%, 61.89%, 58.02%, and 55.70%, respectively. All of the simulation results were "bad". In summary, five models could be used to estimate the daily solar radiation in Northern China Plain, and model IV (the comprehensive model of sunshine duration and diurnal temperature range) showed the highest accuracy.
展开▼
