用遗传算法重构副热带高压特征指数的非线性动力模型

用遗传算法从离散时间序列资料中反演重构了非线性动力模型。首先, 用Lorenz系统的时间积分数据进行模型重构试验，随后，对十年平均的副热带高压形态指数时间序列进行动力模型参数反演和仿真预报试验。结果表明，遗传算法具有的全局搜索和并行计算优势能够较为准确地描述和刻画副热带高压活动，能对副高活动进行较为准确的描述与建模，是诊断和预测副热带高压等复杂天气系统活动的一条有效途径。

A Non-linear Dynamic System Reconstruction of the Subtropical High Characteristic Index Based on Genetic Algorithm

By using Genetic Algorithm (GA), a non-linear dynamic model is retrieved and reconstructed from T106 numerical forecast fields time series data. From this non-linear dynamic model, the dynamic significance is found, and the model is used as the forecast model to diagnose and predict more accurately.After introducing the basal idea of the dynamic system reconstruction and elementary flow of GA, the Lorenz chaos dynamic model is first retrieved for examining the reconstruction effect of GA. Using the known Lorenz chaos dynamic system to prove that the calculational precision of GA is higher and the retrieved model parameters are closer to reality, the authors objectively and accurately reconstruct the Lorenz chaos dynamic system model.Then the authors retrieve and reconstruct an actual dynamic model of the Pacific subtropical high characteristic index from a time series of the 10-year averaged Pacific subtropical high characteristic index. Using the 10-year average (1958-1997) subtropical high ridge index of pentad, the subtropical high area index of pentad and the subtropical high western extension index of pentad as the time series of observational information, the authors reconstruct the non-linear dynamic model of the Pacific subtropical high characteristic index and get the retrieved parameters, then eliminate the illusive parameters which are take up less proportion. Finally the authors get the reconstructed non-linear dynamic model of the Pacific subtropical high characteristic index. On this basis an integral forecasting examination is processed. By initializing actual data which are got from the series of subtropical high characteristic index, the numerically integral results of model are gotten. Comparing with actual data, the model has the characters of longer forecasting cycle and better forecasting effect. In the paper, the authors also give the comparison between the integral results and the actual results. The simulated forecast results of the retrieved dynamic model show that for GA has the connotative advantages of global optimization and parallel calculation, the dynamic model reconstruction method by GA can accurately describe and simulate the subtropical high, and can be used for diagnosing and predicting such complicated weather system as the subtropical high. The efficiency and veracity of the retrieved model parameter are better than the conventional methods, the forecasting temporal effect and maneuverability are better than the conventional statistical methods, such as Neural Net. A new method is found to diagnose and predict the dynamics of complicated weather system especially in which the dynamic model cannot be got.