Studies on a Load Forecasting Method with the Evolutionary Parallel Algorithm

本计画提出一种在平行禁忌搜寻之简化模糊推论中，有效地决定模糊隶属函数之位置与数目的方法。PTS算法是禁忌搜索算法的一种改进算法，具有多种策略。一种是将TS的邻域分解成若干个子邻域，从而减少计算量。另一种是在TS算法中引入多个禁忌长度，使候选解更加多样化，以获得更好的解。因此，PTS允许在解的精度和计算时间方面改进传统TS。具体而言，PTS在优化电力系统负荷预测的简化模糊推理结构方面起着关键作用。将基于PTS的简化模糊推理成功地应用于真实的日最大负荷提前一步预测数据。通过与传统的模拟退火、遗传算法和TS等方法的比较，验证了PTS的有效性。仿真结果表明：1. PTS能够高效地求出全局极小值的高度近似解.所提出的方法允许最小化最大预测误差的信息准则.使用简化的模糊推理，通过隶属函数给出输入变量的有用信息。

This project has proposed an efficient method for determining the location and number of fuzzy membership functions in simplified fuzzy inference with parallel tabu search(PTS). PTS is an improved algorithm of tabu search (TS) for solving a combinatorial optimization problem and has a couple of strategies. One is to decompose the neighborhood of TS into several subneighborhoods so that computational effort is reduced. The other is to introduce multiple tabu lengths into the TS algorithm and make solution candidates more diverse to obtain a better solution. As a result, PTS allows to improve the conventional TS in terms of solution accuracy and computational time. Specifically, PTS plays a key role to optimize the structure of simplified fuzzy inference for load forecasting in power systems. The PTS-based simplified fuzzy inference was successfully applied to real data of one-step ahead daily maximum load forecasting. A comparison between PTS and the conventional methods such as SA, GA and TS was made to demonstrate the effectiveness of PTS. The simulation results have shown the following:1. PTS is capable of evaluating a highly approximate solution of a global minimum efficiently.2. The proposed method allows to minimizes the maximum prediction error with the information criterion.3. The use of simplified fuzzy inference gives useful information of input variables through the membership functions.