基于最小二乘支持向量机的短期负荷预测

发布时间：2018-08-07 14:29

【摘要】：电力系统短期负荷预测是电网安全、经济运行的重要依据之一，正确而精准的电力系统短期负荷预测有助于提高电网运行的安全性、经济性和改善电能质量。因此，寻求最合适的电力短期负荷预测的预测方法，从而对提高短期负荷预测的预测精度是具有十分重要的应用价值。本文基于电力系统短期负荷预测的背景、意义以及国内外发展现状的研究，分析了电力系统负荷的特点、规律以及与各种影响因素之间的非线性关系，给出了对历史负荷数据中异常数据的辨识与处理方法，对历史负荷数据和与短期负荷预测有关的影响因素进行归一化处理。根据最小二乘支持向量机(LSSVM)具有能够较好地解决小样本、非线性、高维数以及局部极小值等实际问题的优势。首先基于支持向量机(SVM)的研究，通过利用训练误差的平方代替松弛变量，将不等式约束改进为等式约束，从而提出最小二乘支持向量机(LSSVM)的电力系统短期负荷预测模型，这样就避免了求解一个二次规划问题，提高预测模型训练的速度。由于LSSVM短期负荷预测模型的参数选择对预测结果精度有着至关重要的影响，，本文提出利用粒子群算法(PSO)对LSSVM中的参数进行优化选择，得到基于PSO-LSSVM的短期负荷预测模型，以求进一步提高预测精度。但在粒子群优化算法进行寻优过程中，容易陷入局部最小值，出现早熟收敛的情况，针对这一问题，提出对标准粒子群优化算法进行改进，避免其在优化过程中出现上述问题。建立基于改进粒子群优化算法的最小二乘支持向量机(IPSO-LSSVM)短期负荷预测模型。并通过平均相对误差和均方差根来作为评价标准，验证该算法的准确性。最后，本文通过对2010年广东某地区的历史负荷数据进行分析研究，分别对基于LSSVM、PSO-LSSVM、IPSO-LSSVM的三种短期负预测模型进行预测仿真。最终结果对比表明：IPSO-LSSVM短期负荷预测模型具有收敛性好、有较高的预测精度和较快的训练速度，验证利用改进的PSO算法进行参数优化有助于提高短期负荷预测的预测精度，由此可以说明对LSSVM短期负荷预测模型的参数优化具有很高的研究价值和社会意义。
[Abstract]:Short-term load forecasting of power system is one of the important bases for safe and economical operation of power system. Correct and accurate short-term load forecasting of power system is helpful to improve the security, economy and power quality of power system. Therefore, it is very important to find the most suitable short-term load forecasting method to improve the forecasting accuracy of short-term load forecasting. Based on the background and significance of short-term load forecasting in power system and the research on the development of power system at home and abroad, this paper analyzes the characteristics, laws and nonlinear relationship between power system load and various influencing factors. The identification and processing methods of abnormal data in historical load data are given. The historical load data and the influencing factors related to short-term load forecasting are normalized. According to the least squares support vector machine (LSSVM), it has the advantage of solving the practical problems such as small sample, nonlinear, high dimension and local minimum. Firstly, based on the research of support vector machine (SVM), the inequality constraint is improved to equality constraint by using the square of training error instead of relaxation variable, and a short-term load forecasting model of power system based on least squares support vector machine (LSSVM) is proposed. In this way, it avoids solving a quadratic programming problem and improves the training speed of prediction model. Because the parameter selection of LSSVM short-term load forecasting model has an important influence on the precision of forecasting results, this paper proposes to optimize the selection of parameters in LSSVM by using particle swarm optimization algorithm (PSO), and obtain the short-term load forecasting model based on PSO-LSSVM. In order to further improve the accuracy of prediction. However, in the process of particle swarm optimization, it is easy to fall into the local minimum, which leads to premature convergence. To solve this problem, an improved standard particle swarm optimization algorithm is proposed to avoid the above problems in the process of optimization. A short term load forecasting model based on improved particle swarm optimization (PSO) based on least squares support vector machine (IPSO-LSSVM) is established. The accuracy of the algorithm is verified by using the mean relative error and root of mean variance as evaluation criteria. Finally, by analyzing the historical load data of a certain area in Guangdong province in 2010, this paper simulates three short-term negative forecasting models based on LSSVMU PSO-LSSVMU IPSO-LSSVM respectively. The comparison of the final results shows that the short term load forecasting model has good convergence, high forecasting accuracy and fast training speed. It is proved that the improved PSO algorithm is helpful to improve the forecasting accuracy of short-term load forecasting. Therefore, it is of great value and social significance to optimize the parameters of LSSVM short-term load forecasting model.
【学位授予单位】：湖南大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM715

【参考文献】