【作者】 刘晓娟；

【导师】 方建安；

【作者基本信息】 东华大学 ， 控制理论与控制工程， 2014， 博士

【摘要】 电力系统负荷预测是能量管理系统和配电管理系统的重要部分,是实现自动发电控制和经济调度控制的前提,它涉及电力系统规划与设计,电力系统运行的经济性、安全性、可靠性和稳定性等。随着电力系统的不断发展和电网管理的日趋现代化、复杂化,电力系统负荷预测问题的研究引起人们广泛的关注。准确的电力负荷预测可以合理地安排电网内部发电机组的运行,保持电网运行的安全稳定性；可以合理的安排机组检修计划,保证社会的正常生产和人民的正常生活；可以决定新装机容量的大小与电网的调度控制,有效的降低发电成本,提高经济效益和社会效益。因此,电力系统负荷预测成为现代电力系统运行和管理中的一个重要研究课题和现代电力系统科学中的一个重要领域。本文主要利用模糊数学、模糊时间序列、灰色系统理论、模糊多目标最优规划等智能方法,对电力系统负荷预测的模型与算法进行了应用研究。首先给出了基于模糊时间序列的基本预测模型与算法,并考虑到负荷自身变化的特点以及温度等综合因素对电力负荷的影响,对该预测模型进行了优化修正；其次基于稳健统计理论和模糊回归的思想,研究了稳健回归预测算法和稳健模糊回归预测算法；第三在离散灰色预测模型的基础上结合模糊多目标规划理论,研究了模糊离散灰色区间预测模型与算法；最后利用遗传算法和Pareto最优化理论来搜索和确定参数的最优区间,给出了基于多目标优化的模糊时间预测算法。针对不同的模型和算法,通过具体算例对不同预测模型与算法的预测结果进行了分析比较。论文的主要研究内容和创新点如下：(1)基于模糊时间序列的预测模型与算法研究。在模糊时间序列基本理论框架下,将模糊时间序列预测模型应用到电力系统负荷预测中来。首先对历史负荷数据进行必要的预处理,以检测和平滑异常数据,其次考虑到电力负荷自身的特征以及影响负荷预测的重要因素,分别引入温度影响因子、负荷变化影响因子及综合影响因素作为权重因子,提出了时变-模糊时间序列预测算法和基于最高温度修正的最大负荷预测算法。最后通过两个数值算例,说明了我们提出的考虑因素分析的模糊时间序列预测算法具有较高的预测精度和较大的有效性和普遍适应性。(2)基于稳健模糊回归的预测模型与算法研究。利用稳健统计和模糊回归的基本理论,研究了基于稳健估计的稳健回归预测算法和基于Tanaka’s体系下的稳健模糊回归预测模型。该稳健回归预测算法通过函数的选择以及反复迭代减弱异常数值对回归方程的不利影响,增强回归方程的稳健性；而稳健模糊回归预测模型,借助于误差评价指标以及Tanaka’s的模糊回归方法,研究并提出了一稳健模糊回归预测模型来预测模糊区间,通过一系列参数的优化和转化得到非线性模型。最后的两个算例,通过与传统算法和模型的比较,可以明显看出,我们提出的稳健算法和稳健模糊模型都能够有效降低异常数据对预测结果的影响,并且具有较好的稳定性。(3)基于模糊离散灰色理论的区间预测模型研究。在离散灰色预测模型的基础上,结合模糊多目标最优化理论,研究并提出了基于离散灰色模型的模糊多目标区间预测模型。针对输入变量是精确数和模糊数,分别给出了不同的区间预测算法。并将本文的预测结果和现有文献中的模糊线性回归区间预测、灰色模糊区间预测结果进行比较,精确度更高；最后将该模型和算法应用于电力系统负荷区间预测,得到了更一般的预测结果,说明我们提出的模型和算法具有较广泛的适应性。(4)基于多目标优化的模糊时间序列预测模型与算法研究。针对预测区间的划分是影响模糊时间序列预测精度的主要因素之一,首先定义和划分带有参数的预测区间,利用遗传算法搜索最优预测区间的长度并进行最优化参数的筛选和初始参数的选择,并借助于Pareto最优化理论识别最优解,提出了时变-基于比率的多目标最优化模糊时间序列预测模型和时不变-多目标最优化模糊时间序列预测模型,并将该模型和算法分别应用于Alabama大学入学率的预测,同已有文献相比,得到了较好的预测结果。最后,将我们提出的预测模型和算法应用于上海市用电量的长期预测,得到了更一般的预测结果,说明我们提出的预测模型和算法的有效性和适应性。更多还原

【Abstract】 Electric load forecasting is one of the important parts in EMS and SMS and the precondition of the auto-generate electricity control and the economically attemper. It connected the programming and the design of the power system with the economical, security, reliability and stability. With the development of the electric power system and the management of electric grid becoming more and more modern and complexity, more and more researchers pay widely attention to the load forecasting problem. The accurate load forecasting could design the generator groups operational normally, keep up with security and stability of the electric power grid. Preparing the examination and reparation of the generator groups to be ensure the normal production and living; and effectively reducing the cost of generating electricity and enhancing the benefits of the societal and economic. So, the electric load forecasting is an important research topic and research field in modern power system operating and the managementIn this thesis, the electricity load forecasting models and algorithms are discussed mainly using fuzzy mathematics, fuzzy time series theory, grey systems theory and fuzzy multi-objective programming etc the mathematics theory and intelligent method. Firstly, the basic fuzzy time series forecasting models and algorithms are given, Considering the changes of loads itself and temperature et al synthetical factors which affects on the loads, the optimal revised forecasting algorithm are studied again; secondly, a robust regression forecasting algorithm and robust fuzzy regression forecasting algorithm are discussed based on robust statistic theory and fuzzy regression theory, then, discrete grey interval forecasting models and algorithms are obtained based on the discrete grey forecasting model combined with fuzzy multi-objective theory, and at last, the fuzzy time series forecasting algorithm based on genetic algorithm to searching for the optimal parameters and Pareto optimization theory to determined the optimal interval are investigated. Considering the different forecasting models and algorithms, the numerical examples are given to comparing the forecasting results. The main research contents and innovative points are the follows:(1) Research about electricity load forecasting algorithm based on fuzzy time series.Based on the fuzzy time series theory, this paper introduce the fuzzy time series forecasting model and algorithm into the electric power system load forecasting. Firstly, the history loads are pre-disposed to recognize and smooth the outlier. then considering the characters of the loads and the important factors which affect on the load forecasting, the temperature influencing factor, the load changeable factor and the synthetical factors etc are introduced as the weight factors. So the time-variant fuzzy time series forecasting algorithm and maximum load forecasting algorithm based on bi-factor revised fuzzy time series model are proposed. At last, two numerical examples are given to validate our proposed forecasting model and algorithm have better forecasting effects and have an universality.(2) Forecasting model and algorithm research based on robust fuzzy regression.Based on robust statistic theory and Tanaka’s fuzzy regression principle, a robust regression forecasting algorithm and a robust fuzzy regression forecasting model are studied. By considering the choice of the objective function and the weight function and the iterative procedure, The robust regression forecasting algorithm weakens the bad influence of the outlier to the regression equation and strengthens the robustness of the regression equation. On the other hands, the robust fuzzy regression forecasting model is studied, which gives the robust fuzzy forecasting interval by means of the error evaluative index and the Tanaka’s fuzzy regression method. The nonlinear forecasting model is obtained by the optimization and the transform of the parameter. At last, the two numerical examples are given to compare the forecasting accuracy our proposed robust forecasting model and algorithm with the traditional forecasting method, and obviously, our proposed forecasting method could reduce the bad influence which the outlier made to the forecasting effects and could have a better robustness compared with the conventional methods.(3) Research about fuzzy discrete gray interval forecasting model.Based on discrete grey forecasting model combined with fuzzy multi-objective optimization theory, the fuzzy multi-objective discrete grey forecasting model is proposed. Considering the input variables are crisp data or the fuzzy data, the different interval forecasting model and algorithm are obtained.The forecasting results using our proposed forecasting model have a better accuracy compared with the forecasting results of the fuzzy linear regression interval forecasting and the grey fuzzy interval forecasting. Finally, we applied the forecasting model into the power system load interval forecasting and obtain the general results which illuminate that our proposed forecasting model have a universal and extensive applications. (4) Research about the fuzzy time series forecasting model and algorithms based on Multi-objective optimization theory.Because of the forecasting interval partitions is one of the important influence factors which affect the precision of the fuzzy time series forecasting, we first define and partition the forecasting interval with parameters, then using the genetic algorithm to searching for the optimal lengths of forecasting interval with parameters and Pareto optimization theory to identify the optimal solution and choice the initialization parameter. a time-variant ratio-multi objective optimal fuzzy time series forecasting model and a time-invariant multi-objective optimal fuzzy time series forecasting model and algorithms are proposed. When we apply our proposed forecasting model and algorithm to the enrollment forecasting of the Alabama, these forecasting results have a better forecasting precision compared with the other reference literatures. At last, we using our proposed two forecasting models and algorithms into the long term electricity consumption in shanghai and obtain the universal forecasting results which explain that our proposed forecasting methods have availability and applicability.更多还原