模糊神经网络广义预测控制在单元机组协调控制中应用研究

发布时间：2018-05-21 08:08

本文选题：广义预测控制 + 模糊神经网络　；参考：《内蒙古工业大学》2015年硕士论文

【摘要】：燃煤发电在我国能源结构中占主要地位,尽管有非常规能源如页岩气、可燃冰等出现,但目前使用的最廉价、最安全的能源仍然是煤炭。火电厂单元机组是电网的基本组成部分,保障电网基本的负荷要求;它担负着电网的调峰、调频任务,影响电网的运行稳定和经济效益。在日益严峻的环境压力下,如何确保火电厂安全、经济和高效稳定运行成为当务之急。机炉协调控制是一个局部线性、全局强非线性,具有双输入双输出的强耦合控制系统。在全工况范围内,采用常规的控制方法难以保证其得到满意的控制品质。本文是在大量查阅了有关广义预测控制、模糊控制和神经网络控制文献的基础上,针对广义预测控制具有在线计算量大,不适用于非线性系统的特点,提出了采用模糊神经模型的模糊神经网络广义预测控制。通过采用模糊神经网络的辨识方法,克服了传统的模糊系统辨识精度较低的问题,可以有效的解决BP神经网络学习算法存在收敛速度慢和局部极小值问题;同时该算法有效的减少了广义预测控制算法在线计算量,使广义预测控制算法的实时性得到很大的提高。最后采用300MW单元机组协调控制系统数学模型进行了仿真研究,仿真结果表明,控制系统保证对象输出功率和主蒸汽压力值能快速平稳地跟踪设定值。在变负荷工况下,系统仍然能够保持良好的控制性能,与传统PID控制相比该算法表现出良好的自适应性和鲁棒性。
[Abstract]:Coal-fired power generation plays an important role in China's energy structure. Although unconventional energy sources such as shale gas and combustible ice appear, coal is still the cheapest and safest energy source. The unit of thermal power plant is the basic component of the power network, which guarantees the basic load requirements of the power network, and it undertakes the tasks of peak shaving and frequency modulation of the power network, which affects the stability and economic benefit of the power network. Under the increasingly severe environmental pressure, how to ensure the safety, economy and efficient and stable operation of thermal power plants has become an urgent task. The coordinated control system is a strong coupling control system with local linearity, global strong nonlinearity and double input and double output. In the whole working condition, it is difficult to obtain satisfactory control quality by using conventional control methods. In this paper, on the basis of a large number of literatures on generalized predictive control, fuzzy control and neural network control, the generalized predictive control has the characteristics of large on-line computation and not suitable for nonlinear systems. A fuzzy neural network generalized predictive control based on fuzzy neural model is proposed. By adopting the identification method of fuzzy neural network, the problem of low precision of traditional fuzzy system identification is overcome, and the problem of slow convergence speed and local minimum value of BP neural network learning algorithm can be effectively solved. At the same time, the algorithm effectively reduces the amount of on-line computation of the generalized predictive control algorithm, so that the real-time performance of the generalized predictive control algorithm has been greatly improved. Finally, the mathematical model of coordinated control system of 300MW unit is used to simulate the system. The simulation results show that the control system ensures the output power of the object and the value of the main steam pressure can track the set value quickly and smoothly. Under the variable load condition, the system can still maintain good control performance. Compared with the traditional PID control, the algorithm has good adaptability and robustness.
【学位授予单位】：内蒙古工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TM621;TP273

【参考文献】