热网系统仿真及一次网优化控制研究

发布时间：2018-03-09 21:54

本文选题：热网　切入点：Flowmaster　出处：《内蒙古科技大学》2014年硕士论文　论文类型：学位论文

【摘要】：当代城市的快速发展，使集中供暖系统成为我国北方城市重点建设的系统。因为供暖系统具有非线性、多变量、大滞后的特点。并且热负荷有时变性与不确定性的特点，所以供暖系统是一个不容易控制的系统。怎样节约能源，降低损耗的同时对热网系统进行有效的控制，，是供热企业面临的一个重要问题。论文首先从供热管网模型着手，由于以往根据热网机理建立的热网模型，为了简化传递函数和减少热网的计算量忽略了一些热网的相关参数，这样建立的模型和真实的热网系统存在一些差距，同时搭建的模型通用性不强，只能针对特定的热网系统。本论文使用一维流体模拟软件Flowmaster搭建可视化的集中供暖系统仿真模型，根据实际的集中供暖系统搭建具有热源、热力站、热用户三大部分的模型，其中热力站的数目是十四个，是一个比较大型的集中供暖系统。搭建的模型能够在多种模拟环境下进行仿真，其需要设置的参数种类多，能够比较贴近真实的系统。由于集中供热系统的热负荷和大气温度的不确定性，需要建立短期预测模型，作为热量分配的约束条件。本文选用基于神经网络的预测模型对各个热力站的热负荷滚动预测，该方法使用动态K均值聚类算法与递归最小二乘法（ROLS）改良RBF神经网络，用逐渐刷新历史数据的办法实现了对热力站短期热负荷预测。针对一次网的优化控制问题，设计一个热源生产最少热量的目标函数。采用粒子群优化（ParticleSwarmOptimization，PSO）算法，对目标函数寻优，分别找到各时刻总供热量然后按照各个热力站预测的热负荷变化趋势给模型分配热量。通过PSO的寻优计算，可以使供暖系统提供最少的热量满足最多的用户，达到节约能源的目的。论文通过仿真实验的方法验证了RBF神经网络对各个热力站热负荷预测值作为分配总热量的依据，然后使用PSO优化算法和Flowmaster搭建的集中供热系统仿真模型进行联合仿真。通过对比仿真后的模型数据曲线可以发现有控制算法的模型能够按照热用户的热量需求变化趋势进行供热，对实际系统节能降耗有一定的指导意义。
[Abstract]:With the rapid development of modern cities, central heating system has become the key construction system in northern cities of China. Because the heating system has the characteristics of nonlinear, multi-variable, large lag, and the heat load is sometimes variable and uncertain, Therefore, heating system is not easy to control, how to save energy, reduce losses while effectively control the heating network system, is an important problem facing heating enterprises. Firstly, the paper starts with the heat supply network model. Because of the previous heat network model established according to the heat network mechanism, in order to simplify the transfer function and reduce the calculation amount of the heat network, some related parameters of the heat network are ignored. There are some gaps between the established model and the real heat network system, and the model built at the same time is not universal. This paper uses one-dimensional fluid simulation software Flowmaster to build a visual central heating system simulation model. According to the actual central heating system, the model has three parts: heat source, thermal station and heat user. The number of thermal stations is 14, which is a relatively large central heating system. The built model can be simulated in a variety of simulation environments, and there are many kinds of parameters that need to be set up. Because of the heat load of central heating system and the uncertainty of atmospheric temperature, it is necessary to establish short-term forecasting model. As the constraint condition of heat distribution, this paper selects the neural network-based prediction model to predict the thermal load rolling of each thermal station. The dynamic K-means clustering algorithm and the recursive least square method are used to improve the RBF neural network. The short-term heat load forecasting of thermal station is realized by gradually refreshing the historical data. In view of the optimization control problem of primary network, an objective function for heat source production is designed. Particle swarm optimization (PSO) algorithm is used to optimize the objective function. Find out the total heat supply at each time and assign the heat to the model according to the change trend of heat load predicted by each thermal station. Through the optimization calculation of PSO, the heating system can provide the least heat to satisfy the most users. To save energy. The RBF neural network is used to predict the thermal load of each thermal station as the basis of the total heat distribution. Then the simulation model of central heating system built by PSO optimization algorithm and Flowmaster is combined. By comparing the model data curve after simulation, we can find that the model with control algorithm can change according to the heat demand of heat users. The trend of heating, It has certain guiding significance to the actual system energy saving and consumption reduction.
【学位授予单位】：内蒙古科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TU833;TP13

【参考文献】