集中供热系统的热负荷预测方法研究

发布时间：2018-11-13 08:55

【摘要】：城市集中供热节能是建筑节能的有效途径之一。供热系统的运行调节直接影响着系统的整体节能。供热系统的供暖热负荷预测指导着供热系统的运行调节和控制。因此，本文提出了对供热系统供暖热负荷预测方法的研究，目的是提供一种预测供热系统供暖热负荷值的方法，在确保满足各热用户（热力站）的实际供热需求的条件下达到供热系统的经济运行。本文以集中供热系统作为研究对象，基于遗传算法优化的神经网络算法，进行供热系统的供暖热负荷预测方法的研究。首先从影响建筑热负荷的因素入手，对建筑供暖热过程和供暖负荷的特征进行了分析。按照供暖负荷的特征将其划分为气象因素、建筑因素和人为因素，并区别供热计量建筑与非热计量建筑。由于用户行为的不可预知性，其为负荷预测中的偶然因素。在此基础上，根据热负荷预测过程中的主要影响因素对热负荷方法进行了研究。对目前可用于热负荷预测的方法进行了评价。由于供热系统热负荷具有时变性、时滞性、随机性和偶然性等特点，神经网络方法对热负荷的预测具有明显优势，其中，BP神经网络算法存在收敛慢、容错差、易陷于局部最小等缺点，而遗传算法能弥补BP算法的不足，因此本课题选择了这种方法。本文以MATLAB软件为平台，，用遗传算法来优化网络结构的权值，再结合BP神经网络算法对历史数据进行学习训练，最终得到热负荷的预测结果。
[Abstract]:Urban central heating energy saving is one of the effective ways of building energy saving. The operation regulation of heating system directly affects the whole energy saving of the system. Heat load forecasting of heating system guides the operation regulation and control of heating system. Therefore, this paper puts forward the research of heat load forecasting method for heating system, the purpose of which is to provide a method to predict the heating load value of heating system. The economical operation of the heating system can be achieved under the condition of satisfying the actual heating demand of the heating users (stations). This paper takes the central heating system as the research object, based on the genetic algorithm optimization neural network algorithm, carries on the heating heat load forecasting method of the heating system. Firstly, the characteristics of heating process and heating load of buildings are analyzed from the factors that affect the heat load of buildings. According to the characteristics of heating load, it is divided into meteorological factor, building factor and human factor, and the heating metering building and non-heat metering building are distinguished. Because of the unpredictability of user behavior, it is an accidental factor in load forecasting. On this basis, the method of heat load is studied according to the main influencing factors in the process of heat load forecasting. The current methods for heat load forecasting are evaluated. Because of the characteristics of time-varying, time-delay, randomness and contingency of heat load in heating system, the neural network method has obvious advantages in heat load forecasting. Among them, the BP neural network algorithm has slow convergence and poor fault tolerance. It is easy to be trapped in local minimum, but genetic algorithm can make up for the deficiency of BP algorithm, so this method is chosen in this paper. In this paper, the weight of network structure is optimized by genetic algorithm on the platform of MATLAB software, and then BP neural network algorithm is used to study and train the historical data. Finally, the prediction results of heat load are obtained.
【学位授予单位】：长安大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TU995

【参考文献】