基于人工神经网络的建筑热负荷预测及控制

发布时间：2018-02-04 09:19

本文关键词： 人工神经网络建筑热负荷预测节能控制　出处：《大连海事大学》2015年硕士论文　论文类型：学位论文

【摘要】：目前,在我国北方城镇的集中供暖大部分是供热中心或者换热站直接把热水送往用户端,易于造成供热不均和能耗浪费。虽然部分地区按国家相关建议和要求进行了按热收费、分户计量的尝试,但由于分户计量需对散热器端的供热装置进行改造且一次性投资较大,同时因用户自主节能意识还不强,分户调节尚不尽人意。代之以建筑分栋计量及调节的供热控制装置投资适中,自动化水平及工业级可靠性程度更高,是调节建筑物热负荷实现供热用户端节能目标的有效手段。建筑供热系统是一个大时滞、大惯性的复杂系统,建筑热负荷与室外环境、建筑围护结构等存在一定程度的非线性关系,利用机理建模涉及参数众多、难度大,预测结果也有较大误差。通过利用人工神经网络不依赖模型本身的特点和良好的非线性逼近能力,选择了BP (Back Propagation)与RBF (Radial Basis Function)人工神经网络的方法,根据采集到的室外干球温度、光照、风速、室内温度以及时间序列分别对建筑热负荷进行建模和预测。对比研究表明RBF神经网络预测更稳定,均方误差低于BP神经网络5.3%,更适合于建筑热负荷的预测。在对环境热负荷预测的基础上,需要为满足建筑热需求进行调节。通常是靠调节电动阀的开度对建筑物供热管网的热媒进行量调节,但这可能因水力失衡导致建筑物内的不利回路增加,造成内部冷热不均,严重时造成局部冻塞事故。因此设计了智能Bang-Bang调节,即开阀就要将阀门开至设计开度且开够一定的时长,保证管网中热媒以一定的压力和流速流经整栋建筑；关就可以彻底关断,实现供热节能间歇；但应注意到建筑物内的散热过程依然是连续的。通过计算开阀时段供给的热量与预测的热需求的差值,可得到满足工艺约束的关阀时间。由此形成了独特的开阀时段固定而关阀时段动态的变周期控制模式。将此模式与室内舒适性、管网防冻、设备故障预防等因素结合,设计实现了具有工程可用性的智能控制器。经实验测定,在天气情况稳定及热源热网供热充足的情况下,平均节能率可达10%,具有良好的社会价值和经济价值。
[Abstract]:At present, the central heating in the northern towns of our country is mostly heating center or heat exchange station to send hot water directly to the user. It is easy to cause uneven heating and waste of energy consumption. Although some regions according to the relevant suggestions and requirements of the country to charge by heat, household metering attempts. But because the household metering needs to reform the heat supply device at the end of the radiator, and the one-time investment is large, at the same time, the consciousness of energy saving of the users is not strong enough. Household regulation is not satisfactory. Instead of building building metering and regulating heating control device investment is moderate and the level of automation and industrial reliability is higher. Building heating system is a complex system with large time delay, large inertia, building heat load and outdoor environment. There is a certain degree of nonlinear relationship between the building envelope structure and so on, so it is difficult to use the mechanism to model the structure with many parameters. The prediction result also has big error, by using the artificial neural network not dependent on the characteristics of the model itself and good nonlinear approximation ability. The methods of BP back propagation and RBF Radial Basis function are selected. According to the collected outdoor dry ball temperature, light, wind speed, indoor temperature and time series, respectively, the building heat load is modeled and forecasted. The comparative study shows that the RBF neural network prediction is more stable. The mean square error is lower than that of BP neural network, which is more suitable for building heat load forecasting. It is usually by adjusting the opening of the electric valve to adjust the heat medium of the building heating network, but this may lead to the increase of the unfavorable loop in the building because of the hydraulic imbalance. Cause internal heat and cold uneven, serious local freezing plug accident. Therefore, the design of intelligent Bang-Bang regulation, that is, the valve will open to the design open to a certain length of time. Ensure that the heat medium in the pipe network flows through the whole building with a certain pressure and velocity; Close can be completely shut off, energy saving intermittent heat supply; It should be noted, however, that the heat dissipation process in the building is still continuous. The difference between the heat supplied during the open valve period and the predicted heat demand is calculated. The closing time to meet the process constraints can be obtained. Thus a unique variable period control mode with fixed opening period and closed valve period is formed. The model is combined with indoor comfort and the pipe network is frostproof. An intelligent controller with engineering availability is designed and realized by combining equipment fault prevention and other factors. The experimental results show that the average energy saving rate can reach 10% under the condition of stable weather condition and sufficient heat supply of heat source and heat network. Have good social value and economic value.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TU111;TP183

【参考文献】