建筑室内环境建模、控制与优化及能耗预测

发布时间：2018-03-21 17:41

本文选题：建筑环境优化　切入点：建筑能量预测　出处：《浙江大学》2013年博士论文　论文类型：学位论文

【摘要】：当今,能源危机和环境污染是世界各国面临的共同挑战。我国是能源消耗大国,随着经济的飞速发展,建筑能耗已占社会总能耗的四分之一以上,且这一比例还在逐年升高。在此背景下,建筑节能技术越来越受到重视。从控制学角度看,建筑能量系统可看作多变量、非线性的复杂系统,建筑节能目标的实现涉及到建筑环境的优化控制、建筑能量的预测与管理等诸多方面内容。本文立足控制学科,在前人的基础上,从建筑环境控制与能量管理等相关领域入手,进行了如下研究工作。 ·出于节能和室内环境优化的目的,暖通空调系统对热环境的控制与优化需要室内温度分布的动态信息帮助决策。通常,计算流体力学(CFD)模型能提供这样的精确信息,但是由于其迭代计算繁杂、耗时长,难以满足实时性要求。本文引入基于本征正交分解(POD)的模型降阶技术,与CFD仿真结合提出一种新的建筑热环境建模方法,可同时满足热环境建模的精度和实时性要求。POD模型降阶是种映射方法,配合离散化技术,该方法可将无限维的非线性复杂系统变为仅与POD模式系数相关的低阶线性系统。具体建模方法为：首先,利用CFD工具对建筑室内热环境进行动态仿真,在此期间用快照的方式采集动态温度场信息；其次,运用有限体积法对能量平衡方程进行空间和时间的离散化,并建立离散能量平衡方程的状态空间表达式：然后,运用POD方法对室内动态温度场进行降阶,并运用Galerkin映射将高阶能量方程投射到降阶子空间上,从而得到阶数十以内的低阶线性系统。在一个二维房间的仿真实验中,室内温度场的POD降阶模型得到与CFD仿真基本一致的瞬、稳态精度,而其阶数低至六阶,证明了该方法的有效性。 ·设计一个基于POD降阶模型的室内温度精确控制系统。该温控系统的特点在于,运用“离线-在线”策略建立了室内温度场的动态降阶模型,以实时反馈热环境信息提高温控精度。降阶温度场的初始状态通过一个温度传感器和卡尔曼滤波器估计得到。本文分别设计了单神经元自适应PID控制器和模型预测控制器对该方法进行仿真验证。结果表明,在速度场不变等假设条件下,该温控系统可利用室内温度场信息精确控制各区域温度,具有提高热舒适度和降低能耗的潜力。 ·针对目前建筑室内环境的优化策略大都忽视环境参数空间分布的问题,本文利用POD降阶技术在环境建模方面的快速性和准确性,运用多维插值和遗传算法,设计一种综合考虑室内热舒适度、室内空气质量(IAQ)及空调能耗的优化控制策略。其中,建筑室内环境参数,包括温度场、气流场、CO2浓度分布、及热舒适度分布等,先通过CFD仿真得到；然后利用POD方法重构上述参数分布的低阶变化空间。优化方法采用遗传算法,控制变量包括置换通风系统的送风温度和速度。优化目标涵盖系统能耗、室内热舒适度、IAQ、以及垂直温差等。在遗传算法的每次优化迭代中,通过POD参数空间内的多维插值快速求解候选控制变量对应的系统响应,确保了优化算法的实时性。一个办公室环境的优化仿真证实该方法的有效性。 ·作为典型的数据驱动建模方法,在过去20年间,人工神经网络在建筑能耗预测领域应用广泛。本文结合自适应模糊推理系统(ANFIS)和遗传算法的各自特点提出一种新的建筑能耗预测方法,即GA-ANFIS方法。其中,ANFIS通过训练输入/输出数据自适应调整T-S模糊系统的隶属度函数参数和结论参数,遗传算法则对ANFIS中的模糊规则参数进行优化以帮助构造最优规则基。设计ANFIS的分级结构用于应对输入变量过多造成的维数灾难问题。利用美国采暖、制冷与空调工程师学会(ASHRAE)提供的建筑能耗数据对该方法进行验证。结果表明,该方法与神经网络方法相比其建模时间在同一尺度内,而预测精度最多可提高20%。 ·本文利用GA-ANFIS方法分别对玉泉图书馆和杭州某酒店的电力能耗作预测实验。能耗数据均由浙大中控能耗监控系统实时采集,气象数据来自浙江省气象局官方资料。实验结果验证了仿真结论,GA-ANFIS方法可结合目前的建筑能量采集系统,应用于建筑未来能耗的预测与分析。
[Abstract]:Nowadays, energy crisis and environmental pollution are common challenges facing the world. China is a big energy consuming country, with the rapid development of economy, the building energy consumption has accounted for 1/4 of the total social energy consumption, and the proportion is increasing year by year. Under this background, building energy-saving technology has attracted more and more attention from the control point of view. That building energy system can be seen as a multi variable, nonlinear complex system, to achieve the goal of building energy-saving control relates to the optimization of building environment, many aspects of building energy prediction and management. Based on the control subjects, on the basis of previous studies, starting from the construction of environmental control and energy management and other related fields, are as follows the research work.
And for the purpose of saving energy and optimizing indoor environment, HVAC system control and optimization of thermal environment dynamic information of indoor temperature distribution decision-making. Usually, the computational fluid dynamics (CFD) model can provide accurate information to this, but because of the iterative calculation of complicated, time-consuming, difficult to meet the real-time requirements. This paper introduced based on the proper orthogonal decomposition (POD) reduction technique combined with CFD simulation model, put forward a new modeling method for building thermal environment, thermal environment modeling can meet the accuracy and real-time requirements of model reduction is.POD
A mapping method, with the discretization technique, this method can get nonlinear infinite dimensional complex system is only related to POD mode coefficients of low order linear system. The modeling method is as follows: firstly, the simulation of indoor thermal environment by using CFD tools, with a snapshot of the way of collecting information of dynamic temperature field during this period; secondly, the energy balance of discrete space and time equation using the finite volume method, and the discrete energy balance equation of state space expression. Then, using POD method to reduce the order of dynamic indoor temperature field, and using the Galerkin mapping of high order equation of energy reduction is projected onto the subspace, resulting in a low order linear system. In order for within a two-dimensional simulation room, indoor temperature field of POD reduced order model consistent with the CFD simulation of transient and steady-state accuracy, and its order The validity of the method is proved by the low to six order.
Design a reduced order model based on POD indoor temperature control system. The precise features of temperature control system, the use of "offline - online" strategy to establish a dynamic indoor temperature field of the reduced order model, real-time feedback information to improve the accuracy of temperature. The thermal environment reduced the initial state order temperature by a temperature sensor and Calman filter estimated. This paper designs a single neuron adaptive PID controller and model predictive controller to verify the method. The results show that the velocity field is invariant under assumed conditions, the control system can use the information of indoor temperature field accurately control the temperature of the area, can improve thermal comfort and reduce energy consumption potential.
For most of the optimization strategy of the construction of the indoor environment neglect environment parameters of spatial distribution, rapidity and accuracy of the POD reduction technique in environmental modeling, using a multidimensional interpolation algorithm and genetic algorithm, design a comprehensive consideration of the indoor thermal comfort, indoor air quality (IAQ) and the energy optimal control strategy. Among them, the indoor environment parameters, including temperature field, flow field, concentration distribution of CO2, and the thermal comfort distribution, first obtained by CFD simulation; and then use the POD method to reconstruct the distribution parameters of the low order change space. Using genetic algorithm optimization method, the control variables include wind speed and temperature to displacement ventilation system the optimization goal of covering the system energy consumption, indoor thermal comfort, IAQ, and the vertical temperature difference. In each iterative optimization of genetic algorithm, the POD parameter space multidimensional interpolation fast The response of the candidate control variable is solved to ensure the real-time performance of the optimization algorithm. The optimization simulation of an office environment proves the effectiveness of the method.
As a typical data-driven modeling method, in the past 20 years, artificial neural network is widely used in the prediction of the energy consumption of the building. This paper combines adaptive fuzzy inference system (ANFIS) characteristics of genetic algorithm and proposes a new building energy consumption prediction method, namely GA-ANFIS method. Among them, ANFIS through membership functions and conclusion the training parameters of input / output data adaptive T-S fuzzy system, genetic algorithm for the parameters of the fuzzy rules in ANFIS are optimized to help construct the optimal rule base. The dimension disaster response caused by excessive input variables for a hierarchical structure design of ANFIS. Using the heating, refrigeration and Air Conditioning Engineers (ASHRAE) to verify the the method of building energy consumption data. The results show that this method and neural network method compared with the modeling time in the same scale, and the prediction accuracy The maximum can be improved by 20%.
In this paper, using GA-ANFIS method of power consumption of the Jade Spring respectively the library and a hotel in Hangzhou for the prediction experiments. The energy consumption data by SUPCON real-time energy consumption monitoring system, meteorological data from Zhejiang Provincial Meteorological Bureau official data. Experimental results verify the simulation results, the GA-ANFIS method can be combined with the building energy acquisition system at present, forecast and analysis the future application in building energy consumption.

【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2013
【分类号】：TU111.195

【参考文献】