【作者】 王彦；

【导师】 刘宏立；

【作者基本信息】 湖南大学 ， 电路与系统， 2012， 博士

【摘要】 在城市化建设发展进程中,建筑能耗比重越来越大。目前,我国建筑能耗占社会总能耗的27.5%,建筑能耗不仅给国家、社会造成了能源负担,很大程度上也制约了经济的可持续发展。据能源界的研究和实践,普遍认为建筑节能是各种节能途径中节能潜力最大、节能效果最好、最直接的方式。建筑节能研究已成为节能研究的一个重要方向,且建设大型能耗监控系统是建筑节能运行的重要手段。本文围绕大型能耗监控系统通信网络、节能控制策略及暖通空调(HVAC)系统故障检测诊断与恢复这三个方面展开了研究,取得了一定的成果。(1)深入研究了大型能耗监控系统通信网络。在能耗数据的近程传输过程中,针对低压电力线载波通信时噪声干扰强、噪声分布具有频段选择性,且接入负载复杂多变导致其信道很难实现有效估计和预测等特性,采用基于跳频OFDM的电力线载波通信实现能耗数据的近程传输。着重研究了跳频OFDM电力线载波通信网络的传输性能及优化策略,通过推导BPP模型下的跳频OFDM电力通信网的传输中断概率函数,分析不同网络通信参数对网络传输性能的影响,在此基础上研究了一种效果和效率兼备的最优参数组搜索算法、以优化跳频OFDM电力线载波通信网络的传输性能。(2)为有效提高能耗数据远程传输系统容量并有效抵抗多址干扰,采用群正交多载波分组通信(GO-MC-CDMA)技术实现能耗数据的远程传输。着重研究了其多用户检测算法和信道估计算法：在考虑载波频偏的情况下,分析了GO-MC-CDMA上行链路采用最大似然多用户检测时的系统误比特率性能,给出了闭式解,分析和仿真结果均表明：在GO-MC-CDMA系统中采用最大似然多用户检测不仅复杂度不高,而且获得了较好的对抗载波频偏的能力；针对GO-MC-CDMA系统中LS信道估计的缺陷,提出了改进方案,并研究了改进的LS信道估计算法中估计初值数目与系统模型失配误差以及噪声影响之间的关系,根据该关系提出了估计初值数L的调整规则,然后用模糊算法实现L的自适应调整,该模糊自适应LS信道估计算法能够根据信噪比的变化自动调整最佳L值,保证理想的系统误比特率性能。经分析,该算法复杂度远低于LMMSE信道估计算法,更具实用价值。(3)针对暖通空调系统存在时滞、非线性的特点,研究了自适应Smith预估器以克服时滞对系统稳定性的影响,引入自适应算法来提高Smith预估器的自适应能力。在分析改进型Smith预估补偿控制方案的基础上,研究了预估器参数不匹配时,不同tf情况下的系统特性曲线,总结出滤波时间常数tf的调整规律,并提出自适应Smith预估补偿控制方案；利用FPAA的高速特性,提出了一种响应速度快、反馈时间短的基于FPAA的模糊自整定PID控制策略,减小了控制器本身时滞。(4)为提高控制精度、减小时变影响,对神经网络PID控制进行优化并应用于HVAC系统中：针对BP神经网络学习过程收敛速度慢及易陷入局部极小值的缺陷,深入研究了阻尼最小二乘(LM)算法。为解决选择学习速率和求解逆矩阵会严重影响训练时间和收敛精度的缺陷,利用LU分解法对LM算法进行了改进和优化,利用MATLAB平台对其仿真,将改进后的LMBP神经网络PID控制器应用于暖通空调冷冻水循环的控制回路中,将其控制效果与PID控制算法、BP神经网络PID控制算法进行了仿真对比研究；针对模糊神经网络PID控制器中参数初始值的设置对控制器性能影响大的特点,提出了一种改进的PSO算法优化模糊神经网络PID控制器参数策略。(5)针对HVAC系统中常用执行器及传感器的典型故障,提出一种基于小波神经网络(WNN)和希尔伯特-黄变换(HHT)相结合的故障诊断方法,并对执行器和传感器的四种故障(完全失效、偏差、漂移、精度下降)进行了仿真研究,仿真实验结果表明该方法可以有效地提高故障诊断的准确率更多还原

【Abstract】 In the process of urbanized construction and development, the proportion of building energy consumption is increasing. In our country, the total building energy consumption occupies27.5%of the total social energy consumption. The building energy consumption has imposed the large energy burden on the state and our society, and restricted sustainable economic development to some extent. According to the research and practice from the field of energy, how to save energy efficiently from building energy consumption is currently considered as the most direct and effective way. The study of how to save building energy has become an important research issue, and it is main way that building of the Energy-efficient Monitor to save energy. This paper focuses on the design of energy monitoring network of the large-scale public buildings energy-saving and control strategies, fault detection diagnosis and recovery of heating ventilating and air conditioning (shorted as HVAC) system. Our main contributions are:(1) We have studied the large-scale energy monitoring network in depth. For consumed energy short range transmission, aiming at some special characteristics including the strong noise interference of low voltage distribution network, the frequency selectivity of the noise distribution and the complex access load leading hard to realize effective estimation and forecast for communication channel, we design a frequecy hopping OFDM-based power line carrier communication scheme for the short range transmission of consumed energy data. Especilly, we emphasize the transmission performance of frequency hopping OFDM power line carrier communication network and optimization strategies. By deducing the transmission interrupt probability function of the frequency hopping OFDM power line communication network for BPP model,and analyzing the influence of different network communication parameters on the network transmission performance,. We propose an optimal parameter group searching algorithm to optimize the transmission performance of frequency hopping OFDM power line communication network.(2) We use the group-orthogonal multi-carrier code-division multiple-access (shorted as GO-MC-CDMA) as the way of remote energy data transmission to effectively improve the system capacity and resist multiple access interference. We emphasize the multi-user detection algorithm and channel estimation algorithm,which take the Carrier-Frequency Offset (CFO) into consideration and give the system bit error rate performance of the GO-MC-CDMA uplink by using the Maximum Likelihood (ML) multi-user detection,and also give the closed-form solution..Numerical simulation results show that the GO-MC-CDMA system using ML detection not only has low complexity, but also has out-standing performance of bit error rate and against carrier frequency offset. Then we present an improved scheme to overcome some defects of LS channel estimation in GO-MC-CDMA system. In addition, we exploit the relationship between the mean-squared error (MSE) and the number of chosen first estimate value in the LS channel estimation algorithms.Based on the above relationship, we propose some rules to estimate the initial value L, and then use fussy algorithm to realize the L self-adaptive modification.The self-adaptive LS channel estimation fuzzy algorithm can automatically choose the optimal L value according to the change of the signal to noise rate,which can guarantee perfect performance of the system error bit rate. The complexity of the fussy algorithm is far below the LMMSE channel estimation algorithm, and more practical.(3) Aiming at the characteristics of the HVAC system, such as time-delay and nonlinear, we design the Smith Predictor for reducing the influence on system stability because of time-delay and also design an adaptive algorithm for improving the adaptive ability of smith predictor. On the basis of analyzing the improved smith predictive scheme, we study the adjusted rules of filtering time constant tf, and propose the self-adaptive Smith predictive compensated controlling scheme. An fuzzy adaptive PID control strategy with fast response and short feedback time based on FPAA is proposed to reduce the time-delay of controller.(4) We optimize the neural network PID control and apply it to the HVAC system to improve control accuracy and to reduce time-vary effect. In allusion to the defects of the slowly converging and easily immerging in partial minimum in the learning process of Back Propagation (BP) Neural Network (NN), a Levenberg-Marquardt (LM) algorithm has been presented. In order to solve the two problems including the choice of learning rate and the inverse matrix solving in LM algorithm, which seriously influence on both training time and converging accuracy, we use the LU decomposition method to improve the LM algorithm, and the effect of which is simulated by MATLAB. In this paper, we apply the LMBP Neural Network PID controller in cooling water cycle of heating ventilating and air conditioning (HVAC) system, also simulate and compare the results of LMBP neural network PID controller, PID controller, and BP neural network PID controller. Because the settings of initial parameter values in fuzzy neural network PID controller have an important influence on the performance of the controller, we propose an improved PSO algorithm to optimize the parameters of fuzzy neural network PID controller.(5) For solving the typical bugs of sensors and actuators in HVAC system, this paper presents a diagnosis method based on HHT-WNN. Simulation results of four typical sensor bugs (i.e. complete failure, bias failure, drift failure and decreased accuracy) show that this method can improve effectly the diagnosis accuracy, compared with BP neural network.更多还原