【作者】 邓盛川；

【导师】 齐维贵；

【作者基本信息】 哈尔滨工业大学 ， 电力电子及电力传动， 2012， 博士

【摘要】 能源问题是21世纪面临的重要挑战。我国作为一个能耗大国，降低能耗，实现低碳经济，是我国目前的一项基本国策。我国三北地区的采暖能耗，约占全社会能耗的27.2%。供热不仅能耗大，而且效率低，每单位面积供热能耗高出发达国家的2～3倍，供热节能潜力很大。因此，研究供热节能及其控制策略的意义重大。调峰炉热力站作为供热系统调节的一种手段，在集中供热网末端加入调峰炉热力站，可以有效地解决供热高峰期负荷不足的问题，对于供热系统的稳定和良好运行起到了重要的作用。本文将研究调峰炉热力站的节能控制策略，需要从三个方面实现：一是通过供热负荷预报，准确地给出供热系统的负荷需求，防止过度供热情况的发生；二是通过优化调度，给出热力站供热系统中热负荷的最佳分配；三是通过先进的预测控制手段，保证供热系统的稳定运行，最终实现节能。供热负荷预报是进行优化调度的前提。由于供热负荷数据存在着趋势性和周期性，本文首先采用乘积季节自回归滑动平均（Autoregressive IntegratedMoving Average，ARIMA）的方法进行供热负荷预报。但该方法不适于负荷存在较大突变的情况，因此采用Kalman在线递归预测乘积季节ARIMA模型参数，该方法提高了负荷突变阶段的预报精度。为了更好处理供热负荷数据中的非线性问题，引入最小最大概率机（Minimax Probability Machine，MPM），将相空间重构与最小最大概率回归相结合进行供热负荷预报。该方法同神经网络预报法、支持向量机（Support Vector Machine，SVM）预报法进行了比较，通过仿真分析各种预报方法的性能。调峰炉热力站的优化调度是实现节能的关键。文中将供热负荷预报结果与国家供热标准相结合，给出准确的调峰炉热力站供热负荷需求，并以此负荷进行调度。调峰炉热力站的优化调度从能耗和经济性两个方面考虑，通过在供热高峰期和非高峰期对能耗和经济性不同侧重，实现优化调度的综合最佳。在寻优算法上先采用非线性规划等算法求解优化调度结果，但这些方法受到初值的选取的影响，用免疫粒子群（Immune Partile Swarm Optimization，IPSO）寻优的计算方法则避免这些方法上的缺陷。通过实例计算结果表明，免疫粒子群算法的优化调度计算更为方便和快速。预测控制是实现调峰炉热力站优化调度结果的手段。优化调度给出节能监控信号，通过预测控制达到控制目标。在建模部分采用飞升曲线和最小二乘相结合的方法，建立调峰炉热力站供热过程中的质调通道模型，锅炉温度变化扰动模型和水量变化扰动模型。为了降低锅炉温度变化扰动和水量变化扰动带来的影响，文中采用加前馈补偿的三冲量预测控制方法。其中对质调通道进行神经网络预测控制，利用改进的差分进化法进行控制律的求解，对锅炉温度变化扰动和水量变化扰动进行前馈补偿。用仿真方法证明三冲量预测控制法的有效性。最后，根据工程要求，开发调峰炉热力站监控装置。对该装置从硬件和软件两方面进行规划，给出基于PLC和GPRS无线通信的底层监控装置，并在监控中心开发优化调度软件。为了更好地满足优化调度和底层控制的需要，在MATLAB程序中实现智能算法的计算，并通过OPC技术实现与调度软件的数据交换。该监控装置样机在大庆让胡30号热力站稳定运行一年多，经过黑龙江省节能检测部门的技术检验，该装置的所有指标均达到设计要求，实现了节能。更多还原

【Abstract】 Energy problem is an important issue in the twenty-first century. Consideringthe large Energy consumption in Chin, the reduction of energy consumption and lowcarbon economy is one of basic national policy of China. Building heating is a majordomain of energy-saving, and the heating energy consumption of China’s northernregions had reached27.2%of total social energy consumption. The current problemof heating is high energy consumption but low efficiency, the heating energyconsumption in China per unit area is2and3times greater than that of thedeveloped countries, there is brilliant future in heating energy-saving. Therefore, itis significant to study the approach of heating energy-saving and control strategies.Heating station with peak-shaving boiler is a heating pattern in heating system,this pattern can effectively solve the problem of insufficient in the heating load inthe peak period, and play an important role in stability and good operation ofheating system. Energy-saving of heating station with peak-shaving can be donefrom three aspects. First, the heating load forecasting provides accurate heat load ofthe system requirements. Second, optimal dispatch makes heating station in theoptimal heat load distribution. Third, advanced control strategy achieves the aims ofenergy-saving.Heat load forecast is the premise of optimal dispatch. For the trend andperiodicity of heating load data, this thesis uses a multiple seasonal autoregressiveintegrated moving average (ARIMA) method in heating load forecasting. As thismethod is difficult to aquire accurate prediction results for where big mutations exist,so the Kalman recursive online is used to predict multiple seasonal ARIMA modelparameters in order to improve the prediction accurace of the load mutation. Tosolve nonlinear problem in heat load data, this paper introduces the minimaxprobability machine(MPM) theory and phase space reconstruction with acombination of MPM in order to forecast heating load, and compare this methodwith neural network and support vector machines(SVM) forecasting methods.Finally the thesis analyzes the performance of different forecasting methods withsimulation.Optimal dispatch of heating station with peak-shaving boiler is the key toachieve heat energy saving. This paper utilizes heat load forecasting results to meetthe national heating standards, and provides accurate output heat load of heatingstation. Optimal dispatch of heating station with peak-shaving considers the heatenergy consumption and economy to achieve comprehensive best results. Nonlinearprogramming method is used to solve optimal dispatching problems, but this method depends on initial value. Immune particle swarm (IPSO) has better search ability.The simulation results show that the immune particle swarm optimization method ismore convenient and faster than traditional optimal computing.Optimal dispatching results achieved through the control, this paper uses acombination of rising curves and least-squares method to establish quality-adjustmodel, quantity-adjust model, boiler temperature disturbance model and flowdisturbance model. In order to reduce the impact of boiler temperature disturbanceand flow disturbance, a three parameters predictive control based on feedforwardcompensation is proposed, and a neural network predictive control is used in thequality-adjust channel, and an improved differential evolution method is used tosolve the control law. Simulation results show the effectiveness of the threeparameters predictive control method.Finally, based on the engineering requirements of heating station withpeak-shaving boiler, the thesis develops a heating station monitoring device whichincludes hardware and software, PLC and GPRS wireless communication. Theintelligent algorithm was programmed by MATLAB, data exchange was realiezedbetween the optimal sofeware and MATLAB through the OPC. Operating in Ranghudistrict of Daqing for more than a year, the monitoring device achieved the purposeof energy-saving, and pass the performance tests of national quality supervisiondepartment.更多还原