【作者】 朱承治；

【导师】 曹一家； 郭创新；

【作者基本信息】 浙江大学 ， 电力系统及其自动化， 2008， 博士

【摘要】 输变电设备是组成电力系统的主要元件。在当前电力需求不断增大和电力企业商业化运营的环境下,设备的可利用率和维护成本直接关系到系统运行可靠性、企业效益与市场竞争力。因此需要设备管理部门能够及时掌握设备的运行状态和健康状态,正确地对设备进行故障检测和诊断,通过合理的检修体制预防和消除设备的故障。在全面总结我国电网企业现有检修体制的基础上,借鉴其它工业领域的研究成果,提出了电网企业优化检修（Optimal Maintenance,OM）的思想:逐步减少定期检修,避免重要设备的事后检修,推行状态检修,制订以可靠性为中心的综合检修计划,保证系统的可靠性,降低检修成本。输变电设备优化检修的核心是状态检修,其实现依赖于状态监测技术、状态评估技术、故障诊断技术、检修计划优化技术,企业信息化技术的发展。本文对这几个领域的若干关键技术进行了研究,并着重以电力变压器为例进行了应用。对状态监测中的故障检测功能进行了分析。基于非线性系统辨识模型的故障检测方法要求模型本身具有较高的辨识精度,因此提出一种基于差异进化算法（Differential Evolution,DE）和粒子群优化算法（Particle Swarm Optimization,PSO）的新型混合进化算法DEPSO,以及基于DEPSO的径向基函数神经网络（Radial Basis Function Neural Network,RBFNN）模型,并应用于预测SF₆气体绝缘变压器表面温度。该模型用DEPSO算法训练RBFNN隐层中心的数量和位置,并采用递推最小二乘法确定网络输出层的权值。对某变电站SF₆气体绝缘变压器的表面温度预测结果表明:与BP网络、基于进化规划、PSO的RBFNN相比,这种建模方法具有更高的辨识精度。输变电设备的状态评估是一个多属性决策问题,需要对其状态进行合理划分,并综合考虑监测资料、工作环境、运行检修记录,建立综合的评估指标体系。以变压器为例,根据状态评估指标具有层次性和模糊不确定性的特点,提出了一种改进的证据推理方法用于评估变压器的状态。证据的初始基本概率赋值通过层次分析法及模糊评估法获取,改进的方法适用于证据间出现高度冲突的情况。实例分析表明了该方法的有效性。指标体系的建立方法和改进的证据推理方法同样适用于输电线路和断路器等输变电设备的状态评估。提出一种新颖的多分类多核学习支持向量机变压器故障诊断方法,相对于传统的两分类支持向量机,该方法具备诸多优点。算法针对单一的优化目标函数求解,只需设计一组参数,大大降低了支撑向量机在解决多类问题中模型构造和参数选择的难度;核函数是多个基核函数的组合,提高了分类的精度;将模型分解为两个凸优问题进行求解,问题的复杂度低,求解速度快。诊断实例表明,该方法能保证很高的诊断准确率,具有很好的实用性和推广性。从输变电设备检修计划编制的实际需求出发,建立了考虑多种约束条件、以系统全年可靠性指标最优为目标的检修计划优化模型。模型中考虑了设备实际状态对其故障率的转化,以及检修对故障率的影响。针对该模型的特点,提出了一种改进的免疫算法,该算法在上一代最优抗体的基础上,通过一个较小邻域范围和一个较大邻域范围的并行搜索,使得该方法具备较强的局部寻优能力和全局寻优能力,有效提高算法的收敛速度和收敛精度。通过马尔可夫链的分析,证明了本文提出该算法的全局收敛性。最后,通过对标准IEEE RTS96单区域系统进行仿真,证明了本文模型及算法的实用性。为了提高输变电设备检修决策信息平台的集成能力及系统的柔性,提出了基于面向服务架构（Services-Oriented Architecture,SOA）的输变电设备优化检修信息系统体系结构。将优化检修系统的服务划分为应用服务、业务服务及业务流程服务,并进行了详细的功能描述及建模。将公用信息模型和模型驱动的思想用于SOA开发,并基于WebServices技术实现。SOA是一种面向动态需求的企业架构模型,它为企业应用提供了一种服务驱动的分布式协同工作新模式。基于SOA的优化检修信息集成框架能适应业务和实现技术的不断变化,极大地实现了软件的重用,降低了集成的复杂性和成本。更多还原

【Abstract】 Transmission&Distribution Equipments are the main elements which constitute power system. Under the environment that electricity demand is increasing and power utilities have been operated commercially, the availability and maintenance cost of the equipments play important roles on the system operation reliability and the benefit and competitiveness of the utilities. As a result, utilities are required to obtain the health condition of equipments in time. Moreover, they are needed to detect and diagnose failures for the equipments and then prevent or eliminate the failures according to some appropriate inspection and maintenance mechanism. Based on full summarization of the present maintenance mechanisms in power grid and research achievements obtained in other industrial fields, a more reasonable mechanism aimed to optimal maintenance is put forward as below. Utilities should reduce the frequency of periodic maintenance, and avoid applying the breakdown maintenance method on critical elements. The condition based maintenance has a promising future which should be carried out gradually. All of these methods must be coordinated by reliability-centered maintenance scheduling in order to ensure the reliability of the system and the reduction of the maintenance cost. The condition based maintenance is the core of the optimal maintenance. Realization of the optimal maintenance mechanism for Transmission&Distribution Equipments is based on the development of several technologies, such as condition monitoring, condition assessment, fault diagonosis, optimization of maintenance scheduling, and enterprise informationization. Some key problems of these technologies are studied in the dissertation, and mainly applied to the maintenance of power transformers.In Chapter 2. the fault detection technology of condition mornitoring is mainly analyzed. Fault detection based on nonlinear system identification demands high identification accuracy, so a novel radial basis function neural network （RBFNN） model based on a hybrid learning algorithm differential evolution and particle swarm optimization （DEPSO） is proposed in this paper to predict the shell temperature for SF6-insulated transformers. The DEPSO automatically adjusts the number and positions of hidden layer RBF centers. The weights of output layer are decided by the recursive least squares algorithm. The proposed DEPSO-RBFNN model is trained and tested based on the field data collected from a SF₆-insulated transformer. The test results indicate that the DEPSO-RBFNN possesses far superior identification precision than BP neural network （BPNN）, EP-RBFNN and PSO-RBFNN.Transmission&Distribution Equipment condition assessment is a multiple-attribute decision-making （MADM） problem. On the basis of rational partition of condition, a synthetic evaluation index system is needed, which includes condition mornitoring records, working surroundings, operation and maintenance history et al. Considering the fuzziness and uncertainty of condition assessment indices, an improved evidential reasoning （ER） approach to the transformer condition assessment is presented. The initial basic probability assignment of evidence is accessed by means of analytic hierarchy process （AHP） and fuzzy evaluation method. The newly approved method is applied to the case in which high conflict occurs. The results of an example analysis present its effectiveness. The method to build the evaluation index system and the improved evidential reasoning are also suitable when applied to the condition assessment of other Transmission&Distribution Equipments, such as breakers and transmission lines.In this thesis, Chapter 4 develops a novel support vector machine （SVM）, i.e. multiple kernels learning multicategory support vector machine （MKLM-SVM）, for the faults diagnosis in transformers. Unlike traditional SVM that may fail under some various circumstances, the MKLM-SVM method has some good theoretical properties, the MKLM-SVM method is only based on a simple objective function, and the classification results can be directly calculated on the basis of a simple decision function; the MKLM-SVM method can calculate an optimal kernel function on the basis of linear combinations of basic kernels, further boosting the overall performance; the solutions for the MKLM-method can be efficiently calculated by iteratively solving two convex optimization functions with a low computation cost. The test results show that the proposed method has high classification accuracy, which proves its effectiveness and usefulness.A new model for Transmission&Distribution Equipments maintenance scheduling which intends to find the most reliable maintenance schedule without violating any restrictions is proposed in this paper. The scheduling problem is solved considering the mapping between the equipment condition and equipment failure rate. A functional relationship between failure rate and maintenance measures has also been developed, which assess the impact of maintenance. An improved immune algorithm is proposed to power system maintenance scheduling optimization problem. In order to realize the parallel global and local search capabilities, this algorithm generates the next population under the guidance of the previous superior antibodies in a small and a large neighborhood respectively. Through analysis of Markov chain, the proposed method is proved to be convergent on whole solution space. The test results on IEEE RTS96 single area system demonstrate that the proposed model and optimization method are effective.In order to facilitate information integration and improve flexibility of Transmission&Distribution Equipment Optimal Maintenance Information System （TDE-OMIS）, a Service-Oriented Architecture （SOA） based framwork is introduced. The SOA based TDE-OMIS include several kinds of service,such as application services, business services and business process services, which have been functional described and modeled in detail. The Common Information Model （CIM） and the Model-Driven （MD） methodology are introduced into the realization of the SOA based on the Web Services technology. SOA is a dynamic demand-oriented enterprise architecture model, which provides a new kind of service-driven and distributed collaborative operation mode. Service-Oriented optimal maintenance information integration framework can adapt the continuous change of business and technology, so it can reuse the software and reduce the complexity and cost of information integration greatly.更多还原