毕设外文翻译

英文翻译

题目:

通过分析变压器中溶解气体而进行故障诊断的专家系统

姓名: 宋日成

学院: 工学院

专业: 自动化

班级: 自动化112 班学号: 32211218

指导教师: 陆静职称: 讲师

2015年3月14日

南京农业大学教务处制

An Expert System for Transformer Fault Diagnosis Using

Dissolved Gas Analysis

W. S. Chan ·Y. L. Xu ·X. L. Ding ·W. J. Dai

Received: 9 November 2005 / Accepted: 11 August 2006 / Published online: 7 September 2006? Springer-Verlag 2006

Abstract In order to automate the transformer fault diagnosis, improve the accuracy of judgment, the introduction of artificial intelligence expert system fault diagnosis. The system is based on fuzzy reasoning confidence by observing the information, the use of knowledge to reach a conclusion. Test proved this method reduces the randomness of judgment, improved diagnosis.

Keywords:Expert System Transformer Troubleshooting

1 Introduction

The power transformer is a major apparatus in a power system, and its correct functioning is vital to system operations. In order to minimize system outages,many devices have evolved to monitor the serviceability of power transformers. These devices, such as, Buchholz relays or differential relays, respond only to a severe power failure requiring immediate removal of the transformer from service, in which case, outages are inevitable. Thus, preventive techniques for early detection faults to avoid outages would be valuable. In this way, analysis of the mixture of the faulty gases dissolved in insulation oil of power transformer has received worldwide recognition as an effective method for the detection of incipient faults. Many researchers and electrical utilities have reported on their experience and developed interpretative criteria on the basis of DGA. However, criteria tend to vary from utility to utility. Each approach has limitations and none of them has a firm mathematical description. Therefore, transformer diagnosis is still in the heuristic stage. For this reason, knowledge-based programming is a suitable approach to implement in such a diagnostic problem. Based on the interpretation of DGA, a prototype of an expert system for diagnosis of suspected transformer faults and their maintenance procedures is proposed. The significant source in this knowledge base is the gas ratio method. Some limitations of this approach are overcome by incorporating the diagnostic procedure and the synthetic expertise method. Furthermore, data bases adopted from TPCS gas records of transformers are incorporated into the expert system to increase the practical performance. Uncertainty of diagnosis is managed by using fuzzy set concepts. This expert system is constructed with rule based knowledge representation, since it can be expressed by experts. The expert system building tool, Knowledge Engineering System (KES), is used in the development of the knowledge system

because, it has excellent man-machine interface that provides suggestions. Moreover, its inference strategy is similar to the MYCIN. A famous rule-based expert system used for medical diagnosis. The uncertainty of human qualitative diagnostic expertise, e.g., key gas analysis, and another quantitative imprecision, such as, norms threshold and gas ratio boundaries etc., are smoothed by appropriate fuzzy models. With the results of such implementation, different certainty factors will be assigned to the corresponding expertise variables. Both event-driven (forward chaining) and

goal-driven (backward chaining) inferences are used in the inference engine to improve the inference efficiency. To demonstrate the feasibility of the proposed expert system, around hundreds of TPC historical gas records have been tested. It is found that more appropriate faulty types and maintenance suggestions can support the maintenance personals to increase the performance of transformer diagnosis.

2.DEVELOPMENT OF DIAGNOSIS AND INTERPRETATION

Like many diagnostic problems, diagnosis of an oil-immersed power transformer is a skilled task. A transformer may function well externally with monitors, while some incipient deterioration may occur internally to cause a fatal problem in the latter development. According to a Japanese experience, nearly 80% of all faults result from incipient deteriorations. Therefore, faults should be identified and avoided at the earliest possible stage by some predictive maintenance technique. DGA is one of the most popular techniques for this problem. Fault gases in transformers are generally produced by oil degradation and other insulating materials, e.g., cellulose and paper. Theoretically, if an incipient oractive fault is present, the individual dissolved gas concentration, gassing rate, total combustible gas (TCG) and cellulose degradation are all significantly increased. By using gas chromatography to analyze the gas dissolved in a transformer's insulating oil, it becomes feasible to judge the incipient fault types. This study is concerned with the following representative combustible gases; hydrogen(H2), methane(C2H 2), ethane(C2H6), ethylene(C2H4), acetylene(C2H2) and carbon monoxide(C0). Many interpretative methods based on DGA to diagnose the nature of incipient deterioration have been reported. Even under normal transformer operational conditions, some of these gases may be formed inside. Thus, it is necessary to build concentration norms from a sufficiently large sampling to assess the statistics. TPC investigated gas data from power transformers to construct its criteria. The developed knowledge base in this paper is partially based on these data. On the other hand, Dornerburg developed a method to judge different faults by rating pairs of concentrations of gases, e.g., CH /H , C H /C2H4, with approximately equal solubility and fusion coefficients. Rogers established mare comprehensive ratio codes to interpret the thermal fault types with theoretical thermodynamic assessments. This gas ratio method was promising because it eliminated the effect of oil volume and simplified the choice of units. Moreover, it systematically classified the diagnosis expertise in a table form. Table 1 displays the ratio method as proposed by Rogers . The dissolved gas may vary with the nature and severity of different faults. By analyzing the energy density of faults, it's possible to distinguish three basic fault processes: overheating(pyrolysis),corona (partial discharge) and arcing discharge.

Corona and arcing arise from electrical faults, while overheating is a thermal fault. Both types of faults m y lead to deterioration, while damage from overheating is typically less than that from electrical stress. Infect, different gas trends lead to different faulty types, the key gas method is identified.' For example large amounts of C H and H are produced with minor arcing fault 4. quantities of CH 2aid C2H4 2 may be a symptom of an arcing fault.

3.THE PROPOSED DIAGNOSTIC EXPERT SYSTEM

This study is aimed at developing a rule-based expert system to perform transformer diagnosis similar to a human expert. The details of system processing are described below.

3.1 The Proposed Diagnostic Method

Diagnosis is a task that requires experience. It is unwise to determine an approach from only a few investigations. Therefore, this study uses the synthetic expertise method with the experienced procedure to assist the popular gas ratio method and complete practical performance.

3.1.1 Experienced Diagnostic Procedure

The overall procedure of routine maintenance for transformers is listed. The core of this procedure is based on the implementation of the DGA technique. The gas ratio method is the significant knowledge source. Some operational limitations of the gas ratio method exist. The ratio table is unable to cover all possible cases. Minimum levels of gases must be present. The solid insulation involving CO and CO are handled separately and the gas ratio codes have been developed mainlyfrom a

free-breathing transformer. Other diagnostic expertise should be used to assist this method. Norms, synthetic expertise method and data base records have been incorporated to complete these limitations. The first step of this diagnostic procedure begins by asking DGA for an oil sample to be tested. More important relevant information about the transformer's condition, such as the voltage level, the preservative type, the on-line-tap-changer (OLTC) state, the operating period and degassed time must be known for further inference. Noms (criteria) Set up by TPC power transformers' gas characteristic data are then used to judge the transformers' cactive fault is present, the individual dissolved gas concentration, gassing rate, total combustible gas (TCG) and cellulose degradation are all significantly increased. By using gas chromatography to analyze the gas dissolved in a transformer's insulating oil, it becomes feasible to judge the incipient fault types. This study is concerned with the following representative combustible gases; hydrogen(H2), methane(C2H 2), ethane(C2H6), ethylene(C2H4), acetylene(C2H2) and carbon monoxide(C0). Many interpretative methods based on DGA to diagnose the nature of incipient deterioration have been reported. Even under normal transformer operational conditions, some of these gases may be formed inside. Thus, it is necessary to build concentration norms

from a sufficiently large sampling to assess the statistics. TPC investigated gas data from power transformers to construct its criteria. The developed knowledge base in this paper is partially based on these data. On the other hanondition. For the abnormal cases, the gas ratio method is used to diagnose transformer fault type. If different or unknown diagnosis results are found from these ratio methods, a further synthetic expertise method is adopted. After these procedures, different severity degrees are assigned to allow appropriate corresponding maintenance suggestions.

3.1.2 Synthetic Expertise Method

The ratio trend, norms threshold, key gas analysis and some expertise are considered as different evidences to confirm some special fault types. In other words, more significant evidences have been collected for some special fault type, better assessment of the transformer status is obtained. The ratio trend can be seen as a modification of the conventional gas ratio and key gas method. Obviously, the above gas trends should be incorporated with other evidences under the experienced procedure for practical use. Noms threshold, the gassing rate, the quantity of total combustible gas (TCG), the TPC maintenance expertise and the fuzzy set assignment are all important evidences considered in the synthetic diagnosis. Other expertise based on a transformer historical data base is also used to analyze the characteristics

of a case transformer. Section 3 . 4 gives some details of these rules. 3.2 Expert System Structure The proposed diagnostic expert system is composed of four components, working memory, a knowledge base, an inference engine and a

man-machine interface. Working memory (global data base) contains the current data relevant to solve the present problem. In this study, most of the diagnostic variables stored in the data base are current gas concentration, some are from the user, others are retrieved from the transformer's historical data base. Note that the fuzzy set concept 1s incorporated to create fuzzy variables on the request of system reasoning.

A knowledge base is the collection of domain expertise. It contains facts and knowledge relationship, which uses these facts, as the basis for decision making. The production rule used in this system is expressed in IF-THEN forms. A successful expert system depends on a high quality knowledge base. For this transformer diagnostic system, the knowledge base incorporates some popular interpretative methods of DGA, synthetic expertise method and heuristic maintenance rules. Section 3.4 will describe this knowledge base. Another special consideration in the expert system is its inference engine. The inference engine controls the strategies of reasoning and searching for appropriate knowledge. The reasoning strategy employs both forward chaining (data-driven) and backward chaining (goal-driven). Fuzzy rules, norms rules, gas ratio rules, synthetic expertise rules and some of the maintenance rules are implemented by backward chaining. Other rules, such as, procedure rules and some maintenance rules, use forward chaining. As for the searching strategy in KES, the depth first searching and short-circuit evaluation are adopted. The former can improve the search efficiency by properly arranging the location of significant rules in the inference procedures. The latter strategy only searches the key conditional statements in the antecedent that are responsible for establishing whether the entire

rule is true or false. Taking the advantages of these two approaches in the building and structuring of a knowledge base improves inference efficiency significantly. As for man-machine interface. KES has an effective interface which is better than typical knowledge programming languages, such as, PROLOG or LISP. With the help of this interface, the capability of tracing, explaining and training in an expert system is greatly simplified.

4. IMPLEMENTATION OF THE PROPOSED EXPERT SYSTEM

An expert system is developed based on the proposed interpretative rules and diagnostic procedures of the overall system. To demonstrate the feasibility of this expert system in diagnosis, the gas data supported by MTL of TPC have been tested. In Taiwan, the MTL of TPC performs the DGA and sends the results to all acting divisions relating to power transformers. In return, these acting divisions are requested to collect and supply their transformer oil samples periodically. After analyzing oil samples, more than ten years' worthy gas records are collected and classified into three voltage levels, 69KV，161KV and 345KV. Thus, gas records for one transformer are composed of several groups of data. In the process of DGA interpretation, all of these data may be considered, but only the recent data which have significant effects on diagnosis are listed in the later demonstration. In MTL, all gas concentrations are expressed by ppm in volume concentration. 100 ppm is equal to 0.01 ml (gas)/100 ml (oil). From the expertise of diagnosis, the normal state can be confirmed only by inspection of the transformers' norms level. In practice, most of the transformer oil samples are normal, and this can be inferred successfully on the early execution of this expert system. However, the Success of an expert system is mainly dependent on the capability of diagnosis for the transformers in question. In the implementation, many gas records which are in abnormal condition are chosen to test the Justification of this diagnostic system. A total of 101 transformer records have been executed and the results are summarized in Table 5. Among those implemented, three are listed and demonstrated. Shown in Table 5 are the test results of 101 units of transformers in three types of remedy: normal, thermal fault and arc fault. After comparing them with the actual state and expert judgment, a summary of results was obtained. As previously stated, one unit of transformer may include many groups of gas data. In evaluation, we depicted some key groups in one unit to justify because some transformers may have different incipient faults during different operational stages. Some mistakes implemented from testing are caused by the remaining oil in the oil sampling container, unstable gas characteristics of the new degassing sample and some obscure gas types. If more information or new techniques support other uncertain membership functions, they can be added into the knowledge bas to enlarge the performance of this prototype expert system. Furthermore, the parameters described in Table 2, 3 and 4 are suitable for TPC power transformer. Different regions may be modified if the maintenance personnel find more suitable system parameters.

5.CONCLUSIONS

A prototype expert system is developed on a personal computer using KES. It can diagnose the incipient faults of the suspected transformers and suggest proper maintenance actions. Fuzzy set concept is used to handle uncertain norms thresholds, gas ratio boundaries and key gas analysis. The synthetic method and diagnostic procedure are proposed to assist the situation which can not be handled properly by the gas ratio methods. Results from the implementation of the expert system shows that the expert system is a useful tool to assist human experts and maintenance engineers. The knowledge base of this expert system is incorporated within the popular interpretative method of DGA, synthetic expertise and heuristic maintenance rules. The data base supported by TPC MTL for about 10 year collection of transformer inspection data is also used to improve the interpretation of diagnosis. Through the development of the proposed expert system, the expertise of TPC MTL can be reserved. In addition, this work can be continued to expand the knowledge base by adding any new experience, measurement and analysis techniques.

通过分析变压器中溶解气体而进行故障诊断的专家系统

W. S. Chan ·Y. L. Xu ·X. L. Ding ·W. J. Dai

Received: 9 November 2005 / Accepted: 11 August 2006 / Published online: 7

September 2006? Springer-Verlag 2006

摘要：为了实现变压器故障诊断的自动化，提高判断的准确性，把人工智能专家系统引入到故障诊断中。该系统采用基于置信度的模糊推理方法，由观测到的信息，利用知识库来推断结论。试验证明此方法降低了判断的随机性，提高了诊断水平。

关键词：专家系统变压器故障诊断

1.导言

电力变压器是电力系统中一个主要的仪器，其正确的运行对系统至关重要。为了尽量减少系统的中断时间，许多设备已变得可以监察电力变压器的可用性。这些设备，例如，巴克霍尔兹继电器或差动继电器，是为了对付一个严重的需要立即拆除变压器的电力故障，在这种情况下，中断是不可避免的。因此，预防技术，及早发现故障，以避免中断将是非常值得的。在这方面的方式，分析电力变压器绝缘油中溶解气体已得到世界各地一致承认的作为一种有效的检测故障方法。许多研究人员和电力部门汇报了他们关于溶解气体分析的经验。然而，标准往往不同的，从一种方法到另一种方法。每种方法都有其局限性，并没有具有准确的数学描述。因此，变压器的诊断仍然是在启发式阶段。基于这个原因，以此知识为基础的编程是一个合适的方式去落实这样的诊断问题。以说明绝缘油中溶解气体为基础，专家建议原型研制一用以监视变压器故障的诊断专家系统的原型。在这方面的设计基础的主要来源是气体比值的方法。这种方法的局限性通过纳入诊断程序和综合专业知识而被克服。此外，以采用TPC变压器气体分析方法为基础的数据资料库被纳入专家系统，以加强实际的表现。不确定性管理的诊断是用模糊集概念。这个专家系统是以基于规则而建造的，因为它可以被该系统所表达出来。专家系统的工程工具，知识工程系统（KES ），是用来在发展的知识体系，因为它具有良好的人机界面。此外，它的推理方法是类似MYCIN的。一个著名的以规则为基础的专家系统用于医疗诊断。人类定性诊断的专业知识的不确定性，例如，关键气体分析，以及定量的不精确性，例如，精确的数值和天然气的比例界限等，被模糊模型变得合适了。由于执行情况等的结果，不同的确定性的因素将被分配到相应的专业知识的变量。这两个事件驱动（向前链结）和目标驱动（反向链）的推论是用在推理机制上，以改善推理效率。证明的可行性建议的专家系统，大约有数以百计的气体的TPC 的历史纪录，已经经过测试。更合适的故障类型及维修的建议，可以支持维护人员以提高变压器的诊断的表现。

2.诊断的发展和说明

许多的诊断问题一样，诊断油浸式电力变压器是一个技术性的工作。变压器可能通过外部显示器的观察能够很好的工作，而一些早期的恶化，可能会在内部出现一些致命的问题。据日本的经验，近八成的故障的结果，都是由于初期恶化导致的。因此，故障应查明并避免在尽可能出现在早的阶段，一些预测维修技术。溶解油分析是其中一个关于此问题的最热门的技术。故障变压器中的气体，一般是由石油和其他绝缘材料的分解所产生的，例如，纤维素和纸张。理论上讲，如果变压器中出现故障，那么个别的溶解气体的浓度，产气率，总可燃气体量和纤维素降解率，都是显着增加。分析变压器的绝缘油中的溶解气体，它成为可行的判断早期故障类型的方法。这项研究是用以下代表可燃气体：氢（H2），甲烷（CH），乙烷（乙烷），乙烯（乙烯），乙炔（乙炔）and一氧化碳（CO ）。许多以DGA为基础的用以诊断初期恶化的方法已经被发表出来。即使在正常的变压器的条件下，这些气体中的一些也可能会在内部形成。因此，这要规范地建立一个足够大的取样浓度，以评估统计。TPC从电力变压器调查的气体数据用以建立其自己的准则。发达国家在这方面的知识库文件，是部分根据这些数据建成的。在另一方面，Dornerburg制定了一个方法对浓度的气体分析来判断不同的故障，例如甲烷/小时，甲烷/乙烯，与溶解的化合物等。罗杰斯母马全面比率代码以理论热力学评估的观点解释热量缺点类型。因为它消除了油容量的作用并且简化了单位，选择这个气体比率法是可行的。而且，它以表格形式系统地对诊断技术进行了分类。表1显示比例的方法，是由罗杰斯提出的。被溶化的气体会因为不同的本质而有所变化。通过分析能量密度的故障，很可能区分三个基本故障的过程：过热（裂解），电晕（局部放电）和电弧放电。电晕和电弧引起的电气故障，而过热是一个热故障。这两种类型的故障导致我的恶化，而损害过热通常小于从电应力。事实上，不同的气体的变化趋势，导致不同的故障类型，而通过判断关键气体变化趋势而确定故障类型的方法是确定的。例如大量的乙炔和氢气产生轻微的电弧故障。大量的乙炔和乙烯产生一电弧故障的症状。

3 、故障诊断专家系统

本研究的目的是发展一种类似人类专家以规则为基础的专家系统来进行变压器的诊断。系统处理详情如下所述。3.1诊断方法诊断是一个需要经验的任务。从只有极少数的调查中确定一种方法，是不明智的。因此，本研究采用合成方法的专业知识与程序，以协助热门气比的方法和完美的实际表现。

3.1.1经验丰富的诊断程序

变压器的例行维护程序在列表中。这个程序的核心是基于DGA的技术。气体比值法是重要的知识来源。气体比值法存在一些限制。比例表是无法涵盖所有可能的个案。最低含量的气体必须含有。介入CO的和CO2的坚实绝缘材料分别被处理，并且气体比率代码主要从一台的变压器开发。其他诊断的专业知识应该用来协助这个方法。准则、综合性专门技术方法和数据库纪录被合并完成这些局限第一步这个诊断过程通过请求DGA开始为一油样进行测试。必须为进一步推断知道更重要的相关资讯，如电压水平，防腐剂类型，上线变抽头（有

载分接开关）状态，经营的期间和去除毒气的时间的更加重要的相关信息。Noms (标准)由TPC典型数据设定然后用于判断变压器的情况的电源变压器的气体。由于不正常的情况，气体比率法用于诊断变压器故障类型。如果不同或未知的诊断结果从这些比率法被找到，将会采取一个进一步综合性专门技术。在这些程序以后，不同的严重程度分别对应适当的维护建议。

3.1.2综合的专业知识

比例的趋势，规范的门槛，关键气体分析和一些专业知识，被视为不同的证据，以确认一些特殊的故障类型。在其他的话，更重要的证据已收集了对一些特殊故障类型，更好地评估变压器的地位得到了。换句话说，更加重大的证据为一些特别缺点类型收集了，对变压器状态的更好的评估获得。比例的趋势，可以被看作是改造传统的气体的比例和关键气体的方法。比例的CH ，以乙烷是有关区分故障。这些气体的趋势可以预测故障类型。根据变压器历史数据库的其他专门技术也用于分析案件变压器的特征。第3，4部分提供这些规则的细节。

3.2专家系统的结构

诊断专家系统组成的四个组成部分，工作记忆，一个知识库，推理机和人机界面。工作记忆（全球数据基地）包含当前有关的数据，以解决目前的问题。在这项研究中，大部分的诊断变量存储在数据基础目前的气体浓度，有些是从用户，其他人是来自变压器的历史资料库。请注意，模糊集的概念，1S的纳入，以创造模糊变量上的要求，系统推理。一个知识库，是收集域的专门知识。它包含的事实和知识的关系，利用这些事实，并以此为基础的决策。生产使用的规则在这个系统中是表示，如果-然后形式。一个成功的专家系统依赖于高品质的知识库。这个变压器故障诊断系统，知识库中包含了一些受欢迎的解释性方法碇泊区，合成专业知识，方法和启发式维修规则。

4 .专家系统的实施

一个专家系统的开发是根据整个系统的提出的解释的规则和诊断过程。为了证明专家系统在诊断上这一举措的可行性，被MTL系统支持的气体数据已经被测试过。在台湾，MTL系统实施DGA方法，并将结果向所有有关的电力变压器代理报告。作为返回结果，这些代理被要求，定期收集和供应他们的变压器油样本。分析后的石油样本，收集的超过十年的有价值的气体纪录，并分为三个电压等级，69kv，161kv和345kv 。因此，一个变压器的气体的纪录是由几组数据组成的。在DGA这一过程中，所有这些数据可能都会被考虑到，但只有最近的数据有重大影响的诊断详列于后。在MTL系统中，所有气体的浓度是在体积浓度所表示的百万分之一。100 ppm平等于0.01毫升（气）/ 100毫升（石油）。从专业方面讲，只有通过检查变压器的正常水平才可使系统的正常状态得以证实。在实践中，大部分的变压器油样品是完全正常的，这可以从早期对这一专家系统的执行中成功地推断出来。然而，问题是成功的专家系统主要是依赖于诊断变压器的能力。在运行中，许多不正常状况下的气体纪录被选择去测试判断这一诊断系统。共有101个变压器纪录已经被处理，结果归纳在表5。在这些实施，三个被列出和证明了。表5所示是测试结果，101个单位的变压

器的三种补救措施类型：正常，热故障和电弧故障。经过实际状况和专家的判断比较，摘要结果获得。如前所述，一个单位的变压器可能包括许多组的气体数据。在评价中，因为有些变压器在不同的运作阶段可能有不同的初始故障，所以我们描述故障是通过一组关键气体的含量。一些故障的测试是由于在石油采样容器中的剩余油所造成的，不稳定的气体的特点，新脱气样品和一些模糊的气体类型。如果想了解更多信息或新技术的支持及其他不认识的函数，他们可以加入知识bas 扩大这个原型专家系统表现。此外，该说明的参数表2，第3和第4是适合的TPC电力变压器。如果维修人员寻找更多合适的系统参数，有些地方可予修改。

5 .结论

原型专家系统在个人计算机被开发使用KES。它可以诊断被怀疑的变压器的缺点和建议适当的维护行动。模糊集的概念是用来处理不明朗，规范的阈值，气体的比例选区分界及关键气体分析。综合性方法的诊断过程用以协助由气体比值方法不能得到妥善处理的情况。从专家系统的实施表现看，专家系统是协助人类专家和维护工程师的有用工具。这个专家系统知识库被纳入DGA、综合性专门技术和启发式维护规则之内。TPC MTL所支持的约十年收集的变压器的检测数据资料库，也是用来改善诊断。通过发展所提出的专家系统，可以继续保留包含专业知识库的TPC MFT。此外，这一工作可以加入任何新的经验，测量和分析技术用以继续扩大知识库。