Various design considerations were taken into account when devising a condition monitoring, analysis and diagnosis system for pumped storage power units. Qingyu Wang, Fulei Chu and Fuzhou Feng report
Unlike the gas turbine, the water turbine has a much lower rotating speed but can be influenced by a variety of operational factors, making it difficult to detect malfunctions. Conditions for a pumped storage unit, which has to be switched from generating to pumping mode or vice versa according to requirements, are even more complicated. To ensure the safe and effective operation of such a unit, a system was designed to handle condition monitoring, along with analytical and diagnostic capabilities.
Research in the field of condition monitoring, analysis and fault diagnosis in large pumped storage units has been in progress for more than a decade. Systems have been developed by various companies but most of them can only perform monitoring and alarm functions — only notifying the operator that a fault has occurred. With this in mind, the authors decided to try and design a system that could also diagnose and analyse problems. Such a system was designed for and installed at the Guangzhou pumped storage power station in southern China.
The primary parameters of the 300MW Guangzhou pumped storage unit are shown in the table below.
Basic design principles
Based on the analysis of operating characteristics and common malfunctions in the pumped storage unit, appropriate sensors and measuring points were selected, analytical methods chosen, and software modules devised so that a practical system could be put into operation. Such a system needed to be reliable, open and flexible. The customer’s requirements were taken into account as much as possible with the aim of designing a simple and user friendly system.
The data acquisition unit was based on industrial personal computers (IPC), and the whole system as upper versus lower computers with network communication between them. There are many advantages to IPC technology, such as abundant sources, low prices, high reliability, openness and good compatibility. Due to the advantages of IPC it will be easier for the system to be upgraded in the future or combined with a SCADA system, which has been established in the power station before. With network communications, the distance between the upper and lower computer can be as long as one kilometre, which also makes it more convenient to operate and manage.
The selection and placement of transducers had to be considered carefully, according to the location and the type of faults in the pumped storage unit. For instance, eddy current transducers can be chosen for measuring the shaft vibration, rotating speed and keyphasor; velocity transducers for measuring support and structure; hydro pressure pulsation transducers for measuring hydraulic pressure variation; and temperature transducers for measuring pad temperature and oil temperature.
Design of the software
The system had to be able to acquire and store information on the condition of the shaft runoff, vibration, temperature and pressure pulsation of key parts of the unit; and to gain detailed information about the unit with regards to monitoring, analysing and diagnosing. Therefore it will be capable of facilitating reliable methods for regular maintenance, component repair and predictive maintenance according to the condition of the unit.
The lower computer carries out data acquisition while the upper computer is responsible for monitoring, analysing and diagnosing. As the design of the software is mainly focused on the upper computer several modules to perform these functions were established.
Condition monitoring is a basic component of the system which should reflect the condition of the whole unit or parts of the unit by on-line monitoring of different parameters. It is paramount that this module illustrates the condition of the unit clearly. First of all the authors designed for global monitoring, which allows users to read current measured values of all points on the unit, giving a complete picture of the condition of the whole unit.
To meet various user needs, the graphical interface was designed in two different ways, namely, as a graphical representation to monitor direct parameters or as a tabular representation to monitor variables of global data.
Sometimes, however, users may find that some of the signal parameter values of the measuring points need greater attention, such as when an abnormal situation occurs. With this in mind, partial monitoring was devised where the whole unit is divided into seven parts for further monitoring, each of which includes several corresponding measured points. Trend monitoring was also incorporated as an option to allow users to select one or several parameters to form a trend chart.
The analysis module is a key component of the system. Whether or not the system can perform expected functions, or whether or not the system can be useful, depends largely on this module. With its help, users should at least be able to deal with such common faults as illustrated in the following table Vibrations in pumped storage units are mainly caused by mechanical factors, electromagnetic factors and hydraulic excitation. Other sources include excitation between different sets and excitation of auxiliary machines. The signals, which reflect the forced vibration owing to electromagnetism, rotor, stream current and many other sources, may be periodic signals or random signals or both. To analyse and diagnose these signals time-domain features, such as the average value of amplitude and the maximum value of peak to peak, need to be used and can be usually achieved through time-domain waveform analysis methods.
On the other hand, frequency-domain features may also be necessary to estimate the condition of the unit, such as the values of a certain characteristic frequency point and the energy distribution of a frequency bandwidth.
There are also other methods of analysing malfunctions in rotating machinery, such as orbit analysis, phase analysis, trend analysis etc. So the analysis module was designed to include both the common analysis methods used in signal processing, and some means and methods more widely used in hydro turbine analysis as well. Users can select them according to their practical needs.
To systematically evaluate the condition of the unit, so as to analyse the monitored signal and find the causes of potential faults in the unit, the signal analysis module has been divided into several parts.
For each part, several preferred methods have been selected for analysis and diagnosis based on its features. Users can decide whether to add or delete the methods related to one part. Moreover, new signal analysis methods beyond those which have been provided can be added to this module, according to user needs. Such configuration demonstrated the openness of the system.
A diagnosis module is needed when experts are absent. The basic idea to establish diagnostic models comes from common analysis methods such as fault tree, failure mode and effect analysis, cause-effect network, physical process description, structure and behavioural function models and so on.
Common faults, other than function failure, usually occur in a large hydraulic runoff. So the methods mentioned above based on analysis of function failure are not effective enough. Therefore another method was developed as the core of the diagnostic model.
The following characteristics were incorporated into the diagnosis module:
• Expression of knowledge. It should be direct, simple, easy to understand, which makes it expedient to build, edit and modify the knowledge base.
• Diversified reasoning mechanism. For instance, besides the logic reasoning based on production rules, there should still be other methods such as reasoning of fuzzy clustering algorithms based on pattern recognition theory, so as to quickly find the set of the main possible malfunctions.
• An open, user friendly knowledge base.
• A variety of diagnosis methods.
• An explanation of diagnosis results would help the users to understand the process of diagnosis.
The system has been in service for over 30 months in Guangzhou pumped storage power station. Although tremendous data has been collected from the system, clear faults still cannot be found and only minor problems have been identified. Solutions have been suggested to the maintenance team. Many aspects still need improvements, such as:
• Hydraulic excitation should be investigated more thoroughly.
• The essential characteristic parameters should be separated from methods such as online pattern recognition, so as to obtain stable curves of the trend variation.
• Other charts or graphics more suitable for field maintenance and repair works, such as relative efficiency curve charts and vibration energy distribution charts, should be used.
The system has been established based on the analysis of features of the usual malfunctions and operating characteristics of the pumped storage unit.
Further analysis of the data will give greater understanding of the mechanisms of the unit. A new condition monitoring, analysis and diagnosis system, in which these improvements have been included, will be installed in Shisanling pumped storage power station later in 2001.