Distributed Infrastructures and Diagnostic Algorithms for Condition-Based Maintenance of Electrical Power Systems

In the field of electrical power engineering, condition-based maintenance of components of electrical power systems is increasingly being seen as a very important approach for ensuring production stability in the energy market.

ALSTOM Power Switzerland offers services for condition-based maintenance of and diagnostic decision support for insulation systems of high-voltage electrical rotating machines (like turbo and hydro generators) to operators of such machines all over the world. In collaboration with the Insulation Competence Center (ALSTOM Power Birr), we have envisioned and developed two scenarios for improving the quantity and the quality of the offered diagnostic and condition-based maintenance services.

Subproject 1 : An Internet-Based Infrastructure for the Evaluation of Partial Discharge Patterns for Condition-Based Maintenance

The goal of the first subproject is to collect and evaluate online sensor data via the Internet to enable condition-based maintenance of the insulation systems of power generators. ALSTOM has developed sensor boxes that periodically send electrical signals to the machines they are attached to and capture the replying signals (partial discharge patterns). The incorporated algorithm evaluates the data to detect changes with respect to the expected, normal pattern. If such changes are noticed, the data are sent to a central server for further evaluation by a human expert. The expert uses an elaborated analysis framework (implemented as a client-server application) to make diagnostic evaluations of the patterns, supported by previous historical evaluations stored on the server. The current results of this subproject are:

• a distributed Internet-based infrastructure for the evaluation of partial discharge (PD) patterns for condition-based maintenance of the insulation systems of electrical power generators
• the ROI (Region of Interests) algorithm, a domain specific algorithm for detecting changes in PD patterns
• an advanced analysis framework that supports diagnostic evaluations of PD patterns (see screenshot below)

Subproject 2 : A Distributed Experience Management System
for Diagnostic Decision Support

The goal of the second subproject is to assist diagnostic engineers in performing off-line preventive maintenance operations. Typically, engineers need to raise and check many hypotheses to arrive at a successful diagnostic decision. The necessary knowledge to solve such a diagnostic problem is usually acquired by engineers only in the course of several years of practical experience. Our contribution in offering engineers decision support is a distributed system that accesses many sources of information, to find out whether a similar problem has been solved and documented previously by other engineers. Currently, the results of this subproject are:

• methods for knowledge extraction by applying data and text mining techniques
• development of experience management principles in collecting and reusing valuable experiences for diagnostic problem solving
• an interactive web-based system that supports the diagnostic decision making process of human experts (see screenshot below)

Funding: ALSTOM Power, Birr , Switzerland (since 2000)

Contact: Eni Mustafaraj