The hydro digest14 December 2000
The Digest analysis and diagnostics system was originally developed for use in thermal power plants but was modified for hydro power plants to meet special technological requirements of hydroelectric operations. Hermann Scheil reports
THE Digest analysis and diagnostics system for turbine plants was originally developed for use in thermal power plants. However, the many possibilities offered by the system soon led to expansion to other areas of application.
The Digest WK branch was created for hydro power in order to meet the special technological requirements of hydroelectric processes, such as weight and large physical dimensions, and both hydraulic and electrical excitation. Digest WK has all of the features and advantages of Digest which have been successfully implemented in steam turbine plants for many years.
In Itaipu Binacional, currently the largest hydro power plant in the world, 18 turbine-generators are generating electrical energy from the power of the Parana river on the borders of Brazil and Paraguay. Siemens supplied the generators for nine of the 18 turbine-generators, for which a monitoring system is currently being implemented with Digest WK in co-operation with voith-siemens. This modular monitoring system is collectively called Mondig (monitoring, diagnosis) and performs a variety of tasks.
Essentially Mondig acts as an early warning system and represents a significant challenge for the Digest WK system. The slightest changes from normal operating conditions are detected and annunciated with process data acquisition as a function of current operating conditions. The status of monitored components can be diagnosed directly. This enables implementation of countermeasures in the event of disturbances so as to prevent any further damage. Transition to less demanding or reduced load operation enables continued electric power generation with a turbine-generator, possibly even until reduced reservoir levels make shutdown essential, bringing the turbine-generator to a standstill and making it redundant in terms of power generation.
Preventive inspections can thus be replaced by condition-oriented, targeted repair and maintenance measures. With such a system, power plant operators anticipate a significant long term reduction in repair and maintenance expenditures.
Unlike conventional monitoring, in which operating personnel are often overloaded by a large number of messages and extensive data, Digest WK compresses the relevant data in such a way as to enable relatively simple analysis and diagnosis.
The system achieves significantly higher information density with a much wider scope of data. Deviations from setpoint conditions are not only detected but will also be evaluated in a later extension of the system. Knowledge bases are used to generate proposals for targeted countermeasures.
The process can be optimised for the monitored components. This is achieved by constant comparison of the current data with reference values which were determined either with lear- ning procedures or through algorithms.
Annunciations are generated automatically if the allowable limits of the data scatterband are violated. The limits can be specified to lie well below any danger levels.
The operator analyses problems using the system’s graphical user interface. He makes use of a wide range of displays such as trend curves and correlations of various data signals, polar plots, bar charts, tables and plant-specific display material.
Although the Digest WK system includes automatic or manual initiation of knowledge-based diagnostics and displays resulting diagnoses which further characterise plant conditions in the event of disturbances, these are not currently implemented in this project.
Conditioned data are stored as a function of type in the analysis and diagnosis computer (workstation) in scatterband, short term and long term storage. The data can be called up as a function of specific disturbances. Where expedient, it is also possible to make on-line notes in various displays, which can then be stored or printed out with the display.
The use of telephone modems enables remote data transfer. More detailed disturbance analyses can thus be performed on line from a remote site.
Comparison with historical data, even if this has been transferred to external data carriers, is possible at any time. The prerequisite for this is that the data have been recorded with the Digest WK system.
The Digest WK system
The system is essentially structured in three hierarchical levels. In level 1 the data are acquired, in level 2 they are conditioned and preprocessed, and in level 3 they are processed, ie linked, stored and visualised.
The systems in level 2 are relatively simple and feature only a slight storage depth or none at all. At this level the data are digitised and checked for plausibility, and if necessary parameters and spectra are derived from the data.
Level 3 makes use of high performance computers (workstations) with high storage capacity. At this level, extensive and complex monitoring programs perform module-specific decomposition of the data, which are then linked, distributed, saved and visualised.
The advantage of this structure lies in parallel transfer of all data in the shortest possible time to the analysis computer (high performance computer), where the actual process monitoring is performed.
For abnormal events, the system has event recorders which are located in the intermediate level. With up to 64 channels for each turbine-generator, the event recorders simultaneously record all vibration and air gap signals as well as representative process parameters with a sampling rate of 5kHz. The records are overwritten at intervals of up to one hour. If an event occurs, all important signals (from the period prior to onset of the event) are available for disturbance analysis.
The Mondig monitoring system in the Itaipu Binacional hydro power plant in its initial stage comprises the following three modules for each turbine-generator:
• SVIB (monitoring of absolute and relative vibrations and pressure fluctuations), implemented with Digest WK module VIB.
• AGS (monitoring of generator/stator and generator/rotor air gaps), implemented with Digest WK module AGM.
• STE (monitoring of turbine-generator temperatures), implemented with Digest WK module GTA.
The fourth module, SPD (monitoring of radio frequency partial discharges in the generator windings) will be implemented later with Digest WK module PDA.
Vibration monitoring (VIB)
The vibrations in each turbine-generator are detected with 22 sensors, and evaluated for relative and absolute amplitudes as well as phase as a function of relevant process parameters.
Compression of the vibration signals and of the relevant process data in characteristic reference variables enables extensive evaluation of vibration conditions at any time, thus establishing a basis for state-oriented maintenance.
The VIB module already enables the expert to perform on-line diagnostics based on the recorded and displayed data, the conditioned reference variables and the processing tools in Digest WK.
The subsequent extension will enable the transition from manual on-line diagnostics to computer-supported (expert system-based) diagnostics.
Air gap monitoring (AGM)
In large generators, the distance between the rotor and stator (air gap) must be monitored, as there is a risk of the rotor not moving concentrically in the stator. This can result in large forces and vibrations unevenly distributed around the circumference. Rotor contact with the laminated stator core would result in significant damage.
The air gap is monitored by 12 special sensors positioned on three measuring planes around the inner circumference of the laminated stator core.
The monitoring principle lies in cyclic (quasi-continuous) measurement of the distance to each pole of the rotor by each sensor, and comparison with reference setpoints as a function of operating conditions with appropriate consideration of influencing variables. The reference setpoints associated with the sensor data are determined by calculation of polynomials with coefficients from the fingerprint recording.
The analysis is based on the fingerprint recording performed on the ‘healthy generator’. The measurements for the fingerprint recording are performed in all representative operating conditions and all possible output ranges. This procedure allows the current air gap data at any time to be compared with reference setpoints which are valid for precisely those data.
• Generation of reference setpoints with appropriate consideration of influencing variables, using a computer model derived from fingerprint recording.
• Quasi-continuous monitoring of current actual values with reference setpoints: determination of nine representative reference parameters.
• Graphical display of measured and calculated values:
rotor profiles and polar displays, stator/rotor deformation, trends in normalised reference variables, x-y display of all measured and calculated values, bar charts, tables.
• Short, intermediate and long term storage.
• Signaling concept on violation of limits based on operating conditions.
Generator temperature analysis (GTA)
The temperature monitoring module is subdivided into two areas:
A) Monitoring of cooling circuit temperatures in stator winding.
B) Monitoring of operating temperatures in generator and bearings.
A) Cooling circuit temperatures in stator winding
There is a risk of blockage of the hollow conductors in water-cooled stator windings. Such blockages lead to high hot-spot temperatures which endanger the insulation.
Conventional monitoring methods are not adequate for timely detection of such blockages, with the result that serious damage can then occur. The Digest WK module diagnoses hot-spot temperatures based on changes in the temperature rises of the water as it leaves the individual cooling circuits.
Monitoring for unallowable temperature rises of the water in the hollow conductors is achieved by comparison of actual temperature rise for the cooling circuit water with load-dependent reference setpoints which are stored in the analysis computer. Deviation from these values is the criterion for issuing a fault alarm.
The reference setpoints are determined from reference measurements in a three-dimensional mathematical model with appropriate consideration of all current and load-dependent influencing variables. The reference measurements are recorded in specific test runs on the ‘healthy’ generator and stored in a ‘fingerprint’. The analysis computer contains a polynomial of the following form, determined for each cooling circuit on the basis of the fingerprint:
Where ai are stored as coefficients, fi as functions of the influencing variables and k as an empirically determined correction factor.
The analysis computer controls the measurements on the cooling circuit temperatures, saves the results and evaluates them with appropriate consideration of the influencing variables. In the event of an unallowable increase in cooling system temperatures, ie unallowable process changes, further increases first lead to the issue of warnings, followed by alarms.
Logs which are automatically generated, extensive graphics and plant-specific display material enable continuing analyses and prognoses. The quasi-continuous monitoring even enables optimum adaptation of generator output to the extent of the cooling disturbance and thus continued operation at reduced output without the risk of damage to components.
B) Generator operating temperatures
Monitoring of the 236 operating temperatures is similar to that described in (A), by comparison of the current data with reference setpoints.
The reference setpoints are determined from reference measurements with appropriate consideration of the load-dependent influencing variables, using coefficients which were determined in specific test runs and recorded as ‘fingerprints’ in a special storage area. Appropriate annunciations are generated automatically on violation of the user-defined limits for max, min and average or standard deviation values. Analyses of the measured and calculated temperatures can be performed in the various displays of the standard and free-form graphics.
The different monitoring systems, which have their own names for the different turbine-generator suppliers, are combined in the Itaipu power plant under the common name Mondig. Following operating experience with Mondig, the systems will be extended to diagnostics systems for which the customer can independently extend the knowledge base.
The Siemens Digest WK system can process uncertain knowledge. This was achieved by developing algorithms ranging from mathematically oriented, probabilistic models to the implementation of fuzzy logic. Evaluations are assigned to the diagnoses, indicating the degree of confidence in conclusions. All diagnoses are sorted by decreasing confidence factors and output for evaluation. A list of diagnoses which can be ruled out with a high degree of probability is also generated.