By Henk C. Smith & Annette Risberg, Rovsing Dynamics, Denmark
Alstom Hydro Power is currently constructing two new hydropower plants in China. These are the Huizhou hydropower plant in the Guangdong province with eight turbine-generator units, and the Bailianhe plant in the Hubei province, with four similar units.
Construction of the upper reservoir of the Bailianhe hydropower plant in the Hubei province, China
The two plants are so-called peak storage hydropower plants. During periods of low power demand, the machines will be used as electric motor-pumps to fill an artificial lake, but when the region’s power demand is high, the water will be released and the machines will operate in turbine-generator mode. For this, an advanced conditioning monitoring system is required. The capability of the OPENpredictor system for automated advanced condition monitoring will reduce operation risk and help the plants prioritize maintenance activities. The complex task of monitoring all the different operation conditions is totally automated to minimize risk and provide trends and forecasts for similar operating conditions.
In total 12 reversible Francis hydro turbine-generator units, each with a capacity of approximately 300 MW, will be monitored. The advanced processing capabilities of the online monitoring system will provide warnings at the central control room about machinery problems in an early stage of development. The integration of the system with ALSPA, Alstom’s control system, allows separation of information for operators and maintenance engineers.
System architecture of the OPENpredictor condition monitoring system
OPENpredictor integrates vibration with process data such as temperature, flow, generator and exciter current. These data facilitate close monitoring of all machine operation, transient and gradient states, to follow their respective dynamic behaviour and report identified changes.
The complex machine operations, with both pump and turbine operation, and more than 20 different operational/gradient states, requires a dedicated approach to efficiently monitor the critical components of these machines. In particular, for peak hydropower plants such as Bailianhe and Huizhou, classification of the data comparison to machines’ different operational states is a pre-requisite for efficient condition monitoring and early fault detection.
Special monitoring needs
Machinery in hydropower plants are usually highly reliable. However, as maintenance has a large cost impact on the total operation it is essential to only perform maintenance when it is absolutely necessary. To guarantee machine availability, avoid major breakdown and to schedule inspection and maintenance, machinery operational health needs to be monitored accurately.
Short-term and long-term risks can be evaluated with the OPENpredictor condition monitoring system, which performs online mechanical and functional health monitoring. The system is designed to provide monitoring, alarming and reporting to minimize maintenance costs and improve plant productivity. The gathered information helps to provide information to help the staff forecast machinery problems, optimize operation efficiency, schedule and cluster inspection and maintenance, and conduct troubleshooting.
Automated early machinery fault detection facilitates prioritization of maintenance actions and reduces the need for time based maintenance tasks to those actually required. Forced outages can be turned into planned outages, resulting in optimal control of the operation. To achieve maximum production uptime, it is essential to coordinate maintenance activities of all critical machines. Therefore, OPENpredictor incorporates fault-monitoring strategies for all types of critical machines, ranging from turbines and generators to multi-stage pumps and fans. Such machines consist of different critical components, with individual potential failure modes, but together describe the overall machine health.
In a hydropower plant different functions are provided by primary machinery (turbine, generator and servo motor) and secondary machinery and systems (drain, air compression, cooling, oil pump, leak oil) in order to produce power. Each machine requires dedicated monitoring techniques to assess its health, to secure availability and the efficiency of production. OPENpredictor integrates the signals and data from all sensors and subsystems, to monitor and present any required information so decisions can be made reliably. Based on this information, the plant operators and maintenance engineers can conclude where short- and long-term risks become too large, so corrective or preventative maintenance actions can be taken.
Right data at the right time
A lot of emphasis is placed on automated fault identification to minimize expert involvement because the machines in a hydropower plant tend to be highly reliable. By separating information flows, the right people will get the right information at the right time.
Monitored components at the two Chinese hydropower turbine generators will comprise the shafts, bearings, generator stator, rotor and head cover. Armed with the information about the machines’ health, the power plant operator is able to optimize the operational condition and prioritize maintenance tasks. This will increase machine component lifetime and reduce maintenance and downtime costs, justifying the investment. The system is capable of monitoring Kaplan, Bulb, Francis and Pelton type turbines.
At a hydropower plant in Peru, OPENpredictor provides online condition monitoring of three Pelton hydro turbines of 30 MW and offline monitoring of auxiliary machinery. The plant generates power from a river flowing down a mountain from 2130 m above sea level. To optimize the pressure conditions and power generating potential, the river water is also led through the mountain via an artificial channel with a considerable height difference.
Thanks to condition monitoring the plant engineer no longer has to dismantle the main bearings of the three power generation units for inspection once a year. Normally, the inspection takes 14 hours with a production loss of around 1260 MWh, so avoiding downtime for unnecessary inspections saves approximately $31 500 per year, plus the time and machine parts spent.
The system also identified two problems with the alignment of the turbine generator set, stator and rotor between the main bearing and the hydro turbine. Fortunately, the misalignment was discovered and corrected before it became too serious, but the units were working close to the vibration alert threshold, which minimizes the risk of rubbing and fatigue development.
Because OPENpredictor helped discover the misalignment problem in time, the staff could plan to address the problem during an upcoming scheduled maintenance. If they had not had condition monitoring, the damage to the main bearing or, in the other case, the number of inspections would have had to have been increased from one to probably two a year. If the misalignment problem had caused damage to the main bearing, most probably the rotor and the stator would have been damaged too. In that case the units would be out of the operation for a minimum of six to 12 months.
Ring the alarm
Changes in the mechanical health and machine function are automatically reported to the relevant people. OPENpredictor operates with a straight- forward utilization of the alarm system, graphical mimics or a machine browser. Alarms are grouped for short-term operator action, long-term maintenance action and identified machine performance reduction. Automated shift and management reporting avoids tedious data retrieval, reporting and printing.
The graphical mimic shows real-time operation data to the operators. In case a parameter reaches an alert level, the colour of the value field will change. By clicking the field, trends and forecasts will be visualized. The diagnostic data are then readily available for root cause analysis by experts.
Information on machinery mechanical health is gathered through signals from dedicated vibration sensors, such as accelerometers and displacement probes giving signals from casing, and shaft vibration, shaft and rotor position and tacho probes providing phase information to identify changes in vector data. OPENpredictor is unique with respect to the way signals picked up by the sensors are analyzed, saved and interpreted. Vast amounts of sensor data are channelled through a signature processing unit (SPU).
To warn the organization about potential risks, individual machinery failure modes are constantly monitored. Early fault identification is achieved with so-called “Fault Selective Signature” monitoring, providing component health assessment by comparing potential fault symptom levels. The health of a machine consists of the development state of all potential failure modes. As a unique function, OPENpredictor employs a suite of different fault selective signatures to cover the widest range of potential problems.
To further increase fault identification sensitivity the signature comparison is classified to operational conditions minimizing symptom level variations due to the process. This means that signature comparison is executed under comparable process conditions. This proactive monitoring strategy consequentially minimizes false warnings and improves forecast reliability of the fault symptom development.
Automatic fault diagnosis
The automatic fault diagnosis is another unique feature of the OPENpredictor system. When component behaviour starts changing, the patented AutoDiagnosis automatically interprets the changes into AutoDiagnosis messages. In addition to early and precise fault diagnosis, it also provides the maintenance staff with a prediction of lead-time to inspection. This function integrates the knowledge of experts into the software, applying the knowledge day-in day-out. The role of experts is thereby changing from trouble shooting in a late stage of problem identification to pro-active verification of potential faults in order to plan and prioritize activities.
The OPENpredictor condition monitoring module offers a suite of analysis tools to verify the identified fault conclusion and provide additional information about developing faults. Examples are bearing wear, shaft instability and misalignment, rotor imbalance, foundation looseness and cavitation.
For hydropower plants air gap monitoring of the rotor-stator gap in the generator is important. OPENpredictor’s signature processing unit can accept signals from dedicated air gap sensors and identify the eccentricity and deformation of both rotor and stator.
Hydro generators have an air gap of only a few centimetres between the rotor and stator, which usually have a rotor with diameter of several metres. This complex design requires continuous online air gap monitoring in order to detect developing faults like eccentricity or deformation of stator or rotor before crucial damages occur due to rubbing.
Partial discharge monitoring is another important condition monitoring function which helps hydro power plants schedule generator inspections. This functionality monitors and reports sparks that occur as a result of deteriorated insulation on the electrical parts inside the generator. The high frequency discharge signals are being converted by a simple converter and partial discharge trends are being monitored in OPENpredictor, using the operational state concept. The discharge forecast is then used to schedule an advanced analysis of the discharge distribution as function of the load cycle.
As all condition monitoring takes place in the system’s intelligent front-end monitors, only one simple server takes care of the administrative tasks for all turbines in a plant. This one-server approach simplifies maintenance of the system, reducing the life cycle costs of the system itself.
Due to the concept of integrating knowledge into the local condition monitoring system at the plant, remote monitoring from service centres is possible even with low bandwidth networks. This allows for a very small group of experts to efficiently monitor a large number of machines and support an operational team.
OPENpredictor has previously been installed to monitor turbine generator sets at hydro power plants in Mexico, Peru and Ukraine in connection with upgrades carried out by Alstom Hydro Power.
The system is highly adaptable and scalable, and is also used in other power sectors for monitoring combined-cycle, thermal and nuclear power plants.