INDEX TEST ANALYSIS OF KAPLAN GENERATING UNIT
Index Test at 69 Ft
ACTUATION TEST EQUIPMENT COMPANY
January 14, 2016
Performance analysis of hydroelectric generating units can be done using a new Hybrid Index Testing Method that does not require the test personnel to go to the dam. A large part of the cost of index testing is the logistics of getting an expert test crew (aka the expensive help) and their equipment on site for the index. Taking advantage of modern powerplant SCADA systems, most of the necessary instrumentation and a data logger are already in place.
This is a report of our recent experience index testing a 5MW Kaplan in this way.
The already extant powerplant data recorder was reprogrammed to add a few signal channels and increase the sample rate to 2 Hz so that suitable data can be continuously recorded en mass and then provided to Actuation Test Equipment Co, Inc., (ATECo) for preliminary PostProcessing using the SteadyState algorithm in the Index Test Box (ITB).
Upon input to the ITB, the software program first sorts and converts the raw data files to a compatible format for analysis. Affinity Law head corrections are applied to the data in its rawest form, and then average values and two accompanying “figure of merit” values are computed as described in the next section.
These results are then analyzed manually utilizing a test code specified methodology to produce the overall efficiency profile and a new head-and-gate-to-blade cam profile to load into the unit’s 3-D Cam.
Figure 1 Overall Unit Efficiency Profile
Description of the Index Test Box
The Index Test Box is a standard IBM® PC Clone computer running any current version of Microsoft Windows®. The ITB software is a Visual Basic 6 program compiled to an executable-program for distribution. Distribution, installation, price and operation are similar to a typical computer utility or video game program.
Using the ITB, data can be acquired several ways:
1. Discreet sensors can be used in powerplants without a SCADA system using National Instruments® Analog/Digital Input/output boards and external I/O modules.
2. Data can be acquired digitally from another instrument using any standard machine to machine communication interface (OPC, IEE-488, RS-232 etc.)
3. Data can be digitized and recorded by another test system, such as a typical powerplant datalogger to record streamed data to files that are transported via CDs, disks, flash drives or over the Internet.
ITB Functions currently available include:
1. Stripchart-style recording of live-field data for later analysis.
2. SteadyState data analysis and collection of live field data.
3. Continuous Efficiency Monitoring of machine health.
4. Digitization of Cartesian coordinate charts of cam profile data
5. The ITB was especially designed for index testing Kaplan turbines, but can also be used to index test Francis, Pelton or any other type of turbines.
Description of Data Reduction Process
Data reduction takes place in several steps:
1. The input “raw-data” is the continuous strip-chart file from the powerplant data logger.
2. The reduction process is facilitated by rearranging the data columns of the raw-data file to conform to the ITB input format and then truncated to cover just the 4-hour period during which the index-test data was collected.
3. The data is normalized in its “rawest” form as it is read into the ITB for SteadyState analysis by applying the code specified affinity laws to correct for any head variations.
4. SteadyState points are gleaned from the continuous raw-data file using a sliding 100-sample window to bracket a block of data for statistical analysis.
5. This analysis consists of least-squares linear regression computation of the 100-sample block that draws a line through the “center of mass” of the 100-sample data block.
6. The center point of this line is the “Average Value” for the data point and the slope of the line indicates its “SteadyStateness.”
7. Next a Standard Deviation is computed between each of the individual data points and the line to quantify data coherence.
8. To be deemed SteadyState, the slope of the line must be near-zero and the standard deviation must be small.
9. An average value is deemed “steady-state” when both Slope and Standard Deviation values are within operator pre-set limits.
Results of the SteadyState analysis are output in three data file formats:
1. "AllData.csv" contains every analyzed data point.
2. “AllSSData.csv” contains only the “SteadyState” data points that are deemed to be steady-state.
3. The third output is a set of files that are named for the head, gate, blade, flow power and efficiency values of their contents. For example, the file named “H683,_G953,_B686,_F8056,_P352,_E756,_SSData.csv” contains all data points at 68.3 ft head, 953 gate, 68.6 blade, 805.6 cfs flow, 3.52 MW and a generating efficiency of 75.6%.
For ease of evaluating these results, data within all output files is arranged in chronological order.
The AllData.csv file is then analyzed further using the industry accepted “smooth-curve” method to develop the new 3-D cam profile line for 69 feet head.
The AllData.csv file is a chronological list of SteadyState-filtered ITB output data points with line breaks wherever a non-SteadyState point is encountered. Averaged values for each column are the data set to go to the next step in the analysis process.
The number of consecutive identical test points is a figure-of-merit for the set of identical data point - reasoning that if the same value comes out many times in succession then the data point is steady-state and the average values are good and stable.
The average value is at the bottom of the head, gate, blade, flow, power and efficiency columns in each data block.
In the next step the ITB program consolidates these averaged values chronologically in the "AllDataNew.csv" file into the "Results.csv" file to create a chronological listing of all of the averaged SteadyState points.
The final result of the ITB SteadyState process is the "Results.xlsx" file.
The data was sorted again to remove any SteadyState points of less than 10 consecutive SteadyState measurements, leaving only the longer, more steady-state dwells.
Next the steady-state data points are sorted by blade angle, and then parsed at each blade change.
Each block with equal blade angles is next sorted by and gate stroke.
Average values are then computed for each equal blade and gate block as the final output of the SteadyState analysis routine.
A Cartesian coordinate graph of this data set for the 2nd index test at 48 feet gross head shows this result (Figure 1).
Figure 2 Index Test Box Output results for Dorena test data
CLASSICAL SMOOTH-CURVE METHOD OF DEVELOPING INDEX TEST BEST-CAM PROFILES
Figure 3 Dorena Index Test Analysis Spreadsheet (Click image for spreadsheet download)
Figure 2 shows the layout of the data in the final spreadsheet.
In a classical “fixed-blades, swept gates” index test, the blades are held at a series of 5 or more fixed pitch angles while the wicket gates are positioned at 5 or more openings and then efficiency is measured by a code accepted method.
In 1984 DoE BPA’s Lee Sheldon provided a copy of USACE’s tutorial on Kaplan index testing as a recipe of guiding principles, methods and techniques to guide Woodward’s Index Test Box software development effort.
This index testing manual is detailed, step-by-step instruction set that will allow someone unfamiliar with Kaplan index testing to quickly get up to speed.
Lee tells of the time when he was given several large boxes full of these manuals as if they were not needed anymore; Lee grabbed a few and the rest of them got lost.
Figure 3 shows an overview (the Big Picture) of the raw data from the powerplant data logger with blade and gate positioning and computed gross unit efficiency.
Figure 4 Big Picture of index test points
At each blade/gate pair test point, a 3-5 minute dwell is waited out to allow everything to settle so that steady-state data points are assured.
Steady-state data is assured because the analysis technique does not rely on “inefficient carbon units” to establish and maintain steady-state operation for a known time interval while supposedly steady-state data points can be captured and recorded.
Instead, the unblinking precision of the computer scans a continuous data stream for steady-state operation and then captures and records a steady-state data point whenever the unit happens to be running steady-state.
Added confidence in the “steady-stateness” of the data is gained when a long, contiguous series of data points is captured that have exactly the same values.
This stream of readings is recorded for subsequent off-line ITB PostProcessor SteadyState analysis that consists of a linear regression by least squares fit and standard deviation for each blade/gate pair. These form a data set for every individual fixed blade angle as shown by the blue highlighted tables in Figure 2.
The next step is to plot power and flow versus gate opening separately as two "smooth curve" graphs, which are shown directly under each blue data set as shown in Figure 2.
These graphs are used to check for any random or precision errors in the data sets.
All of the data was very precise with no randomness evident.
Next the data was recombined and the performance values computed for every 0.5% of gate opening.
This data is identified with pink highlight on Figure 2 and is plotted on the graph entitled, "Dorena 1 - Index Test Data Curves" located at the far right of the spreadsheet in Figure 2.
Figure 5 Final Graph of Classical Smooth Curve Analysis Method
The single data table highlighted in purple at the far left of the spreadsheet of Figure 2 is the index test data input from the ITB software.
As a double-check for this first index test, this data table was compared to hand calculated results by plotting it right on top of the two 0-Degrees (flat-blade angle) smooth curves to verify and demonstrate the coincidence of the ITB software result with the industry accepted method’s result.
Observations and Conclusions
· On the upper portion of the Dorena 1 - Index Test Data Curves graph a tangent line is drawn connecting the tangents of the fixed blade profiles.
· This graph shows the overall efficiency performance of this machine at a gross head (forebay - tailwater) of 69 feet.
· This means the intake losses, the hydraulic losses in the penstock (friction and turbulence from sharp direction and diameter changes) and the sudden expansion losses of the fluid exiting the draft tube are all charged to the turbine.
· This unit has a maximum efficiency of 76.7% at 2.9 MW where it is using 645 cfs.
· The accuracy of an index test data reduction can be judged by the lack of randomness in the line of interpolation points on the lower portion of this main graph.
· All of the data points fall on the smooth curves which exemplifies the extreme accuracy with which this index test data was reduced.
· The tangent points of blade and gate are plotted onto the lower graph and a line drawn through them to form the "blade to gate cam curve."
· Again, this data shows no randomness whatsoever.
· It does show the lift off point for the blades at this gross head is 52.5% gate opening, which is a significantly large gate opening, and that full steep blade angle is not achieved until 97.25%.
· These two characteristics, along with the low maximum efficiency (even when including the penstock losses) indicate that design head or the head of best efficiency will be noticeably higher.
End of file
 ASME PTC-18 Pg. 64, Section 5, pp. 5.2.1
 An “overall efficiency profile” is useful for higher level optimization schemes, such as WaterView, Opt-Ease or USACE’s GDACS program.