Our Insights

Seattle City Light’s Denny Substation – Bench Testing the Master-Follower Transformer Scheme

September 11, 2019

The Denny Substation is a gas-insulated station in Seattle, WA, owned by Seattle City Light. This station consists of two 115 kV flat buses and three 75 MVA transformers feeding twelve 13.8 kV buses connected in a ring. The low side of each transformer is bifurcated, connecting each transformer to two different 13.8 kV buses. City Light operates the station’s 13.8 kV distribution either as a flat bus or in a ring during normal operation based on transformer availability and distribution loading.

Bench testing allowed for all possible logic scenarios to be tested. While we did not have all of the planned equipment available during the bench test, we were able to incorporate hardware and software simulators for IEC 61850 protocols which allowed for very accurate modeling of the substation logic. These bench tests helped us avoid potential pitfalls during field testing, which saved us time troubleshooting and fixing any incorrect logic.

Initial Design

City Light desired to have a Master-Follower solution for this station. The load tap changer (LTC) controllers specified in each transformer had pre-programmed Master-Follower logic that used IEC 61850 GOOSE messaging. A second unrelated automation scheme at this station using IEC 61850 was also being developed by POWER. The initial plan included two options to move forward with the design:

  1. Create custom logic to operate each transformer’s LTC when operating in parallel, or,
  2. Use Vendor A’s Master-Follower scheme included with the specified LTC controller.

City Light and POWER ultimately decided to use the LTC controller’s internal scheme logic. Creating a custom solution would have taken more time and money, and the LTC controller’s vendor had logic pre-programmed that we believed would operate appropriately. Each transformer was designed with an LTC controller to regulate the 13.8 kV buses. This scheme was based on paralleling transformers connected to a flat bus and relied on receiving hardwired inputs for the tie breaker status (left and/or right) and line breaker status.

Using this scheme at the station was not straightforward due to the 13.8 kV ring bus configuration and the bifurcated transformer secondary connections. A simplified one-line of the distribution bus is shown below in Figure 1. A typical application of a Master-Follower scheme would have a line breaker status, right-tie breaker status, and left-tie breaker status hardwired to the LTC controller.  However, more than one breaker could potentially comprise these statuses; therefore, hardwiring was not an option.

To address this issue, we introduced a programmable logic controller (PLC) in each transformer’s cabinet. Each PLC would monitor the system’s status and provide the appropriate hardwired inputs to its respective LTC controller. GOOSE messaging provided the information to the PLCs. Each breaker’s protective relay published a GOOSE message containing the 52 ‘a’ and truck status. Each PLC subscribed to these messages for the required data. This structure allowed quick communications and simple integration of an additional transformer at a later date. The future transformer’s PLC need only subscribe to the GOOSE messages on the station’s existing ethernet network. Using this information, each PLC would determine if one transformer is electrically connected to another transformer and provide the appropriate signals to the LTC controller’s input.

Consider the topology shown in Figure 1. There are 220 (1,048,576) different combinations of tie- and low-side transformer breakers states. Even with some of the breakers logically grouped for bench testing, we would have had to test 216 (65,536) different combinations. With this number of combinations to be tested, hardcoding each scenario into the PLC’s program was not practical.

Instead, we implemented concepts from graph theory, the study of mathematical structures made of vertices connected by edges. The topology of the 13.8 kV system was modeled as a graph; each bus was modeled as a vertex, and each breaker was modeled as a potential edge connecting appropriate vertices. An edge existed between two vertices if its respective breaker was racked-in and closed. Finally, we used a depth-first search algorithm to determine whether two or three transformers were operating in parallel. This implementation allowed for an expandable method to determine if any bus was connected to any other bus in the 13.8 kV system.

First Bench Test

POWER drafted and executed a bench test plan to verify the PLC’s logic functioned as specified. The goal was to determine if the PLC provided the correct hardwired outputs for every breaker status scenario. A PLC and an Intelligent Electronic Device (IED) test set were used to complete this test.

The IED test contained IEC 61850 GOOSE support and was able to publish multiple customized messages simultaneously. Again, considering the number of the possible combinations of breaker states, performing this manually was not practical. Instead, we built a custom script to interact with the IED test set directly to automate the testing. The script stepped through each possible breaker combination, edited the values in each published GOOSE message, and obtained feedback from the PLC through hardwired inputs into the test set. At the same time, the script calculated the expected outcome based on the breaker statuses and compare the calculated values to the IED test set values. Each test result, either passing or failing, was recorded for later review.

We confirmed the PLC was successfully generating appropriate input data for each LTC controller upon review of the recorded test values.

During the first bench test, failures of the testing script uncovered errors with the script, not the PLC programming. We needed to add logic to mimic the electrical connections being verified. Script errors appeared in a matter of minutes once the script was running, so troubleshooting the edits did not take long. Once the errors were corrected, the script was left to run overnight since it took many hours to complete.

The first bench test successfully proved the graph topology logic worked for all the possible breaker states. The next step was to test hits logic in combination with the LTC controller’s logic on-site at the station.

Second Bench Test

On-site commissioning revealed inconsistencies in normal operation and failure modes of the built-in Master-Follower logic of the LTC controller. These inconsistencies would have resulted in mismatched LTC tap positions and possibly even LTCs being tapped to either the minimum or maximum position during a scheme lockout. POWER presented various options to City Light:

  1. Wait for the LTC controller vendor to update their logic and issue a new firmware
  2. Move the Master-Follower logic to PLCs in the transformer cabinets
  3. Replace the LTC controller with another vendor

Ultimately City Light decided to move the Master-Follower functionality to each PLC. This required a custom logic solution to be developed within the PLC. The output contacts of the PLC would now be rewired to directly raise or lower the LTC.

We drafted a new bench test for this scheme. Since the logic was now contained in the PLCs, a thorough in-office test could be completed. The setup for the second bench test consisted of three PLCs, an IED test set, and a Remote Terminal Unit (RTU) hosting a Human-Machine Interface (HMI). The HMI was necessary to test enabling/disabling the scheme and reset a lockout alarm.

The second bench test was executed in two major stages:

  1. Re-verifying the depth-first search algorithm due to minor logic modifications
  2. Verifying normal operation and failure modes of the new scheme

To verify the depth-first search algorithm, the same tools from the first bench test were used. This testing allowed us to verify the PLCs properly determined which transformers are parallel with each other for any breaker status combination.

Since all three PLCs were in one location, we tested all operating scenarios of the new logic. The PLC’s ability to determine the topology had been confirmed in the previous test, so we focused this test on verifying the PLCs ability to:

This was achieved through the different parallel combinations and manipulating GOOSE messages while monitoring the output contacts of each PLC.

The tests confirmed the failure modes by simulating various conditions such as disabling scheme logic, stopping a GOOSE message from being published, etc. Through these scenarios, each PLC was monitored to verify appropriate lockout behavior per the scheme’s functional specification.

Moving the scheme logic to PLCs introduced more areas for error that were quickly found and resolved in a lab/office environment. Adjustments for program execution order were made for scheme robustness. More significantly, a corner-case scenario bug was found that would have prevented a scheme lock-out from being reset. In both these examples, updates were made, and tests rerun efficiently to resolve and improve the end product. City Light had numerous technicians and engineers witnessing the commissioning and were confident the scheme would operate as intended.

Benefits of Bench Testing

POWER’s Advanced Technology Lab environment allowed us to test over 65,000 possible operating scenarios for Denny Substation to ensure the logic would function as expected. On-site testing spot-checked normal operation scenarios and failure modes to verify the bench test results. The on-site testing was scheduled to take place over five days. Actual commissioning was completed in only two days, including all time for coordination with dispatch and switching. The reduced commissioning time could be attributed to the bench test exposing any bugs in advance.

Bench testing can increase the initial project budget, but uncovering potential risks with a scheme while in the office is far better than discovering problems while in the field, especially depending on the complexity of the solution. Such testing also provides an extra level of confidence when the system is ultimately running under live operating conditions.

If you’re interested in any aspect of bench testing or are considering how it can save time and money for your projects, please contact Chris Dyer to discuss possible options.