Upgrade of an Aging Indoor Substation Supporting Smart City Infrastructure

1. Executive Summary

Keentel Engineering Solutions was appointed as owner's engineer and design authority for the refurbishment and digital retrofit of a four-decade-old indoor substation embedded in a dense urban core. The station is the sole 132 kV injection point for a district undergoing coordinated smart-city redevelopment, and it could not be taken out of service at any point in the works. Every element of the design and construction methodology was therefore shaped by one governing constraint: the upgrade had to be delivered live, around a rolling outage plan, with supply security preserved throughout.

A structured condition assessment — dissolved gas analysis with Duval Triangle interpretation, partial discharge measurement, tan-delta and bushing power factor testing, contact resistance, thermography and SF6 leak rate assessment — established that both the primary plant and the secondary system had reached the end of economic life. In parallel, load growth from electric vehicle fast-charging hubs, district cooling plants, data-adjacent commercial load and city-wide sensor and lighting networks was driving a capacity requirement and a materially higher expectation of availability and power quality.

Keentel delivered the full engineering package: design basis and specifications, staged system studies for every intermediate network configuration, replacement of the 132 kV switchgear and both 132/33 kV transformers, retrofit of the 33 kV and 11 kV switchboards, protection and control replacement to IEC 61850 Edition 2 with GOOSE-based interlocking and breaker failure, new station DC systems, arc-flash mitigation, fire detection and suppression, and noise attenuation for the surrounding residential envelope. Studies were executed in PSS®E, DIgSILENT PowerFactory, ETAP and PSCAD/EMTDC.

The completed works raised 132 kV switchgear short-circuit withstand to 40 kA against a prospective duty of 31.5 kA, reduced worst-case incident energy at the 11 kV switchboard from above 40 cal/cm² to below 8 cal/cm², reduced 11 kV voltage total harmonic distortion from 7.8 % to 3.4 %, and delivered the entire programme with zero unplanned customer interruptions. The original footprint was retained in full, avoiding land acquisition in a location where none was available.

Figures presented are representative of the delivered design and have been generalized to protect client confidentiality.

1.1 Project at a Glance

2. Project Context and Business Drivers

The substation occupies a masonry building of its original vintage on a constrained inner-urban plot, bounded on three sides by residential development and on the fourth by an arterial road. There is no adjacent land, no viable temporary compound, and no route by which a replacement station could be built alongside and cut over.

Three drivers converged. The first was condition: the 132 kV minimum-oil circuit breakers had exceeded their supported life, spares were being cannibalised from decommissioned sister stations, and operating records showed lengthening and increasingly variable opening times. Electromagnetic relays and an obsolete remote terminal unit left the Client without disturbance recording and with slow fault location. The second was capability: charging hubs impose sharp converter-dominated demand, district cooling adds large variable-frequency drive load, and municipal sensor, lighting and traffic networks collectively define public perception of whether the district works. The third was fault level: upstream reinforcement had raised prospective 132 kV three-phase fault duty from about 21 kA to 31.5 kA, beyond the legacy switchgear's rated breaking current.

Legacy asbestos-containing building materials were surveyed and remediated by a specialist contractor appointed directly by the Client. Keentel's role was interface management: sequencing isolations around remediation zones and confirming that removal enclosures did not compromise ventilation, egress or fire compartmentation of live electrical rooms.

The business case was therefore a life-extension and capability upgrade, not a like-for-like renewal, justified against a new-build station on acquired land priced at several multiples of the refurbishment.

3. Design Basis and Technical Requirements

The design basis was issued at Month 2 and became the controlling reference for design, procurement and testing. Nominal voltages are 132 kV, 33 kV and 11 kV, with highest equipment voltages of 145 kV, 36 kV and 12 kV. The 132 kV and 33 kV systems are effectively earthed; the 11 kV system is resistance-earthed to 1,000 A for one second, retained because changing it would have required re-rating extensive cable screens. Indoor ambient design temperature was set at 45 °C maximum with a 40 °C daily average. The site lies in a temperate maritime climate zone with low indoor pollution severity and no seismic qualification beyond base building code.

Owner's requirements added further constraints. Noise at the nearest residential façade was capped at 45 dB(A) at night. No equipment was permitted that would require breaching the building envelope for future replacement, so every item had to dismantle into components passing the existing goods access. All secondary equipment had to be maintainable by the Client's own protection staff, which influenced the process bus decision. Every stage required a documented reversion path to a secure, protected, energized configuration inside its own outage window.

4. Substation Configuration and Single-Line Architecture

The legacy station used a 132 kV double busbar with bus coupler, two transformer bays and three cable feeder bays; a 33 kV double-busbar switchboard normally run as two independent sections; and three 11 kV sections fed from retained 33/11 kV transformers, also normally split.

Keentel evaluated retention of double busbar, migration to single busbar with sectionalizer, and a one-and-a-half breaker arrangement. The decision turned on footprint and staged constructability rather than steady-state reliability. One-and-a-half breaker was eliminated immediately: the breaker count could not be housed in the available floor area. Single busbar was cheapest and smallest but removes the ability to transfer a bay between buses during maintenance, the capability the Client relies on to manage planned work here.

The decisive insight was spatial. Gas-insulated switchgear at 145 kV occupies roughly one-third the floor area of the oil-filled equipment it replaced. That released area unlocked the programme: a new double busbar GIS lineup could be built in space vacated by decommissioned ancillary rooms — a capacitor room, a former workshop and a disused cable flat — while the legacy switchgear remained energized in position. Cut-over became a cable and busbar transfer rather than a demolish-then-build sequence, avoiding an N-0 condition lasting weeks.

Civil provision was retained for two further 132 kV bays, and the spare bay was fully equipped rather than left blank, because a future bay installation here is an outage problem before it is a cost problem.

5. System Studies and Analysis

Studies were run not once for a final state but for every intermediate configuration. Eight construction stages, each with its own energized topology, generated a study matrix validated individually before an outage was authorized.

5.1 Modelling Basis and Software

The network model was built in PSS®E from the Client's planning case and extended to cover the full 33 kV and 11 kV distribution. DIgSILENT PowerFactory carried staged short-circuit and coordination work, holding eight topologies in one project through variation and expansion stages rather than eight divergent copies. ETAP handled arc flash, harmonics and DC sizing; PSCAD/EMTDC handled transients. Validation used impedances from routine test certificates, cable capacitances from commissioning records, and issued source impedance with a stated tolerance of plus or minus 8 %.

5.2 Staged Short-Circuit Studies

Fault duties were computed for each stage under maximum and minimum plant, including transitional periods with one new transformer in parallel with a legacy unit of different impedance. That parallel case proved binding for several 33 kV feeder breakers, because the mismatch shifted fault current away from the balanced assumption of the original design.

The 40 kA selection sits deliberately above the 31.5 kA duty. Further reinforcement is indicated within the asset's design life, and 40 kA cost under 4 % more than 31.5 kA in a 145 kV GIS lineup — a low price for removing a future replacement trigger.

5.3 Load Flow and N-1 Contingency Analysis per Stage

Each stage was assessed against summer peak, winter peak and the emerging shoulder-season peak created by charging behaviour. No stage was to present a first-contingency condition exceeding continuous ratings, except where emergency ratings were consciously relied upon.

Stages 3 and 5 were the programme's genuine exposure. Each was confined to a defined low-load season, executed with a pre-agreed 11 kV back-feed from an adjacent network through a normally open point supporting 6.5 MVA of critical load, and backed by contracted mobile generation on site.

5.4 Transformer Loading and Emergency Rating Study

An IEEE C57.91 study established what the station could survive during N-0 stages, assessing the single in-service transformer at 60 MVA for four hours — 120 % of ONAF rating — at design ambient.

This gave operations a bounded envelope rather than an anxiety. Cumulative loss of life across the programme closed out at 0.09 %.

5.5 Arc Flash Study

Arc flash was assessed to IEEE 1584-2018 for the 11 kV switchboards, which fall inside the model's validated range. For 33 kV and 132 kV equipment, outside that range, incident energy was estimated by the Lee arc-in-open-air method, with its limitations stated explicitly rather than obscured.

Three measures acted together: arc flash detection on light and overcurrent criteria, cutting detection to 8 ms; new vacuum breakers, cutting interrupting time from five cycles to three; and a maintenance setting group, selectable at the HMI and annunciated to SCADA, removing intentional grading delay on the incomer during work. Internal arc classification was confirmed as IAC AFLR 25 kA 1 s.

5.6 Harmonic Study

Three weeks of baseline measurement established 11 kV voltage THD at 7.8 %, above the 5.0 % IEEE Std 519 limit at that voltage, with the fifth harmonic at 6.1 % against a 4.0 % individual limit.

Mitigation combined two 4.8 Mvar detuned banks at 11 kV, tuned below the fifth harmonic at 4.2 times fundamental, with a connection policy requiring active front-end or 12-pulse topologies for new district cooling installations above 500 kW. A resonance scan across the full switching matrix confirmed no parallel resonance within 10 % of a characteristic harmonic — a check that would have exposed a near-coincident eleventh-harmonic resonance had a conventional 6 % detuning ratio been used.

5.7 Protection Coordination and Settings Migration

Coordination was re-derived from first principles rather than transcribed, the legacy inverse-time grading having been chosen decades earlier for slower breakers. Margins were 300 ms between electromechanical and numerical devices during transition, reducing to 200 ms once every device in a grading path was numerical. Every numerical relay carried four populated setting groups, so a stage transition became a group change under a controlled switching programme.

5.8 EMT Studies in PSCAD/EMTDC

PSCAD/EMTDC was used for transient studies. Vacuum interruption of the retained 33/11 kV transformers under light load produced prospective chopping overvoltages of 2.6 p.u.; relocating surge arresters to the transformer terminal box reduced this to 1.7 p.u. against a 2.9 p.u. withstand. Very fast transient overvoltages from GIS disconnector operation reached 2.1 p.u. with rise times below 20 ns, addressed by damping resistors on the disconnectors of the two bays closest to the transformer bushings. Energization inrush drove point-on-wave closing on both transformer bays, cutting peak inrush from 7.2 p.u. to 1.4 p.u.

5.9 Ferroresonance Assessment

The retrofit created arrangements in which a wound voltage transformer could be left connected to busbar isolated through the grading capacitance of an open breaker. Simulation with saturable VT models found two 33 kV configurations sustaining subharmonic oscillation beyond two seconds. The fix was partly procedural, through switching rules preventing the configuration arising, and partly physical: damping resistors across the open-delta tertiary windings.

5.10 Grounding Study and Verification

The grid was original to the building and substantially built into structural elements, so wholesale replacement was impossible. Adequacy was assessed to IEEE Std 80 against the increased fault duty using a two-layer soil model from Wenner traverses. The design current is derived from the symmetrical single-phase-to-ground fault at the 132 kV busbar, 29.8 kA, not from the 31.5 kA three-phase value: only zero-sequence current returns through earth, so a three-phase fault produces no net current in the grid and would overstate the design duty by about six percent here.

Compliance was achieved without rebuilding the grid: a perimeter conductor in the cable basement, the new GIS enclosure bonded at four points to different grid nodes, and a 150 mm crushed rock surface layer in the two external areas where operators handle equipment. Fall-of-potential testing required a current electrode route along a footpath easement and correction for buried metallic services.

6. Protection, Automation and Control Philosophy

6.1 Protection Scheme Architecture

Protection is duplicated at 132 kV and on the transformer bays. Main 1 and Main 2 use relays from different product families on separate DC supplies, CT cores, trip coils and fibre paths. The 132 kV cable feeders use line differential as Main 1 and distance as Main 2; autoreclose is blocked on all cable circuits.

The final breaker failure time is 150 ms, from 60 ms interrupting time plus reset and margin. Where legacy five-cycle breakers remained in a zone it was widened to 200 ms, then narrowed once the last legacy breaker was removed. Missing that adjustment is a common source of spurious breaker failure operations during phased replacement.

6.2 IEC 61850 Architecture and the Process Bus Decision

The station bus is a dual-star gigabit Ethernet network with Parallel Redundancy Protocol at all protection IEDs, and High-availability Seamless Redundancy on the ring serving the retrofitted 11 kV boards. GOOSE carries interlocking, breaker failure initiation, bus zone tripping, arc flash blocking and maintenance group indication; MMS carries SCADA and engineering traffic. VLANs separate GOOSE, MMS and time traffic, with GOOSE priority tagged and multicast filtered at each port.

Process bus with non-conventional instrument transformers was evaluated seriously and rejected, with the rejection documented. The retained switchboard structures could not accept non-conventional units without full replacement, destroying the constructability case; the copper between switchgear and relay room already existed in serviceable condition; staged cut-over of a sampled-value scheme alongside legacy hardwired protection introduces failure modes difficult to prove out live; and the Client's protection staff had no sampled-value competency. The option is preserved: bay controllers are merging-unit capable and spare fibre reaches every bay.

6.3 Time Synchronization

The requirement was ±1 ms for sequence-of-events records and ±1 µs for any future sampled-value or point-on-wave application. IEEE 1588 PTP to the C37.238 power profile was adopted as primary, with transparent clocks in the station switches. IRIG-B was retained in parallel for legacy fault recorders and as a fallback independent of Ethernet health. Two GNSS receivers feed redundant grandmasters specified for better than 1 µs holdover over 24 hours.

6.4 Cyber Security

Zoning follows IEC 62443, with the station control zone at Security Level 2 and conduits defined for every crossing between the station zone, the operations WAN and vendor engineering access. There is no routable connectivity from the corporate environment; engineering access is via a hardened jump host with multi-factor authentication inside an electronic security perimeter. Role-based access follows IEC 62351-8, with individual accounts and privilege separation between viewing, operating and setting-change roles. Configuration baselines are captured per device, firmware verified against vendor signatures, and unused ports and services disabled. Controls were aligned to NERC CIP-style obligations, including a 35-day patch evaluation cycle.

6.5 The Legacy-to-Numerical Transition Interface

For roughly eighteen months, electromechanical and numerical protection operated on the same primary plant. This interface generated more engineering effort than any other element, and it is where phased retrofits most often fail.

Current transformer sharing was the first problem. Legacy cores were 5P20 class with burden allowance sized for electromechanical relays; new numerical relays present under 0.1 VA. Where a core served both, composite burden was recalculated and a saturation check performed at the new fault duty. Two cores failed and were replaced during an already-planned switchboard outage.

Busbar protection was the second. The legacy 132 kV scheme was high-impedance with 1200/1 cores; the new scheme is low-impedance with 2000/1. High-impedance schemes cannot tolerate mixed ratios, so the two could not overlap. The resolution was a temporary low-impedance bus zone scheme, commissioned on the new relays but fed from legacy cores through interposing CTs, in service for eleven weeks. It received a full primary injection stability test rather than being accepted unproven on a live double busbar.

Interlocking was the third. Rather than attempt hybrid logic across the boundary, Keentel set a hard rule: interlocking within a fully migrated zone is GOOSE; anything crossing the legacy boundary stays hardwired, through binary outputs from the new bay controllers into existing contact chains. The translation is one-directional and its truth table was verified by physical operation of every permissive and block, not by simulation. Intertripping across the boundary stayed on hardwired DC pilot, because the legacy end could not participate in GOOSE supervision.

7. Primary Plant, Insulation Coordination and Physical Design

The replacement transformers are 40/50 MVA ONAN/ONAF, 132/33 kV, with on-load tap changers giving plus 10 % to minus 15 % in 21 steps and 14 % impedance on the 50 MVA base. Impedance was raised from the legacy 11.5 % specifically to moderate 33 kV fault duty under the stronger upstream source, allowing the existing 25 kA switchboard structure to be retained. Losses and sound level were guaranteed with liquidated damages: no-load loss fell from approximately 38 kW per unit to 19 kW, load loss at 50 MVA is 215 kW, and sound pressure was specified at 62 dB(A) at one metre against approximately 78 dB(A) on the units removed.

All margins exceed the 1.2 minimum of IEC 60071-2 with substantial reserve, appropriate for an indoor station where arrester replacement is disruptive and where the very fast transient environment inside GIS is less well characterised than the standard lightning impulse case.

Physical design was governed by access rather than clearance. Every major component was checked against a three-dimensional route from the goods entrance to final position, including the 90-degree turn at the head of the transformer bay ramp that had defeated an earlier replacement attempt. Transformer handling used skidding and jacking with temporary steelwork spreading load onto identified structural lines.

8. Auxiliary Systems

Two independent 110 V DC systems were installed, each with its own battery, charger, distribution board and earth fault monitoring, feeding Main 1 and Main 2 protection. Sizing followed IEEE Std 485 against a duty cycle of continuous load, a one-minute initial inrush, an eight-hour discharge and a final minute covering simultaneous tripping of all breakers.

A separate 48 V DC system of 200 Ah serves communications and station bus switches, so a protection DC fault cannot silence telemetry when it is most needed. AC station service is dual-fed from separate 11 kV boards with automatic changeover, and a standby generator connection point with mechanical interlock was added; its absence had previously extended a restoration by several hours.

Fire strategy follows NFPA 850 principles adapted to an occupied urban building. Aspirating smoke detection covers the relay room, battery rooms and cable basement. Transformer cells use high-pressure water mist, selected over deluge because building drainage limited acceptable water volume and because mist performs better in a confined cell; the control and relay room uses an inert gas system with pre-discharge alarm and abort. Cells are separated by two-hour rated construction, and cable penetrations were re-sealed after a survey found 41 breached by later cable additions. Oil containment sumps hold 110 % of the largest tank volume plus firewater.

Noise attenuation was engineered rather than assumed. Transformer cells received acoustic doors, absorptive linings and splitter attenuators on supply and extract ventilation, with fan selection re-checked for the added pressure drop. Measured levels at the nearest residential façade were 41 dB(A) at night against the 45 dB(A) requirement.

9. Construction, Commissioning and Energization Support

9.1 Outage Staging Methodology

The staging plan was built backwards from the operational constraint rather than forwards from the construction sequence. Keentel first established, with the Client's control room, the network conditions that were absolutely prohibited, then those permitted only within defined seasons and hours, then those permitted freely. Only after that envelope was fixed was a construction sequence developed to fit inside it. Sequences developed from a contractor's logic and tested against operational limits afterwards almost always require renegotiation late.

Six N-0 windows totalling 14 hours were required. Each was authorized individually, no earlier than 72 hours in advance, against a checklist covering confirmed weather, confirmed back-feed availability, mobile generation load-bank tested within 24 hours, and a named reversion decision-maker on site. Any window that could not start within 90 minutes of plan was abandoned rather than compressed.

Every stage carried a written reversion plan specifying the configuration to be restored, the settings applicable to it, and the maximum elapsed time to a secure state. Reversion was invoked twice: once when a cable termination test result fell outside acceptance and once when a delivery arrived damaged. On both occasions the station was returned to a secure, fully protected configuration inside the window.

9.2 Risk Register

9.3 Testing, FAT and Commissioning

Factory acceptance testing covered the GIS lineup, the transformers and the complete protection and control system, the last on a simulated station with the actual SCD file loaded. Substituting a representative configuration is a false economy: the true SCD exposed 14 discrepancies, including two GOOSE subscriptions bound to the wrong logical node instance. SCD and ICD management followed a single-master rule — one file under Keentel configuration control, all vendor files issued from and returned to it, every change carrying a controlled revision and difference report.

Site testing followed five levels: device settings and hardware verification; scheme-level secondary injection; end-to-end functional proving including primary injection where CT circuits changed; system integration including GOOSE performance measurement under load; and an operational readiness demonstration in which control room staff worked the new interface against scripted scenarios before any bay was declared available. GOOSE transfer times measured a mean of 3.1 ms and a 99th percentile of 5.8 ms against a 10 ms requirement.

10. Results and Value Delivered

Three outcomes carried value beyond the table. The footprint was retained in full, avoiding a land acquisition that was not realistically available. The station is now remote-operations ready, with condition monitoring on both new transformers and remote switching under the Client's existing authorization framework. And the 38 kW reduction in no-load loss removes roughly 333 MWh of annual loss energy.

11. Challenges and Engineering Lessons Learned

The staging plan is the design. On a live brownfield station the outage sequence is not a construction detail to be resolved after the technical design; it is the primary design constraint and must be fixed first. The spatial arbitrage that made this project feasible was identified in the first six weeks. Found six months later, the programme would have been considerably worse.

Study the transitional states, not just the end state. The binding 33 kV fault duty case was not the final configuration but a transitional one, with a new and a legacy transformer of different impedances in parallel. Several of the most demanding settings issues applied only to configurations that existed for weeks.

Mixed-vintage protection needs a hard boundary rule. The most useful single decision in the secondary works was that GOOSE is used only within fully migrated zones and everything crossing the legacy boundary stays hardwired. It cost a little copper, eliminated a whole class of transition failure modes, and made every interface reviewable in an afternoon.

Retained assets deserve as much design attention as new ones. Raising transformer impedance from 11.5 % to 14 % was a small specification change that saved an existing 25 kA switchboard from replacement. Such opportunities are only visible if retained plant is modelled with the rigour applied to new plant.

Document rejected options as carefully as chosen ones. The process bus decision was revisited twice, once by a new stakeholder and once in a value engineering exercise. On both occasions the documented rationale closed the question in under a week.

12. Keentel Capability Summary

• Condition assessment and residual life evaluation: DGA and Duval Triangle interpretation, tan-delta, partial discharge, thermography, contact resistance, SF6 leak rate
• Brownfield outage staging, reversion planning and stage-by-stage risk registers
• Staged short-circuit, load flow and N-1 analysis across intermediate topologies
• Transformer loading and emergency rating studies to IEEE C57.91
• Arc flash studies to IEEE 1584-2018 and Lee method above its voltage range
• Harmonic penetration, resonance scanning and filter design to IEEE Std 519
• EMT studies in PSCAD/EMTDC: switching transients, VFTO, inrush, ferroresonance
• Grounding assessment to IEEE Std 80 and verification to IEEE Std 81
• Protection philosophy, coordination and staged settings migration
• IEC 61850 Ed. 2 architecture, SCD/ICD control, GOOSE performance verification
• Cyber security zoning to IEC 62443 and role-based access to IEC 62351
• Station DC and UPS sizing to IEEE Std 485, station service and standby supply
• Fire detection and suppression, oil containment and acoustic attenuation design
• FAT, SAT, commissioning management, energization support and punch-list closure

13. Frequently Asked Questions

14. Glossary of Terms and Abbreviations

15. Confidentiality and Use Statement

This case study has been prepared for informational purposes. All client identities, project locations, contract details, and proprietary data have been withheld or generalized. Technical parameters, study results, and figures presented are representative of work performed by Keentel Engineering Solutions and have been adapted so that no individual project, owner, or facility can be identified. Nothing in this document constitutes a design recommendation for any specific installation. Any reuse of the methodologies described requires project-specific engineering analysis by a qualified professional engineer.