One platform for the condition of Ghana's transformer fleet — national oversight for the Energy Commission, real-time fault detection at the distribution transformer, condition diagnostics for transmission assets, and a field engineer's application that turns an alarm into a repaired transformer. Built in Accra, aligned to Ghanaian law and regulatory practice from the first screen.
Ghana's distribution network is largely uninstrumented. Transformers fail from sustained overload without warning. LV fuses are stolen. High-voltage cutout failures single-phase customer supplies for hours while being misdiagnosed from the ground. And the outage durations the Energy Commission regulates are reported by the utility being regulated. GridWatch closes all four gaps from a single platform.
National load, regional capacity outlook, peak-window stress and fleet health across every operator — ECG, NEDCo, GRIDCo, Enclave and private distribution — in one independent view the Commission controls.
Sustained overload, phase imbalance, LV fuse loss, HV cutout single-phasing, tamper and supply loss — each classified before a ticket is raised, each with its own dispatch instruction and equipment list.
Dissolved gas analysis, hot-spot temperature, insulation life consumed and remaining service life for transmission and bulk-supply assets, with a structured field inspection record against every unit.
Three findings shape everything that follows, and we state them at the front rather than in a footnote.
Electricity theft detection cannot be delivered today. It requires interval meter reads from every downstream customer and an accurate transformer-to-customer mapping. Neither exists at scale (Section 06).
Asset protection alone does not pay for itself. On our own figures the benefit is smaller than the running cost. The case for funding it is regulatory capability and public safety, not return (Section 09).
POA Nexgen previously drafted terms of reference for this system. We disclose this unprompted and we are not bidding for a build contract on that basis (Section 12).
POA Nexgen Technologies is a Ghanaian IT solutions company registered with the Public Procurement Authority. We build, deploy and support software for public and private institutions across Ghana — with local implementation teams, fixed transparent pricing, and support that answers the same day.
GridWatch is our platform for the energy and utilities vertical, delivered end-to-end: hardware supply and installation, platform deployment, staff training and ongoing support under SLA. Where this proposal involves imported electronics, we assemble and calibrate locally and commit to knowledge transfer to Commission and utility engineers.
| Condition | How it is found today | Consequence |
|---|---|---|
| Sustained overload | Customer complaint, or the failure itself | Transformer burnout and unplanned replacement at full cost |
| Phase imbalance | Rarely detected at all | Customer equipment damage, neutral conductor stress |
| LV fuse loss or theft | Customer complaint | Extended partial outage on one phase |
| HV cutout failure | Frequently misdiagnosed as an LV fault | Crew sent to the wrong equipment; motors burn out while single-phasing persists |
| Tamper, oil and copper theft | Discovered on the next inspection | Asset loss and extended outage |
| Insulation ageing | Periodic oil sampling, if resourced | End-of-life failures that were forecastable years in advance |
| Outage duration | Self-reported by the licensee | No independent basis for regulating reliability performance |
When a high-voltage fuse fails, the transformer does not switch off. Power back-feeds through its windings and the affected low-voltage phase sits at a reduced but non-zero voltage, commonly 30–58% of nominal†. Customers report dim lights or that some appliances work — a low-priority call. Meanwhile three-phase motors overheat and burn out, and the transformer itself is damaged.
From the ground it resembles a blown LV fuse. A crew arriving with a fuse link achieves nothing and must return with different equipment for work at a different height. GridWatch separates the two, and tells the crew which one it is before they leave the depot.
GridWatch consolidates three previously separate POA Nexgen prototypes — the national transformer load monitor, the GridWatch operations console, and the fault-detection engine — into one platform. Every module is switchable per deployment: the Commission runs the full regulatory stack while a distribution utility enables only the operational modules it needs.
Installed capacity against peak demand, total load and system frequency, with the 6–10pm peak window tracked separately from off-peak behaviour.
All sixteen regions with average load, transformers monitored and units needing attention, scoped by the signed-in role.
Ownership mix across ECG, NEDCo, GRIDCo, Enclave Power and private distribution, so the Commission can hold each licensee to its own numbers.
Regional headroom projected forward, identifying where reinforcement is needed before load overtakes installed capacity.
Amber at 90% and red at 100% of nameplate, confirmed over consecutive intervals rather than on a single reading. Duration filtering matches how thermal ageing actually works.
Deviation from the three-phase mean, gated on minimum load so a lightly loaded unit does not generate noise.
Voltage and current collapse on one phase while the others remain live — a pulled or blown fuse, dispatched to ground level with the correct fuse rating.
Back-feed detected between 15% and 60% of the live-phase mean with current still flowing. A different fault, a different crew, a different instruction.
Readings that are physically impossible are identified as device faults, not grid faults — preventing a wasted line-crew dispatch on a failed voltage tap.
Enclosure opening, tilt and vibration; and a last-gasp transmission on total supply loss that timestamps the outage independently of the utility.
Hydrogen and acetylene concentrations tracked against action thresholds, flagging incipient internal faults on transmission and bulk-supply transformers.
Estimated winding hot-spot temperature and ageing acceleration factor, the two variables that actually determine insulation life.
Cumulative life consumed and projected remaining service life per asset, turning replacement planning into a schedule rather than a reaction.
Structured checklist — oil level and leaks, silica gel breather, bushings, cooling, earthing, corrosion, abnormal noise — logged against the asset with inspector and date.
The notification names the fault class and lists the equipment to carry, so the crew arrives once with the right kit.
Android-first mobile application: job list, live readings at the pole, resolution capture with photographic evidence, and an encrypted offline buffer for dead zones.
Per-phase correction factors captured against a reference instrument at installation. This is what makes every alarm threshold in the platform defensible.
Barcode capture, binding to the asset register, GPS lock and an end-to-end connectivity diagnostic before the engineer leaves site.
Critical dispatch delivered by SMS where data coverage fails, so an alert is never stranded in a dead zone.
A single queue of every open job in the supervisor's scope, with severity, the asset, the fault classification and who it is assigned to. The head of field engineering sees the whole district on one screen.
Jobs are assigned to a named engineer from the on-shift roster, with each engineer's current open and closed count visible — so work is distributed on evidence rather than on whoever answers the phone.
Every job carries a running clock against its severity target. Jobs approaching or past target are flagged before they are breached, not reported afterwards.
An engineer cannot close a job the network still reports as faulty. When resolution is submitted the platform re-reads the asset; if the fault signature persists the closure is refused and the job returns to the queue.
Every closure and every rejected closure recorded with the asset, the engineer, the time and the reason — an evidence trail for both the utility and the Commission.
An unacknowledged critical job escalates on a timer to the supervisor and then to area management, and the delay is itself recorded as a compliance event.
Reliability indices computed from measured outage events by region and by operator, on an independent record rather than a licensee submission.
Fleet, incident, reliability and compliance reports generated from live data, exportable as CSV for the Commission's own analysis.
Transparent benefit model with every assumption exposed and editable, so the Commission can test the case against its own figures rather than ours.
Energy Commission, transmission engineer, distribution engineer north and south, and field engineer — each seeing only the assets in their mandate.
Every state-changing action recorded with actor, role and timestamp, with a read-only auditor export.
A monitoring platform is only as good as its willingness to say when it does not know. Three design commitments carry that.
Clip-on current sensors carry roughly ±3.4% combined uncertainty, so a single threshold crossing is meaningless. Faults must hold across consecutive intervals before a ticket is raised. A platform that cries wolf is ignored within a month.
Sensor seating error is the dominant source of measurement uncertainty. A per-phase correction captured against a reference at installation reduces combined uncertainty from about ±5.8% to ±2.3%, and is stored against the asset record.
These are Class 1–3 instruments, not revenue meters. GridWatch data directs an inspection; the inspection produces the evidence. It must not be the sole basis for a financial penalty, and we will not represent it as such.
Energy-balance reconciliation compares what a transformer exports against the sum of what its customers consume. That requires time-synchronised interval reads from every downstream meter — the residential base is largely prepaid meters that record purchase, not time-stamped consumption, and are not remotely read at scale — and an accurate transformer-to-customer mapping, which is incomplete and stale.
Even with those in place, a naive threshold would fail. Measurement uncertainty totals roughly ±10.4% at 95% confidence, and unmetered street lighting plus unmapped legitimate customers add a further +5% to +13% bias — pointing in the same direction as theft. A transformer with no theft at all would routinely read 5–20% unaccounted loss. A threshold set at 8%, as commonly proposed, sits entirely inside that noise and would raise continuous false alarms across the fleet.
What would work is measuring change against each transformer's own ninety-day baseline, which cancels the systematic errors, combined with time-of-day signature analysis — after street lighting is metered or modelled out and the customer mapping is cleansed.
We recommend the Commission treat metering and asset-register investment as the enabling programme. That is where the large recoverable value sits†. GridWatch is designed to consume that data the day it exists, and is complementary to it — not a substitute.
| Framework | How GridWatch satisfies it |
|---|---|
| Energy Commission Act 1997 (Act 541) | Technical oversight, device standards and independent data access under the Commission's own mandate. ✓ |
| PURC Act 1997 (Act 538) | Reliability indices supplied to the economic regulator; any performance-linked financial consequence remains PURC's decision, not the platform's. ✓ |
| L.I. 1816 · Electricity Supply & Distribution Rules | Correct vehicle for technical obligations — device installation on licensee assets and mandated data access. Not used to impose penalties. ✓ |
| Data Protection Act 2012 (Act 843) | Transformer-level aggregation by default; documented lawful basis, DPC registration and impact assessment before any customer-level data is processed. ✓ |
| Public Procurement Act 663 / Act 914 | Competitive tender pathway proposed; prior terms-of-reference involvement disclosed in full. ✓ |
| National Communications Authority | Radio module type approval initiated in parallel with hardware design, not after it. ✓ |
| Ghana Standards Authority | Device approval for equipment attached to distribution infrastructure; calibration traceable to a reference standard. ✓ |
| IEC 62443 · cyber security | Signed firmware and secure boot, device certificate lifecycle, private carrier tunnel, documented incident response. ✓ |
A regulator-owned device attached to a licensee's live network raises unresolved questions of ownership, liability, permit-to-work and maintenance access. We recommend the units are owned and operated by the distribution utility under an Energy Commission technical directive, with a guaranteed independent read-only feed to the Commission. This places the equipment with the party already holding the safety case, and removes the licensee's structural incentive to obstruct — the largest non-technical risk in the programme.
Fastest to launch. Managed hosting with uptime SLA, nightly encrypted backups and patch management. Suited to the pilot phase.
Data residency on Government of Ghana infrastructure. Recommended for a platform holding national critical-infrastructure telemetry.
Installed on Commission servers with hardening and knowledge transfer. Air-gapped analysis environments available.
Ghana's distribution fleet is mounted two ways, and the difference affects cost, access and signal. The asset survey in Section 09 establishes the actual mix in the chosen district before any unit is ordered.
| Mounting | Typical asset | What it means for installation |
|---|---|---|
| Pole-mounted | Smaller units serving residential streets. The most common arrangement in Ghanaian towns. | Elevated work under live-line procedure, insulated tooling, rated PPE. Carries the higher installation rate. Good cellular signal; the standard antenna suffices. |
| Ground-mounted | Larger, heavier units on concrete plinths inside cages or walled compounds, serving commercial districts, estates, institutions and industrial areas such as Tema. | No climbing, so cheaper and safer to fit. But metal cages and enclosures attenuate the signal, so an external antenna on a lead is required. Access is controlled, which needs scheduling with the asset owner. |
Ground-mounted units are also disproportionately the larger ratings, which is precisely the population that carries oil monitoring. The condition diagnostics in Section 04 therefore concentrate on this group, while the fault detection applies to both.
Government and ECG replacement and reinforcement programmes are installing new distribution transformers across the country†. Every one of those is an opportunity to fit monitoring at a fraction of the retrofit cost.
When a unit is being replaced the crew, the vehicle and the outage window are already committed, and the transformer is de-energised. Fitting a monitoring unit at that moment removes the separate site visit, removes the live-line work entirely, and adds no customer outage beyond the one already taking place. The incremental cost falls to roughly a third of the retrofit rate.
We therefore recommend the Commission direct that new and replacement distribution transformers are fitted at the point of installation, and that retrofit is reserved for the existing fleet. On a programme of any size this is the single largest cost reduction available, and it is a policy decision rather than a technical one.
Figures are indicative and in US dollars, because the electronics are imported and priced in dollars; the Ghana Cedi equivalent is fixed at the prevailing rate on the date of firm quotation. Hardware engineering and platform development are one-time costs that amortise across any later expansion — they are not pilot overhead. Import duty and VAT are included at full rates; no Free Zones exemption is assumed.
| At 1,000 monitored units | Annual | Basis |
|---|---|---|
| Benefit from avoided transformer failures | US$ 117,000† | 9 failures avoided per year at US$13,000 all-in |
| Platform operating cost | US$ 171,000 | Hosting, connectivity, support, field replacement, recalibration |
| Net position | − US$ 54,000 | Before any capital recovery |
Asset protection alone does not pay for itself, and this remains true at larger fleet sizes. A regulator may quite properly fund a system for public safety, network resilience and independent oversight — but it must do so knowing that, not on the strength of a return that would not survive audit.
Two things do change the position. HV cutout discrimination requires no additional hardware, because three-phase voltage sensing is already fitted, and pays back its incremental cost in roughly eighteen months. And commercial loss recovery is roughly an order of magnitude larger than asset protection — but it is gated on the metering investment described in Section 06.
The benefit stream accrues to the distribution utility; the cost falls on the Commission. That mismatch should be resolved in the funding structure before a contract, not after.
Hardware and firmware development, EMC pre-compliance, calibration rig. NCA and Ghana Standards Authority approvals initiated in parallel. Detection rules run in advisory mode — logging, not dispatching.
Pilot manufacture and local assembly, asset survey of the pilot district, crew certification for live-line installation, platform deployment.
150 units installed without customer outages. Controlled HV fuse-removal testing across representative transformer types to establish real back-feed thresholds.
Cutout detection promoted from advisory to dispatch-grade once validated against measured behaviour. Field application released to production.
Instrumented fleet compared against the matched control group. Real failure rates, real event rates, business case re-derived on measured evidence — replacing every assumption in this document.
Nobody knows Ghana's real distribution transformer failure rate, what fraction of those failures early warning could actually prevent, or how often high-voltage cutouts fail. Every figure in Section 09 marked † rests on an assumption.
150 instrumented units against 150 matched controls over eighteen months is the only design that can answer them. Without a control group no observed change can be attributed to the system, and the Commission would be asked to approve a national programme on uncontrolled evidence. With one, the scale decision is taken on measured fact — and the resulting business case survives scrutiny by the Auditor-General.
We disclose this unprompted. In most procurement regimes a bidder who wrote the specification is disqualified or heavily scrutinised, and a specification drawn from one vendor's architecture cannot support genuine competition.
We therefore ask that our earlier document is not adopted as the Commission's terms of reference, and we are not bidding for a build contract on that basis.
The pathway we propose: this concept note is received as an unsolicited proposal; the Commission or a development partner commissions an independent feasibility study, which we may or may not be engaged to perform; that study produces the Commission's own terms of reference; and any subsequent build is competitively tendered under the Public Procurement Act, where we compete on merit.
We recognise this forgoes a shorter commercial path. We regard a procurement that survives audit as worth considerably more than one that does not.
A pilot without pre-agreed metrics becomes an argument about interpretation. These are proposed as contractual acceptance criteria, measured against the matched control group, and agreed before a single unit is installed.
| Indicator | Target at month 18 | How it is measured |
|---|---|---|
| Platform availability | ≥ 99.5% | Monthly uptime of the console and API, excluding notified maintenance windows |
| Device reporting rate | ≥ 97% | Expected telemetry intervals actually received per unit per month |
| Measurement accuracy | ± 2.3% | Post-calibration verification against a reference instrument on a 5% annual sample |
| False alarm rate | < 5% of tickets | Tickets closed as "no fault found" as a share of all dispatched tickets |
| Fault classification accuracy | ≥ 90% | Field-confirmed fault type against the classification the platform issued |
| Detection lead time | ≥ 30 min median | Time between platform alert and the first customer report of the same event |
| Avoided failures | Measured, not targeted | Instrumented fleet failure rate against the 150 matched control transformers |
| Outage duration record | 100% of events | Independently timestamped outages usable for SAIDI and SAIFI computation |
| Field adoption | ≥ 80% of tickets | Tickets closed through the mobile application rather than outside the system |
Avoided failures is measured, not promised. Nobody currently knows Ghana's real distribution transformer failure rate or what fraction early warning can prevent. Committing to a number before measuring it would be the same error as the payback claim we withdrew. The pilot exists to establish it.
An alert that reaches nobody is worse than no alert, because it creates a false record of having been raised. Every notification path below is logged, acknowledged and escalated.
| Event | Severity | Channel | Escalation if unacknowledged |
|---|---|---|---|
| HV cutout single-phasing | Critical | Push, SMS, console | District engineer at 15 min, area manager at 45 min |
| LV fuseway open phase | Critical | Push, SMS, console | District engineer at 15 min, area manager at 45 min |
| Enclosure tamper | Critical | Push, SMS, console | Security desk immediately, area manager at 30 min |
| Sustained overload, red | Critical | Push, console | District engineer at 30 min |
| Supply loss, dying gasp | Critical | Push, console, outage log | Control room at 10 min |
| Sustained overload, amber | Warning | Console, daily digest | Planning review, next working day |
| Phase imbalance | Warning | Console, daily digest | Planning review, next working day |
| Dissolved gas elevated | Warning | Console, asset owner email | Oil sample scheduled within 14 days |
| Device sensor fault | Warning | Console, maintenance queue | No line crew dispatched at any stage |
No notification is issued on a single reading. A fault must persist across three consecutive reporting intervals, and a change of fault type restarts the count so a crew never receives a stale instruction.
Critical dispatch falls back to SMS where coverage will not carry a data session, so an alert is never stranded in a dead zone. Acknowledgement by reply is logged against the engineer.
An unacknowledged critical alert escalates on a timer and is recorded as a compliance event in its own right. The Commission can see not only what failed, but how long it took anyone to respond.
Figures marked † throughout this document are assumptions requiring verification from named Ghanaian sources. They are listed here in full. None should be quoted as a finding. Several can be closed in a single working session with ECG and GRIDCo; the remainder are what the pilot is designed to measure.
| Assumption | Source required to close it |
|---|---|
| Distribution transformer failure rate of 3% per year | ECG and NEDCo maintenance records |
| 30% of failures preventable by early warning | Pilot control-group comparison |
| All-in replacement and outage cost of US$13,000 per failure | ECG procurement and outage records |
| Commercial loss rate of approximately 25% | ECG and PURC loss reporting |
| Average end-user tariff | PURC published tariff schedule |
| HV cutout failure frequency | Pilot measurement — currently unknown to anyone |
| Back-feed voltage band of 30–58% of nominal | Controlled characterisation testing, months 6–11 |
| Installation cost of US$220 per pole | Competitive local tender |
| Import duty and VAT at approximately 21% aggregate | Licensed clearing agent |
| Unmetered street lighting at 3–8% of throughput | ECG unmetered load register |
| Machine-to-machine connectivity pricing | MTN and Telecel quotation |
| Type approval cost and lead time | NCA and Ghana Standards Authority |
| Energy Commission and PURC mandate boundary | Ghanaian counsel |
| Availability of an unsolicited-proposal route under Act 663/914 | Public Procurement Authority |
| Pole-mounted to ground-mounted ratio in the chosen pilot district | Asset survey, and ECG district records |
| Volume and schedule of the national transformer replacement programme | Ministry of Energy and ECG directly — press reporting is not a sufficient basis |
Stated so that neither party discovers it during delivery.
| Risk | Severity | Mitigation |
|---|---|---|
| The utility declines or obstructs installation on its assets | High | Adopt the ownership structure in Section 07: units owned and operated by the utility under an Energy Commission directive. Engage ECG and NEDCo as co-sponsors from day one, not as subjects of the programme. |
| Real event rates prove far lower than assumed | High | The pilot is explicitly designed to measure them against a control group. The scale decision is deferred to month 18 rather than committed now. |
| Lightning and surge losses exceed the 4% annual allowance | Medium | Surge protection specified in the bill of materials; environmental qualification to IEC 61000-4-5 within the engineering budget; failure rate measured on the pilot before scale. |
| Type approval takes longer than programmed | Medium | NCA and Ghana Standards Authority submissions run in parallel with hardware design from month 1, not after design freeze. |
| Cellular coverage insufficient at rural pilot sites | Medium | Site survey before unit allocation; 2G fallback in the modem specification; store-and-forward buffering on the device. |
| Cedi–dollar movement on a dollar-denominated contract | Medium | Allocation of exchange risk agreed explicitly before contract; quotation validity limited to 90 days. |
| Demonstration mistaken for a deployed system | Medium | Simulation notice on every screen of the console and in every written reference to it. |
| Mandate dispute between the Commission and PURC | Medium | Joint governance proposed from the outset, with PURC copied on this document. |
| Field crews stop trusting alerts | Medium | Three-interval confirmation hold, install calibration, and a false-alarm rate carried as a contractual acceptance metric in Section 13. |
| Key personnel loss at POA Nexgen | Low | Documented architecture, source code escrow available on request, and knowledge transfer to Commission engineers scheduled within the pilot. |
18.1 Validity. This proposal is valid for 90 days from the date of issue. Pricing is indicative pending scoping; a firm quotation follows within five working days of agreed scope.
18.2 Currency and pricing. Figures are stated in United States dollars because the electronics are imported and priced in dollars. The Ghana Cedi equivalent is fixed at the prevailing interbank rate on the date of firm quotation. Prices exclude VAT, NHIL, GETFund and COVID levies unless expressly stated.
18.3 Payment milestones. Mobilisation on contract signature; a second instalment on completion of hardware build and platform deployment; a third on completion of field installation; and a final retention released on acceptance against the Section 13 metrics. Exact percentages are agreed at contract.
18.4 Change control. Any variation to scope, quantity or specification is raised in writing, priced, and agreed by both parties before work proceeds. No verbal instruction alters scope or price.
18.5 Warranty and defects liability. Twelve months on supplied hardware from the date of installation, covering manufacturing defect. Excludes lightning strike, vandalism, third-party damage and consequential loss. Platform defects are corrected under the support agreement at no additional charge.
18.6 Support and service levels. Severity-based response is defined in the support agreement. Availability is carried as an acceptance metric in Section 13 rather than as an uncosted promise.
18.7 Intellectual property. POA Nexgen retains ownership of the GridWatch platform and grants a perpetual, non-exclusive licence for the deployed instance. All telemetry, asset data and reports generated by the deployment belong to the Commission. Source code escrow is available on request.
18.8 Data protection. POA Nexgen acts as data processor; the Commission is data controller. Processing is limited to the documented purpose, with a defined retention schedule and a data-return-and-deletion obligation at termination.
18.9 Confidentiality. Each party treats the other's commercial and technical information as confidential and does not disclose it to third parties without written consent, save where disclosure is required by law or by the Auditor-General.
18.10 Termination. Either party may terminate for material breach unremedied within 30 days of written notice. On termination the Commission retains all data and any installed hardware for which it has paid.
Provided so that this document can be read by procurement, finance and legal reviewers as well as by engineers. Terms appear in the order a reader is likely to meet them.
Acceptance below indicates agreement to proceed to scoping and firm quotation on the basis set out in this document, including the procurement pathway in Section 12 and the disclosure it contains. It does not of itself constitute a contract, and no work commences until a signed contract and mobilisation instruction are in place.
The most useful next conversation is a working session with the Commission's technical directorate alongside ECG, NEDCo and GRIDCo engineers. We demonstrate the live platform, and we test the assumptions in Section 13 against your operational data — several can be closed in that one meeting. A firm quotation follows within five working days of agreed scope.