By Gleb Tsipursky, Ph.D.
Nigeria’s power sector is under pressure to become more visible, more responsive, and more accountable at the same time. Digital monitoring can help. But the more the grid depends on automated alerts, dashboards, and eventually AI-assisted analysis, the more important it becomes to know who owns the conclusion when something goes wrong.
That question became concrete this week. Recall that Energy Worth Online reported on August 27 that the Transmission Company of Nigeria (TCN) disputed Abuja Electricity Distribution Company’s (AEDC) explanation for a major reduction in power available to Abuja. AEDC had attributed the decline to transmission-line constraints, while TCN said low generation was the real cause.
The important point is not which organization ultimately proves correct. It is that modern power systems generate many possible explanations for the same operational failure. More data can narrow the possibilities, but it can also multiply dashboards, alerts, and competing interpretations.
Nigeria therefore needs a failure-attribution ledger as the power sector digitizes.
A failure-attribution ledger is a simple record attached to consequential grid incidents. It should show what happened, what the monitoring system detected, which data sources were used, who made the diagnosis, which alternatives were considered, what action followed, and what evidence later confirmed or changed the explanation.
•Gleb Tsipursky, Ph.D
This matters even before advanced AI becomes common in grid operations. Energy Worth Online recently covered concerns from generation companies about the rollout of SCADA, the supervisory control and data acquisition technology intended to improve real-time monitoring, operational visibility, and grid reliability. The generation companies did not dispute the value of the system. Their concern was that practical implementation conditions still mattered.
That is the right frame for AI as well. Better prediction does not eliminate the need for ownership. Faster anomaly detection does not answer who decides what the anomaly means. A model can identify correlations in generation, transmission, weather, load, or equipment behavior, but a consequential operational judgment still needs an accountable owner.
A useful failure-attribution ledger would contain six fields.
First, record the incident. What changed in supply, frequency, voltage, equipment condition, or another operational measure, and when did it happen?
Second, record the evidence. Which sensors, operational logs, maintenance records, market data, or system alerts informed the diagnosis? This prevents a later explanation from becoming detached from the information actually available at the time.
Third, name the diagnostic owner. Every consequential conclusion should belong to a defined role or team. “The system said so” is not sufficient when the result affects customers, dispatch decisions, maintenance priorities, or public explanations.
Fourth, record competing hypotheses. If low generation, a transmission constraint, equipment failure, data error, or another cause could plausibly explain the same event, the ledger should show which alternatives were examined and why one explanation was favored.
Fifth, record the operational response. Did the organization redispatch power, inspect equipment, change a constraint, communicate with a distribution company, or take another action? A diagnosis matters because of what it causes people to do.
Sixth, verify the attribution after the event. Once more data becomes available, did the original explanation hold up? If not, the organization should record what changed and what the monitoring process needs to learn.
This structure becomes more valuable as AI enters operational analytics. AI systems are good at finding patterns across large volumes of data, but they can produce false positives, overweight incomplete data, or offer an answer that appears more certain than the underlying evidence deserves. A ledger creates a disciplined place for human reviewers to test those outputs rather than simply accept or ignore them.
The psychological benefit matters too. Engineers and operators are more likely to use new technology responsibly when they know its boundaries and understand how accountability works. If every error becomes an argument over whether the human or the technology was to blame, employees will learn to protect themselves. Some will over-trust automation because it diffuses responsibility. Others will resist useful tools because they fear being held responsible for outputs they did not control.
A failure-attribution ledger changes that incentive. It makes the decision process visible. It tells people that using a digital tool does not erase professional judgment, and that challenging an automated conclusion is part of the workflow when the evidence warrants it.
Nigeria’s power ambitions make this especially relevant. The Federal Government said on August 26 that it wants the system to reliably wheel 6,500MW by the end of 2026 and 8,000MW by the end of 2027. More capacity will increase the value of reliable operational information and fast coordination across generation, transmission, and distribution.
The sector should measure digital modernization accordingly. The metric should not be how many control rooms have new dashboards or how many alerts an algorithm produces. It should be whether operators identify failures faster, reach better-supported conclusions, coordinate responses more effectively, and reduce the time between an incident and verified restoration.
Nigeria can build a more digital grid without making responsibility more opaque. The practical test is simple: when two systems or two institutions disagree about why the lights went out, can the sector reconstruct the evidence, the decision, and the accountable owner?
If it can, AI can become a useful layer of operational intelligence. If it cannot, adding more automation may simply make old accountability gaps harder to see.
•Gleb Tsipursky, Ph.D., is a behavioral scientist and CEO of Disaster Avoidance Experts.
