Siren v2.4.2

Nithya Menon
March 4, 2021


Re-architected for reliability and trend analysis

Siren is designed to detect trends and highlight recurring issues, and the underlying architecture plays a big role in making this happen. To ensure reliability, Siren now has a retry mechanism to handle errors more effectively during its daily run.

At the lowest level, Siren is split into 2 tiers: the first tier processes raw sensor data from the Okra Pods and the second takes a higher-level look at which rules should be triggered for the day, and over a lookback period, to identify trends. Sometimes issues come up for a brief moment and resolve themselves, not warranting labour spent to investigate, so Siren is careful to only raise concern over issues that seem to be persisting.

Reliable blackout triggering

As Siren learns from more data and more installations, the majority of field failure modes have been captured in the algorithm. If there is a blackout due to known failure mode, Siren highlights both the failure and the underlying source.

In some cases, we haven’t yet identified why a blackout occurs but instead of letting that go unnoticed to the customer, we restructured Siren to ensure reliable electricity for the household is the highest priority. So if a household experiences recurring blackouts, and the root cause is unknown, Siren will alert the maintenance manager to investigate further. This “catch-all” also serves as a learning tool for our data team to pinpoint gaps where specific logic can be applied.

Power thief developed and maturing

In numerous markets, we received feedback that power theft is a huge concern and continual problem both on the national grid and in off-grid solutions. Siren’s “power thief” algorithm detects when there are sudden drops in sensor readings that can’t be justified, or there is an imbalance in power entering and exiting the system.

This rule continues to be fine-tuned and is currently only visible for Okra staff to scrutinize, as we need to be extremely confident of our findings before approaching a household with suspected tampering. The rule has been trained on existing confirmed cases of power theft and is steadily becoming much more reliable as we gather more field data.

Stability improvements

Bug/stability fixes in this release included:

  • Minor refinements to all sensitivity thresholds
  • All rules (other than power theft), released for external usage and action

Nithya Menon is an engineering graduate from Harvey Mudd College who has spent her career developing technology targeted towards empowering marginalized and developing communities worldwide. She has been pivotal in designing Okra's key power-sharing algorithms, IoT firmware, and grid management software - and now drives the direction and strategy of Okra's technology as Product Development Lead.