Home/Insights
InsightsField notes and engineering deep-dives from real projects: PV forecasting, SCADA telemetry, meter normalisation, market-data pipelines and grid-system integration. Straight from the systems we build and keep running in production.
Irradiance is the easy part. We break down where day-ahead PV forecasts really go wrong — clear-sky index, soiling, inverter clipping — and which features earn their place in the model.
Read article →From Modbus and OPC UA tags to durable, queryable history. How we buffer high-frequency plant telemetry, handle backfill and gaps, and keep the TSDB fast under continuous writes.
Read article →Every meter and inverter speaks its own dialect of units, timestamps and registers. The canonical schema and mapping layer we use to make multi-vendor fleets look like one clean dataset.
Read article →Day-ahead and intraday prices, fetched, validated and stored so forecasting and trading logic can rely on them. Handling late publications, revisions and the gaps nobody warns you about.
Read article →Setpoints, alarms and state changes don't fit a request/response world. Why we lean on events and a message bus to connect SCADA, ERP and control logic without brittle point-to-point links.
Read article →Most energy data initiatives die between proof of concept and production. The integration debt, ownership gaps and missing operational hooks that kill them — and how we get past each one.
Read article →When to read plant data over Modbus versus OPC UA, and how to migrate without a rip-and-replace.
Read article →When high-frequency telemetry needs a TSDB for compression, aggregates and retention — and where relational still belongs.
Read article →A clear-sky model is the reference that turns raw PV output into a number you can compare across hours, sites and seasons.
Read article →Read-only, outbound-only connectors and an OT DMZ get plant data out without exposing SCADA or PLCs.
Read article →Edge buffering, idempotent backfill and quality flags that close telemetry gaps without poisoning models or reports.
Read article →Separating grid- and market-driven curtailment from real underperformance, and why mixing them breaks your KPIs and forecasts.
Read article →From raw alarm floods to deduplicated, prioritised events on a bus, without the fatigue.
Read article →Why SCADA energy never ties out with settlement volumes, and how a reconciliation layer with tolerance bands handles the gap.
Read article →A local queue at the plant so SCADA and metering data survives backhaul outages and replays cleanly.
Read article →Reports get ignored when the numbers disagree with the meter. Lineage, confidence flags and reconciliation fix that.
Read article →No newsletter noise — just new engineering write-ups and field notes from building energy software, sent when they go live.