A German waste-management technology company needed to see the near future of its process — the calorific and steam values its operation would produce minutes from now. We built a custom time-series model that forecasts both, 5 to 60 minutes ahead, at 95% accuracy.
The forecasting console — predicted vs actual calorific value across selectable horizons. (Interface recreation.)
Incoming waste varies constantly in composition — which makes calorific value, and therefore energy output, hard to anticipate from one moment to the next.
Knowing what's coming 5, 30 or 60 minutes ahead is the difference between reacting to a swing and steering through it — for both calorific and steam values.
The client wanted evidence, not promises. The engagement had to demonstrate real predictive power on their data before any larger commitment.
A forecasting model trained on the client's own process data, predicting calorific and steam values rather than relying on generic off-the-shelf approaches.
Forecasts across a 5–60 minute window, so operators choose the lead time that matches the decision they're making.
Alongside forecasting, the system flags abnormal patterns in the data — early warning for readings that don't fit expected behavior.
A feasibility-first, success-based structure: the model proved itself against the client's live data via API before the engagement progressed — accountability built into the commercial terms.
forecast accuracy on calorific and steam values.
of operational lead time, across selectable horizons.
Passed feasibility evaluation and progressed to a paid second phase — under success-based terms.
Novanexom took our forecasting problem seriously from day one — they proposed a feasibility-first approach, delivered on it, and were straightforward to work with throughout.
Forecasting, anomaly detection or AI automation — we'll tell you honestly if it's feasible, then prove it.
