Data is rarely perfect. The analysis still has to work.
In the real world, data is rarely available in the quantity and quality required. It is incomplete, inconsistent, sparse or missing. Onsieve builds the analysis around that reality, reconstructing what is missing and showing, step by step, how every value was obtained.
Select a step, or let it play.
The client receives a detailed explanation of how the data was constructed, generated and rebuilt.
Each data point is identified as measured, reconstructed or estimated, so nothing is presented as better than it is.
The method behind every gap is documented: the engineering model, the physical relationship, the reference data or the machine learning model used, with the assumptions made.
Reconstructed data is checked against the measurements that exist, and the uncertainty is reported together with the result.
Machine learning is used where it improves on conventional models, to recreate information that was never recorded.
The monthly energy of a plant becomes a realistic hourly generation profile that keeps the measured monthly totals.
Profiles at minute resolution are rebuilt from the few measurements taken during the year.
The inertia and the stability of a power network are assessed even where measurements are scarce.
The hourly electricity load of a whole region, as the basis for sizing generation and storage. See the load forecasting method.
Anomalies, sensor drift and inconsistencies are detected and flagged, so the analysis starts from data that can be trusted.
Once the data chain is built, the analysis and the reports can be repeated automatically as new data arrives.
From a business question to a decision supported by evidence.
The decision the analysis must support: an acquisition, a maintenance plan, a capacity increase or an investment.
Operating data, maintenance records and design documents are collected and checked for gaps, errors and sensor drift.
Engineering and optimization models are built for the specific asset. Machine learning is added where the data supports it.
Results are presented with their assumptions, uncertainty and supporting data, so the decision can be reviewed and challenged.
Send a short description of the asset, the data available and the question to be answered. The scope is defined from there.