What utilities ask us first

EUNIQ is an enterprise AI intelligence platform for energy and utility networks. It unifies meter, network, billing and service data into a single topology-aware model, then forecasts demand, detects anomalies and scores asset risk. EUNIQ stands for Energy & Utility Network Intelligence Quotient, and it covers electricity, gas and water distribution on one foundation.

A dashboard reports what happened. EUNIQ models the network itself. The substation, feeder, transformer and consumer hierarchy is a first-class object, so every signal is interpreted in the context of the asset above it. That is what allows a statistical anomaly to be attributed to a physical cause rather than simply flagged. It also pairs machine learning with the engineering methods utilities already trust — distribution state estimation, load flow analysis, energy balance, minimum night flow — so a model output can be reconciled against physics.

Three things are essential: interval load data at feeder and transformer level (15-minute or hourly, ideally 24–36 months of history), an asset master with the network hierarchy and rated capacities, and the utility's own threshold definitions for high loading and abnormal conditions. Outage logs, load-regulation schedules, calendar data and weather improve accuracy and can follow. Neither forecasting nor anomaly detection requires consumer-identifying information — anonymised or pseudonymised identifiers are sufficient.

That is the case it is designed for. Most platforms assume a clean, fully instrumented network with complete asset registers and source systems that agree with each other. Real distribution networks are mid-rollout on smart metering, carry incomplete registers, and hold conflicting consumer-to-transformer mappings across billing, GIS and metering systems. EUNIQ includes entity resolution and topology reconstruction specifically to work with data in that condition.

LSTM and GRU recurrent networks for short-term load and demand forecasting; LSTM autoencoders for unsupervised anomaly detection, chosen because confirmed and dated asset failure records are scarce in every utility; graph methods including GNN, GraphSAGE and probabilistic matching for topology and identity resolution; and classical baselines such as XGBoost, SARIMA and Kalman filtering to keep every deep model honest. Explainability uses SHAP, LIME and attention attribution so every alert ships with the signals that produced it.

Yes. EUNIQ is cloud-native by design and deployable inside a utility's own estate where regulation, data residency or critical-infrastructure policy requires it. Integration follows international standards — CIM IEC 61968 / 61970 for network and asset data, DLMS / COSEM for metering, MQTT and OPC-UA for telemetry, with open REST and event APIs throughout.

It reports probable contributing factors with the evidence behind each, ranked and scored. It does not present them as confirmed causes. A confirmed cause requires field validation, which only the utility can supply. Those inspection outcomes are captured back into the platform as ground truth, so accuracy improves with each cycle on the utility's own network. In a regulated business where a flag becomes an inspection and possibly a penalty, that distinction is not optional.

We do not quote an accuracy figure before seeing a utility's data. Achievable accuracy depends on data completeness, meter reliability and the volatility of the load itself. The target is agreed with the utility at scope freeze after a sample extract is reviewed, and measured on a held-out period the model never sees during training, reported per asset and in aggregate using MAPE and RMSE.

Three, all available for deployment on the same foundation. EUNIQ Grid covers electricity — technical and non-technical loss, theft, outage, feeder and transformer intelligence, demand and resilience. EUNIQ Gas covers leak and pressure intelligence, demand forecasting, pipeline anomaly and unaccounted-for gas. EUNIQ Water covers non-revenue water, leakage, pressure, demand and asset intelligence. A utility can start with one domain and extend into the next without a rebuild.

Innopas Technology Solutions, with product engineering in Asia Pacific and an established practice in North America. The founding team has delivered mission-critical enterprise systems for a decade — real-time credit authorisation platforms, programmes of 2,500 engineers, and an agile engineering foundation for 30,000 people. Model selection and methodology are reviewed by doctoral advisors alongside subject-matter advisors with distribution-sector operating experience.