
Every plant gets an energy bill: one number for the whole site, weeks late. Ask which line, shift or product drove last month's increase, and the room goes quiet. The data usually exists. It just never makes it from the panel to the people who could act on it. Here's the architecture we use at Crescenza Consulting Group to fix that, from the meter on the machine to analytics on a central server.
Why plants can't see their energy
Meters sit in the switchgear, PLCs count every part, and utility systems run on their own controllers. Each piece works on its own, but none of it is tied to production. So nobody can say what a part cost in energy, which machine runs when the line is down, or what tomorrow's peak will be.
Our architecture, layer by layer
The design principle is simple: model the data once, as close to the machine as possible, and let every layer above reuse it.

Plant floor: meters connect to the machine's PLC over Modbus, next to machine state and part counts.
Edge: Ignition Edge on each line wraps meters and machines in UDTs, publishes them into the unified namespace (UNS), and stores data locally if the network drops.
Messaging: MQTT with Sparkplug B carries the UNS to the center and flags any device that goes offline.
Central server: one Ignition gateway and historian for all sites, running the Perspective dashboards, alerts, alarms and reports.
Analytics: AI models and Grafana read the same historian; model results flow back to Ignition as tags.
Build it once, roll it out as templates
Most energy projects stall at the second line because the tags were built by hand. We standardize three things before the first meter is wired:
One PLC block per meter type that polls the meter, scales the values, handles comms faults and keeps the energy counter.
One UNS path following the ISA-95 hierarchy, so energy and machine state always travel together:
Enterprise/SiteA/Press/Line3/Compressor01/Energy/kW
Enterprise/SiteA/Press/Line3/Compressor01/Energy/kWh
Enterprise/SiteA/Press/Line3/Compressor01/State/Running
One Edge template per line, configured from a worksheet that lists its meters and PLC addresses. The second plant becomes a configuration exercise, not a rebuild.
From visibility to AI: Three stages
We roll out in three stages, and each one has to earn trust before the next begins.
Stage 1: Visibility and reporting in Perspective

Every site starts with the same six metrics: kWh, CO2e, carbon intensity, energy mix, energy cost and energy per good part.
Stage 2: AI on the central historian
Because every value in the historian already carries its asset, line and machine state, AI works without a data-cleaning project first. Results are written back to Ignition as tags, so they appear in the same dashboards and alerts:
Expected-energy baselines per asset, so alerts fire on deviation from normal, not on fixed limits.
Anomaly detection for power spikes, rising idle load or a compressor drawing more for the same output.
Peak-demand forecasting so production can stagger start-ups before a new monthly peak.
Plain-language questions, such as why Line 3 used more energy last week, answered from the historian.
Stage 3: Control and automation
What the first month usually shows
Equipment that never turns off: compressors, chillers and fans running all weekend with the lines stopped.
Compressors feeding leaks: high air flow during shutdowns, now visible as a trend.
Shift change sets the monthly peak: everything starting at once creates the spike you pay for all month.
Twin lines that aren't twins: identical lines using different energy per part.
Lessons from the field
Check Modbus byte order once per meter model, and handle counter rollover in the PLC block.
Agree on UNS naming first. Renaming later breaks history and every model trained on it.
Calculate each KPI in one place, so Perspective, Grafana and AI never disagree.
Verify against the utility meter before anyone sees a dashboard.
Where to start
Most early savings come from running existing equipment better, not from buying new equipment. Start with one site: the main incomer and the five biggest energy users. Build the PLC block, UNS and Edge template properly, get Stage 1 trusted, then roll it out.
Crescenza Consulting Group builds this end to end in Ignition, from PLC blocks and Edge templates to the central historian, dashboards and AI analytics. If you'd like to see what your existing meters could already tell you, talk to us.
Other Stories










