One history, three field protocols

In a mid-sized plant the instrumentation that matters is already installed. The problem is not measuring: it is that every instrument speaks its own language, and nobody ever asked them to speak to each other.

An ordinary case, three devices:

• the heat meter in the boiler room, exposing its registers over M-Bus, because that is what thermal-vector meters speak;
• the line PLC, exposing its data area over Modbus TCP, counting pieces, running hours and alarms;
• the PV inverter on the carport, speaking Modbus with a SunSpec register map — Modbus, but with a different dictionary from the PLC’s.

Three protocols, three dictionaries, three ideas of what a sample is.

Why the one-off integration does not hold

The recurring answer is a script that polls the three devices and writes to a file or a database. It holds until one of three things happens, and eventually all three do.

The field changes. A line is added, an inverter is replaced, a meter moves. Every change is a change to the script, and the script lives on a machine nobody documented.

The network tightens, rightly. The plant network should not accept inbound connections. Any integration that asks for a port open into the plant is an argument with the people who defend that network — and that argument should be lost, not won.

The link drops. A line down for two hours is not a fault, it is a Tuesday. If those samples are not held somewhere during the outage they are gone, and a hole in the history is exactly the kind of thing that surfaces when somebody has to verify that period.

How it works instead

A gateway sits in the field, inside the plant network, and does three things.

It reads locally. One container per protocol on the same box: one speaks M-Bus to the meters, one speaks Modbus to the PLC and to the inverter, each with its own map. Adding a protocol adds a container rather than rewriting the integration.

It calls out, it does not answer. The gateway opens the connection to the system, not the other way round. No inbound port is needed, which is what keeps the conversation with the network owner short.

It buffers. If the link drops, samples stay on the gateway and go up when it returns. The history closes itself instead of carrying a hole dated Tuesday.

From there provenance stops mattering. A kWh read from an M-Bus meter and a kWh read from an inverter over Modbus are two measurements in the same history, on the same time reference, queryable together.

Which protocols actually matter

It depends on the industry, and the answer is less often “Modbus” than people assume. An energy audit is per vector: electricity, heat, gas, water. Meters for the non-electrical vectors almost always speak M-Bus rather than Modbus, and without those you are measuring current only.

Which protocols each domain speaks is mapped out on the industries pages: every industry declares its own, because a page promising a reading we cannot do is worth less than a page that is not there.

The feature, with the steps, is in gateway and protocols.