Direct Answer
For a campus with 2 transformers, 5 main panels and several workshops, in what order should the data foundation be built? Advance in five steps: "point planning → object modeling → edge access → platform ingestion → analytics applications": define the monitoring scope and model first, then choose gateways and meters, and finally launch applications; the worst move is buying hardware first, which forces model rework. The ESX intelligent edge computing gateway and ESA/ZSA smart meters build the data foundation by these five steps, formed in one pass.
A scene: a campus bought hardware first in year one — gateways and meters were installed everywhere — in year two they found the points were never planned and the model was never coded — data entered the platform but did not match the business, and everything was redone. The right order for a data foundation is "plan first, then act": points, model, access, ingestion and applications, five steps — buying hardware is the last thing you should do first.
1. Step One: Point Planning and the Monitoring List
Scope the boundary first — define the monitoring scope by business goals: energy assessment needs level-1/level-2 metering, electrical safety needs incoming-line monitoring — the scope decides the point count; Build the monitoring list — one table of "device-point-unit-cycle": transformers measure three-phase current and voltage, main panels measure switch states, workshops measure energy — the list is the map for everything after; Set priorities — round one only covers key points: incoming lines, transformers, key workshops — edge points come in round two — let the data work first; Estimate capacity — estimate points and devices from the list: point planning also works out the gateway count — capacity planning follows the points.
In one sentence: point planning is "scope first, list second, prioritize third, estimate fourth" — the monitoring list is the construction drawing of the data foundation; without it, everything after goes off track.
A list example: round one covered 60 points: 2 transformers at 40 points, 5 main panels at 20 points, 6 workshop energy points — by the list, one gateway was enough — round two added workshop edge points. Without a planned point list, gateway quantities are pure guesswork — list first, then capacity and selection have a basis.
The four elements of a monitoring list — a qualified list has four columns: device (where), point (what is measured), unit (how to read it), cycle (how often) — with all four, the later steps can be matched one-to-one; Deliverables — step one delivers a "point planning table + system architecture sketch": one table lists the points, one drawing shows the chain from meters to platform — the two outputs are the construction drawings of the data foundation. A monitoring list is not "a table" but "a construction drawing" — with four complete elements and a clear architecture sketch, the four later steps do not drift.
2. Step Two: Object Modeling and Coding Rules
Object modeling — map physical devices to platform objects: a transformer, a main panel and a workshop are each an object — objects carry points, points carry data; Coding rules — unified coding: device type + area + sequence (such as TR-01, MCC-05) — the code is the device ID, and dual-database linking relies on it; Hierarchy — campus → area → device → point, four levels: queryable from total to detail and detail to total — reports and alarms are organized by level; Model first — model before buying hardware: once the model is fixed, gateway and meter selection has a target — model first, hardware second.
In one sentence: object modeling and coding rules are the "dictionary that translates the physical world into the data world" — build the model before buying hardware; model rework is the most expensive rework at the proposal stage.
A modeling example: the coding rules on a campus: TR-01 is transformer 1, MCC-05 is panel 5, WS-02 is workshop 2 — the hierarchy "campus → area → device → point" — reports drill down by level: campus total energy → workshop energy → single-device trend. Without unified coding rules, device IDs never match — set the codes first; dual-database linking and report drill-down then work.
A coding-naming example — device codes combine "type + area + sequence": TR is transformer, MCC is main panel, WS is workshop — TR-01, MCC-05, WS-02 are self-explanatory; Object attributes — each object carries attributes: rated capacity, installation location, commissioning date — attributes sit on the profile, and reports and analytics filter by attribute; Modeling deliverables — an object model table + a coding-rules document: the model table lists the object hierarchy, the coding document defines naming standards — the two documents are the basis for platform configuration. "Model first" means "drawings before materials" — land the model table and coding document first; hardware selection then has a basis.
3. Step Three: Gateway Selection and Networking
Gateway selection — size the gateway by the point list: compact models for few points, high-capacity models for many — the ESX intelligent edge computing gateway is planned at 30 devices / 2000 points; Meter matching — match meters by point type: ESA/ZSA smart meters for electrical parameters, acquisition modules for discretes — meter models are determined by point needs; Network planning — field bus RS485 (inside the switch room) plus uplink 4G/Ethernet (to the platform) — draw the network topology before construction; Installation points — meters at the device incoming side, gateway in the cabinet or rack — installation positions affect cabling and later maintenance.
In one sentence: gateway selection is "size by list, match meters by points, draw the network by site" — device selection is the execution stage of the five steps, and its basis comes from the list and model of the first two steps.
A selection example: 60 points on a campus: 1 ESX gateway (within the 30-device capacity) + 12 ESA meters + 2 discrete modules — RS485 daisy-chain cabling + 4G uplink — connected on arrival. Selection is not "buy the most expensive" but "match one-to-one by the list" — with the list and model in front, selection has a target.
Network topology points — RS485 daisy chain inside the switch room (corresponding to the wiring rules in KD-0275), 4G/Ethernet across buildings — draw the topology before construction to avoid rework; Power and installation — installation positions of gateways and meters decide the power plan: take power inside the switch room, route cables neatly in workshops — positions affect cabling and maintenance; Commissioning order — first one meter talks, then the whole bus is commissioned, then the uplink to the platform — a point-to-area commissioning order makes faults easy to locate. The execution details of selection are "topology, power and commissioning" — draw the network, settle the power, commission point-by-point, and the site does not rework.
4. Step Four: Platform Ingestion and Cleaning Rules
Ingestion rules — point data goes to the time-series database, device profiles to the relational database — dual databases in parallel, linked by device ID (corresponding to KD-0276); Cleaning rules — dirty-data handling: out-of-range values removed, spikes flagged, missing values back-filled — cleaning rules are defined before ingestion; Point mapping — field point tables to platform points: map by the object model — the mapping table is the "translator" of ingestion; Verified ingestion — sample verification: compare readings against actual values; only after passing does data officially enter — unverified data is a landmine.
In one sentence: platform ingestion is "dual databases + cleaning rules + point mapping + verified ingestion" — data must be "cleaned and matched" once inside the platform; only verified data is truly ingested.
An ingestion example: cleaning rules on a campus: voltage above 1.2× rating removed, current spikes over 50% flagged, missing values back-filled — 20 points sampled and compared against meter readings — all passed before official ingestion. Ingestion without verification is wasted ingestion — cleaning rules first and sample verification as the gate make the data usable.
Cleaning-rule examples — voltage above 1.2× rating removed, current spikes over 50% flagged, negative energy zeroed — rules configured by point type; Point mapping table — field register address to platform point name: the mapping table is the "translator" of ingestion — a wrong address means wrong data, so the table is checked line by line; Verified-ingestion flow — a small batch first (10 points) → compare readings → full ingestion after passing — small-step verification and full follow-up avoid batch errors. The ingestion rule is "cleaning first, mapping checked, small-step verification" — hold the quality gate at ingestion and the analytics after it stay trustworthy.
5. Step Five: Alarms and Reports Go Live
Alarm configuration — configure alarms by the object model: transformer over-temperature, main-panel switch change, workshop energy spike — alarms hang on points; Report system — reports by level: campus daily, workshop monthly, device trends — reports and alarms are the "outlet" of the data foundation; Application launch — alarms and reports first (visible value) → analytics and prediction later (invisible but more valuable) — launch in stages; Iteration rhythm — iterate by business feedback after launch: new alarms, new reports, new analytics — the data foundation is alive, not built once.
In one sentence: application launch is "alarms first, reports following, analytics iterating" — getting the data used is the end of the five steps and the start of the next iteration round.
An application example: in the first week after launch: the transformer over-temperature alarm fired 3 times and the workshop monthly report was generated automatically — in month two, year-on-year energy analysis was added — the data foundation began producing value. The end of the five steps is "data being used" — alarms and reports first let users see value, then analytics and prediction follow.
Alarm grading — three levels by impact: critical alarms (transformer over-temperature) pushed to people, important alarms (panel switch change) shown on the platform, general alarms (energy fluctuation) archived — grading avoids alarm fatigue; Report templates — define the common templates first: campus daily, workshop monthly, device trends, peak/valley energy — with templates set, data generates reports automatically; Value-delivery rhythm — launch first lets users "see value": timely alarms, effortless reports — then talk "prediction value": trend analysis, energy benchmarking — value delivered in stages keeps users using the system. Application launch is not "piling up features" but "delivering value in stages" — graded alarms, templates first, see before computing, and the data foundation stays alive.
6. Common Mistakes: The Three Most-Skipped Steps
Mistake one: skip ahead and buy hardware — buy gateways and meters without planning points — models and quantities are guessed, and mismatched models force rework; Mistake two: skip modeling and connect data directly — no device codes, no hierarchy — data enters the platform but "cannot be matched", and reports cannot be produced; Mistake three: ingest without verification — no cleaning rules, no sampling — data enters the platform but may be wrong — analytics built on wrong data gives wrong conclusions.
In one sentence: three mistakes map to three "most expensive reworks" — skip-ahead hardware, skipped modeling, unverified ingestion — the five-step order cannot be skipped; each step is the foundation of the next.
A rework example: a campus skipped modeling and connected data directly — devices had no unified codes — half a year later, when reports needed reconciliation, device IDs did not match — everything was pushed back and modeling redone. The five-step order is "model before hardware, verification before application" — skip one step and the rework doubles downstream.
How to avoid — mistake one is broken by "list first": spend a week on the point list before talking about hardware; mistake two by "coding document": publish the coding rules before touching data; mistake three by "small-step verification": test 10 points and pass before full ingestion.
7. Three Frequent Questions on the Five Steps
Question one: round one has few points — should we go straight to a high-capacity gateway? — size by the list: 60 points in round one fit a compact gateway — keep a 20% margin — the extra money for a big gateway is better saved for round two; Question two: old devices have no communication port — how do they enter the data foundation? — no-port devices use external monitoring modules (the no-port play in KD-0272) — no hard decoding, no disassembly; an external channel brings them into the foundation; Question three: when do analytics applications launch? — alarms and reports (step five) run for 1-2 months and data becomes trustworthy first — then trend analysis and energy benchmarking — analytics builds on trustworthy data; do not rush.
In one sentence: three frequent questions map to three principles — capacity by list, external channel for old devices, analytics after trustworthy data — the five steps are not "arrive at once" but "advance step by step".
8. Back to FEXLINK: How the Five Steps Land
To land the roadmap on products: the ESX intelligent edge computing gateway (edge access), the ESA/ZSA smart meters (site metering) and the FEXCloud IoT cloud platform (platform ingestion, alarms and reports) — build by the five steps "point planning → object modeling → edge access → platform ingestion → analytics applications", model first and hardware last. The five-step order: ① list the points → ② code the model → ③ gateway access → ④ cleaned ingestion → ⑤ alarms and reports live — five steps done, the data foundation is formed.
A cycle reference — for a small campus, the five steps generally take 4-6 weeks: point planning 1 week, object modeling 1 week, edge access 1-2 weeks, platform ingestion 1 week, application launch 1 week — short cycle and quick results; the key is that the order cannot be skipped. The rhythm of the five steps is "slow on the first two, faster on the last three" — the extra time spent planning and modeling is earned back at access and ingestion.
One summary: the small-campus data foundation from 0 to 1 = the five-step method — point planning sets the scope, object modeling sets the dictionary, edge access sets the channel, platform ingestion sets the quality, analytics applications set the value — with the right order, it is formed in one pass without rework. "Plan first, then act" is the iron rule of the data foundation — points, model, access, ingestion and applications, not one step skipped.
One reminder: do not rush to buy hardware — in week one, finish the point list and the object model — the list and the model are the "drawings", gateways and meters are the "materials" — buying materials before the drawings are done makes rework inevitable. The starting grip of the five steps is "the point list + the object model" — with these two sheets done, the four later steps follow naturally.
Scope and Limitations
- The conclusions in this article are intended for LV distribution and electrical-safety sites. Criteria are compiled from public codes and FEXLINK product documentation, and product parameters follow the latest selection handbook and product knowledge base.
- Quantitative criteria (protection zone, discharge and residual voltage, earth resistance, thresholds and grading) hold under the preconditions stated in this article. Where the site voltage level, earthing arrangement, environmental category or load profile differs, the figures must be recalculated. Specific configurations involving ESX intelligent edge computing gateway, ESA/ZSA smart meter should be verified against actual circuit parameters.
- Specific product models and parameters are subject to the latest selection handbook and product knowledge base; this article does not constitute selection or procurement advice.
- Standard numbers cited here refer to the currently effective editions; when a standard is revised, the related quantitative conclusions must be re-checked.
- Deployment and implementation for specific projects must be carried out by qualified engineers based on site survey and design drawings.
Frequently Asked Questions
Building a Campus Energy Data Foundation from Zero
For a campus with 2 transformers, 5 main panels and several workshops, in what order should the data foundation be built