Direct Answer

A 50,000-square-meter campus wants energy-and-carbon digitalization, but the first phase has a limited budget. How should the sensing, edge and platform layers be built without rework? Phase one should solidify full-loop metering, critical-cabinet temperature sensing and edge gateways (200 to 300 points), with the platform limited to basic dashboards and alarms; phase two adds the three AI engines and the digital twin; phase three's energy-carbon autonomy grows naturally from the accumulated data. The key is to reserve the object model and point numbering in phase one — numbering decides everything, and phase two has zero rework.

A scenario: A 50,000-square-meter campus has an 800,000 RMB phase-one budget — the planner says "install 500 points first and make the big screen dazzling" — phase one overspends, the points are too dense and nobody uses the screen — when phase two wants AI, the point numbering is chaotic and there is no object model — so everything is torn down. The right way to build the data foundation is solid sensing in phase one, smart analysis in phase two and autonomous operation in phase three — fix the numbering and object model in phase one, and phases two and three never rework.

1. The Three-Layer Architecture: Sensing, Edge, Platform

Layer one, the sensing layer — metering and monitoring: meters, temperature and leakage — the source of data; Layer two, the edge layer — edge gateways: aggregation, computation, data safe when offline; Layer three, the platform layer — FEXCloud: dashboards, alarms and analysis — the entry to decisions; The division of the three layers — sensing produces data, edge protects data, platform consumes data; The order of the three layers — sensing before platform: without data, the platform is an empty shell.

In one sentence: the three layers are sensing (produces), edge (protects) and platform (consumes) — sensing before platform — without data, the platform is an empty shell.

A layer example: One campus plans three phases: phase one senses (full-loop metering plus critical-cabinet temperature), phase two adds edge (edge computing plus linkage strategies), phase three adds platform (AI engines plus digital twin) — each phase consumes the previous one. The logic of the three layers is layer-by-layer consumption — the platform consumes edge data, the edge consumes sensing data — when the lower layer is not solid, the upper layer is a castle in the air.

2. The Phase-One Checklist: Four Categories Done Solidly

ESA all-element smart meter — full-loop metering: energy, power and harmonics in one meter; EST multi-channel temperature controller — critical-cabinet sensing: busbar, joint and cable hot spots; ESX edge gateway — edge aggregation: local processing and offline caching; FEXCloud platform — basic dashboards and alarms: usage, trends and alarms visualized; The phase-one principle — fewer but solid: 200 to 300 points done well, not 500 spread thin.

In one sentence: the phase-one four-piece set is ESA metering, EST temperature, ESX aggregation and FEXCloud dashboards — fewer but solid.

A checklist example: One campus phase one: ESA on 13 loops, EST on 8 cabinets, two ESX gateways and the platform's basic modules — landing at 800,000 RMB — after six months, the data completeness rate is above 99%. The point of phase one is coverage of the critical and quality first — fewer but solid, every point must measure accurately, transmit reliably and be visible — density is less valuable than solidity.

3. Phase-One Point Design: How to Choose 200 to 300 Points

The selection principle — cover critical loads: loops carrying 80% of usage must be installed; Metering points — inlets plus key loops: sub-metering complete, the usage structure clear; Temperature points — critical-cabinet busbars and joints: high heat-risk points installed; Leakage points — damp and aging areas: insulation hazard-prone zones covered; The boundary of points — no full coverage pursued: phase two adds as needed, phase one keeps quality.

In one sentence: selection means covering critical loads, key-loop metering and high-heat temperature points — 200 to 300 points keep quality, not full coverage.

A selection example: One campus phase one: 4 inlet loops plus 9 key loops with ESA (13 metering points), 8 critical cabinets with EST (64 temperature points) and leakage control in the charging area — 230 points in total — covering 82% of usage. The logic of selection speaks in usage share and risk level — loops covering 80% of usage plus high-risk cabinets — 200 to 300 points in phase one gives the best value.

4. The Role of the Edge Gateway: Data Safe When Offline

The gateway's duty — aggregating sensing data: unified access for multiple device categories; The gateway's computing — local processing: cleaning, caching and preliminary calculation; Data safe when offline — offline caching: after the network recovers, data is re-uploaded; The gateway's deployment — distributed by zone: near distribution rooms and machine rooms; The gateway's value — data without loss: network fluctuation does not hurt completeness.

In one sentence: the edge gateway aggregates, caches and computes locally — data safe when offline — completeness does not depend on network stability.

A gateway example: One campus's optical cable maintenance cuts the network for six hours: the ESX gateway caches all metering data and auto-uploads after recovery — month-end reconciliation has zero gaps. The gateway's logic puts network fluctuation outside the data chain — sensing data is stored at the edge first — the network recovers, the data completes — completeness never bets on the network.

5. Platform Phase One: Basic Dashboards and Alarms, Not Greed

The phase-one platform — basic dashboards: usage, trends and sub-meter shares; The phase-one alarms — threshold alarms: over-limit alarms that can be located; The phase-one boundary — no AI: analysis engines come in phase two; The phase-one focus — complete data into storage: acquisition, cleaning and storage closed; The phase-one value — make data usable first: business gets used to reading numbers.

In one sentence: platform phase one is basic dashboards plus threshold alarms with complete data into storage — AI analysis comes in phase two.

A platform example: One campus's phase-one platform shows sub-meter usage on dashboards and pushes alarm lists to phones — operations start reading data daily — in three months they proactively find two abnormal usage points. The phase-one platform's logic is to let data flow first — dashboards are the entry, alarms are the grip — once business uses it, phase two has a foundation for AI.

6. Point Numbering and the Object Model: Do It in Phase One, Never Rework

Point numbering — globally unique: loop, device and measurement point map one to one; The object model — structured description: device type, location and hierarchy; Do it in phase one — fix the numbering standard: coding rules and naming standards written into documents; The phase-two benefit — new points extend the numbering: AI and the twin consume data directly; The cost of rework — chaotic numbering: phase two rebuilds the whole database, history data invalidated.

In one sentence: numbering and the object model are globally unique and structured — the standard is fixed in phase one — phase two extends with zero rework.

A numbering example: One campus's coding standard is six levels: plot, building, floor, loop, device and point — for example B3-F2-R05-CT01-T — the 230 phase-one points are numbered by the standard — 120 phase-two points extend it — the object model is reused directly. The logic of numbering is that in the data world, a name is an address — with good numbering, AI, the twin and reports all consume directly — chaotic numbering makes data an island however much there is.

7. Phase Two Upgrade: The Three AI Engines and Analysis

Phase two's role — from seeing to understanding: data analysis and anomaly recognition; Engine one, Tianyan — trend warning: abnormal energy, load forecasting; Engine two, Qianzhi — alarm governance: false-alarm compression, root-cause location; Engine three, Wanxiang — knowledge graph: equipment relations, impact analysis; The phase-two premise — phase-one data complete: AI's fuel is history data.

In one sentence: phase-two AI is Tianyan (trend), Qianzhi (alarm governance) and Wanxiang (knowledge graph) — the fuel is phase-one complete data.

A phase-two example: One campus launches the three engines: Tianyan finds abnormal night usage in one building, Qianzhi compresses alarms by 80%, Wanxiang locates the air-conditioning units — after handling, usage falls. Phase two's logic is that phase one produces data and phase two consumes it — complete data is AI's fuel — an unsolid phase one makes phase two spin in the air.

8. Phase Three: Digital Twin and Energy-Carbon Autonomy

Phase three's role — from understanding to autonomy: closed-loop control and optimized operation; The digital twin — the equipment map: alarm location and scene linkage; Energy-carbon autonomy — automatic optimization: load adjustment and efficiency improvement; The phase-three premise — phase-one and phase-two data and models: the twin and autonomy consume all assets; The phase-three effect — continuous efficiency improvement: energy costs fall month by month.

In one sentence: phase three is the digital twin (location) plus energy-carbon autonomy (optimization), consuming all phase-one and phase-two assets — continuous efficiency improvement.

A phase-three example: One campus: the twin locates alarms and autonomy automatically adjusts air conditioning and lighting — energy costs fall 12% year on year. Phase three's logic is that data assets begin to appreciate themselves — the twin and autonomy grow naturally from the data foundation — phases one and two lay the ground, phase three harvests.

9. The Baseline Standard: GB 51348-2019 and Monitoring Points

GB 51348-2019 — the civil-building electrical design standard: the basic baseline of electrical system design; What the standard means — design with evidence: loop, protection and monitoring configurations follow the code; Its relation to the data foundation — monitoring points obey the electrical design: no interference with operation, no harm to safety; The standard's landing — compliant implementation: adding points does not change the original protection logic; The standard's value — the rework firewall: compliant design makes acceptance and renovation smooth.

In one sentence: GB 51348-2019 is the electrical-design baseline — adding monitoring points must not change the protection logic — compliant design, no rework in renovation.

A standard example: One campus's point design: monitoring installation avoids protective devices and never taps the main loop — reviewed and passed against GB 51348-2019 — later capacity upgrades have zero rework. The standard's logic is that monitoring is an add-on, not a modification — do not touch the original protection logic or violate the electrical design code — a compliant design makes every future renovation smooth.

10. Phase-One Investment Scale Reference

The investment composition — sensing 60%, edge 15%, platform 25%: sensing is the main body; The scale reference — a 50,000-square-meter campus: phase one in the 600,000 to 1,000,000 RMB range; The per-point cost — metering and temperature points: roughly 3,000 to 5,000 RMB each overall; The platform share — basic modules: about a quarter of the total investment; The investment order — sensing first: make data production solid before anything else.

In one sentence: the phase-one scale is sensing 60%, edge 15%, platform 25% — 600,000 to 1,000,000 RMB for 50,000 square meters — sensing first.

A scale example: One campus's phase-one final accounts were 830,000 RMB: sensing 500,000 (ESA plus EST), edge 130,000 (two ESX), platform 200,000 (FEXCloud basic) — matching the 6/1.5/2.5 composition. The investment logic is that data production is the main body — the sensing layer is the source of data and holds the largest share — edge and platform are supporting and need to be sufficient, not lavish.

11. The Phase-Two Upgrade Path and Migration

The upgrade path — add points plus engines: extend the numbering, open new platform modules; The migration guarantee — the object model reused: new assets connect automatically; Data without interruption — smooth platform upgrade: history data fully preserved; Team preparation — a data owner: phase two needs an operations role; The upgrade pitfall — plan ahead: avoid "phase two tears down phase one".

In one sentence: the phase-two upgrade adds points and engines — numbering extends, the model reuses and data is preserved — plan ahead, no tearing down phase one.

An upgrade example: One campus phase two: 120 new points extend the numbering and the platform opens the three-engine module — completed in two weeks with no history loss — phase-one assets wasted nothing. The upgrade logic is that phase one is a foundation, not a burden — with numbering and the model fixed, phase two is an addition, not a rebuild — a solid foundation makes additions the cheapest.

12. Back to YESLON: The Three-Layer Product Combination

The sensing layer — ESA meters, EST temperature and leakage control: the source of data; The edge layer — ESX edge gateways: aggregation, caching and local computing; The platform layer — FEXCloud plus the Taiyi hub: dashboards, alarms and AI analysis; The solution foundation — the Digital Electrical Intelligent Supervision System: three layers delivered as one; The combination's effect — all three layers built from zero to one: phase one without rework, phase two upgradeable.

In one sentence: the three-layer combination is ESA/EST (sensing), ESX (edge) and FEXCloud/Taiyi (platform) — the Digital Electrical Intelligent Supervision System delivered as one — from zero to one without rework.

A combination example: One campus uses the Digital Electrical Intelligent Supervision System: ESA for full-loop metering, EST for critical-cabinet temperature, ESX for edge aggregation and FEXCloud for basic dashboards — 230 points land in phase one and the three engines are smoothly added in phase two. The solution's logic is one system and one plan — sensing, edge and platform come from the same system — numbering and the model are naturally unified, and phase two upgrades without friction.

13. Implementation Self-Check: Five Questions to Verify the Three Layers

Question one, is sensing solid — point quality: critical loads covered, completeness meeting targets; Question two, does edge protect data — offline caching: data not lost during network interruption; Question three, is the platform usable — dashboards and alarms: does business use it daily; Question four, is numbering standard — point uniqueness: can new points extend seamlessly; Question five, is phase two easy to upgrade — path reserved: are engines added with zero rework.

In one sentence: the five questions are sensing solid, edge protecting, platform usable, numbering standard and phase two upgradeable — all five pass and the foundation is qualified.

A self-check example: One campus self-checks: sensing completeness 99% (solid), six offline hours with zero loss (protecting), operations reading daily (usable), 230 points numbered by standard (unique), phase two completed in two weeks (upgradeable) — all five pass. The success of the data foundation is how smoothly phase one connects to phase two — solid sensing, protected edge, usable platform, standard numbering and smooth upgrades — from zero to one without rework, from one to N without strain.

Scope and Limitations

  • The conclusions in this article are intended for LV distribution and electrical-safety sites. Standard numbers (GB 51348-2019) refer to their currently effective editions, and product parameters follow the latest FEXLINK 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 ESA all-element smart meter, EST multi-channel temperature controller, ESX edge gateway 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 Mid-Size Campus Energy Data Foundation

A 50,000-square-meter campus wants energy-and-carbon digitalization, but the first phase has a limited budget