Digital Energy

What the Correlation Engine V5.0 NILM Evolution Is

If a power-using device can be heard out, energy analysis need not install a separate collection device for every piece of equipment. Non-intrusive load monitoring (NILM) is exactly this idea: from the current waveform of the main circuit, identify which device is operating and how much power it consumes. Based on the wording listed in the product documentation, this article explains the NILM evolution direction of the Correlation Engine toward V5.0, the signature identification model of the Prediction Engine, its position in the selection correspondence, and its placement in the family data chain; it infers no recognition accuracy for any specific site.

2026-09-26 Digital Energy FEXLINK 7 min
Correlation Engine V5.0: Non-Intrusive Load Monitoring (NILM) Evolution
Correlation Engine V5.0: Non-Intrusive Load Monitoring (NILM) Evolution

Direct answer

If a power-using device can be heard out, energy analysis need not install a separate collection device for every piece of equipment. Non-intrusive load monitoring (NILM) is exactly this idea: from the current waveform of the main circuit, identify which device is operating and how much power it consumes. Based on the wording listed in the product documentation, this article explains the NILM evolution direction of the Correlation Engine toward V5.0, the signature identification model of the Prediction Engine, its position in the selection correspondence, and its placement in the family data chain; it infers no recognition accuracy for any specific site.

1. From V4.0 to V5.0: one line of evolution

The product documentation states that the evolution direction of the Correlation Engine is an upgrade to V5.0, with non-intrusive load monitoring (NILM) at its core; the family overview marks its version as V4.0 (V5.0 NILM evolving) and its positioning remains the situation-awareness brain, judgement layer. This wording gives both the current state and the direction.

V4.0 is the currently effective version, and V5.0 NILM evolving shows that the evolution is not yet complete. Viewing the version with the positioning makes clear the nature of this line: Correlation Engine's spatial and causal judgement capability is unchanged, and what is added is the capability to identify loads. The product documentation also states that this evolution involves two documents, the Correlation Engine Upgrade Technical Solution V5.0 and the Correlation Engine_V5_NILM 100% Advancement Plan. The existence of the documents shows that the evolution has a clear technical plan and advancement arrangement. The product documentation gives only the direction and the document names, and this article does not expand the plan content.

2. What non-intrusive load monitoring (NILM) is

Non-intrusive means there is no need to install a collector separately on every device; load monitoring means each type of device has its characteristics in the current waveform. Together, NILM means inferring the device composition from the main-port waveform.

The value of this capability is a light footprint. Traditional sub-metering requires sensors on every branch, with many points and high cost; NILM gathers the collection point to the main circuit port and replaces distributed hardware with an algorithm. The product documentation lists it as the core direction of Correlation Engine V5.0, showing that it is regarded as a key step for the judgement layer to extend to the energy-usage side. This article explains only the method idea and does not infer the device types it suits or its identification boundary.

3. Prediction Engine E-01: identification without extra hardware

The product documentation states that the signature model of the Prediction Engine is non-intrusive load identification (E-01), whose feature is that no extra hardware is needed and that the specific device is identified directly through the current waveform. The Prediction Engine handles prediction, and E-01 puts identification capability into energy analysis.

No extra hardware is the essence of this model: identification relies on existing collected data rather than new sensors. This turns the non-intrusive nature of NILM from a concept into an implementable condition. The model belongs to the Prediction Engine, showing that it shares the same data foundation with other prediction-type models; at the same time, it echoes the Correlation Engine V5.0 load-monitoring direction, forming a recognition-and-judgement pairing. The product documentation gives the model name and features, and this article does not infer its identification scope or misjudgement situation.

4. Identification basis: start-up feature, steady-state power, harmonics

The product documentation states that Prediction Engine E-01 identifies devices through a combination of start-up features, steady-state power, and harmonic features. The three classes of feature each have their meaning.

The start-up feature is the waveform change at the instant a device powers up, which differs among devices; steady-state power is the power level when a device runs steadily, used to distinguish devices with similar consumption but different nature; and the harmonic feature is the harmonic content in the current, reflecting the electrical characteristics of the device. Looking at one class of feature alone easily causes confusion, and only the combination of the three raises discrimination. This explains why NILM is a fingerprint rather than a reading: it compares the combination of waveform forms. The product documentation gives the feature classes, and this article does not expand the specific algorithms and decision thresholds.

5. Position in the selection correspondence

The product-selection and AI-capability correspondence in the product documentation maps non-intrusive load identification (NILM) to Prediction Engine E-01 plus the Correlation Engine V5.0 load monitoring. This correspondence links the capability requirement to specific models.

The reading of the correspondence table is this: if the requirement is non-intrusive load identification, the corresponding capability combination is the identification model of Prediction Engine and the load-monitoring direction of Correlation Engine. One leans toward identification and the other toward judgement, precisely covering the two segments of identifying what it is and judging accordingly. The product documentation gives the correspondence, and this article does not infer the corresponding combination of other capability items.

6. The planning scale of the energy-usage analysis section

The product documentation states that the energy-usage analysis section of the Prediction Engine plans 15 items in V2.0, of which nine are disclosed, and the first-batch P0 model includes non-intrusive load disaggregation (E-01). This section manages energy-related capabilities centrally.

Planning 15 items shows the full envisaged scale of the section; the nine disclosed are the items within it that can be explained externally; and the first-batch P0 is a batch landed first. E-01 appears both in the first-batch models and in the signature models above, showing that it is both a representative of the section and an object of priority implementation. The product documentation gives the section scale and the first-batch model, and this article does not infer the functions and schedules of the remaining models.

7. Position in the family architecture chain

The product documentation states that the data chain of the family architecture is: sensor data, the Coordination Engine back end (access for more than 40 kinds of protocols, four-level cleaning), the front-end layer (safety red-line pre-check), Diagnostic Engine (recognising things), Correlation Engine (situation awareness), Prediction Engine (prediction), the standards engine (1200 national standards), and the decision interface.

Placing NILM in this chain makes its position clear: data is accessed and cleaned through the Coordination Engine back end, passes the front-end layer pre-check, enters Diagnostic Engine for recognising things, then Correlation Engine for situation awareness, and subsequently Prediction Engine for prediction. E-01 sits in the Prediction Engine segment, and the Correlation Engine V5.0 load monitoring sits in the Correlation Engine segment, and the two are adjacent along the chain. Understanding the chain order explains why NILM needs the cooperation of several layers rather than being an isolated model. The product documentation gives the chain composition, and this article does not infer the data formats and calling methods between the layers.

Scope and limitations

First, this article restates only the wording listed in the product documentation, and its factual boundary is limited to: the evolution direction of the Correlation Engine toward V5.0, with non-intrusive load monitoring (NILM) at its core, involving the two documents Correlation Engine Upgrade Technical Solution V5.0 and Correlation Engine_V5_NILM 100% Advancement Plan; the family overview marking its version as V4.0 (V5.0 NILM evolving) and its positioning as the situation-awareness brain, judgement layer; the signature model of the Prediction Engine, non-intrusive load identification (E-01), requiring no extra hardware and identifying devices through a combination of start-up features, steady-state power, and harmonic features; the selection correspondence mapping non-intrusive load identification (NILM) to Prediction Engine E-01 plus the Correlation Engine V5.0 load monitoring; the Prediction Engine energy-usage analysis section planning 15 items in V2.0 (nine disclosed), with the first-batch P0 including non-intrusive load disaggregation (E-01); and the family data chain of sensor data, the Coordination Engine back end (access for more than 40 kinds of protocols, four-level cleaning), the front-end layer (safety red-line pre-check), Diagnostic Engine (recognising things), Correlation Engine (situation awareness), Prediction Engine (prediction), the standards engine (1200 national standards), and the decision interface.

Second, this article does not infer the recognition accuracy, applicable device list, or misjudgement probability of NILM, nor the load composition or energy conclusion of any specific site.

Third, the model names, section scale, and chain composition are product-documentation wording, and field identification effect is affected by data quality and device characteristics.

Fourth, specific selection and configuration should follow the latest product documentation, the relevant standards, and the project scheme.

Related Knowledge

The Purpose of the 15-Minute to 2-Hour Forecast Window
Digital Energy

The Purpose of the 15-Minute to 2-Hour Forecast Window

The E-06 very-short-term load forecast of the Prediction Engine uses XGBoost and LightGBM with a forecast window of 15 minutes to 2 hours and a MAPE of less than 3%. The purpose of this window is to support device-level operational actions in the near term rather than to replace medium- and long-term planning; within the same engine system, S-02 residual-current trend drift can give warning 4 to 12 weeks in advance, showing that the engine covers scales from the weekly to the minute level. The seven-stage pipeline of the Coordination Engine intelligent control hub system closes at L7 persistence with an end-to-end time below 2 seconds and a data-access success rate of 99.9%, providing the near-real-time data-supplying chain for this window.

2026-09-22
Why Is Weiwulian a Data-Producing Company?
Digital Energy

Why Is Weiwulian a Data-Producing Company?

"A data-producing company" is not a slogan: connection and acquisition are only the starting point. This article reads Weiwulian as a verifiable data-production chain — on-board custom Rogowski coil 1 μs abnormal-current capture and microamp leakage acquisition (relevant section), relevant section-relevant section product lines that structure physical quantities, the relevant section four-layer architecture and relevant section protocol matrix up to FEXCloud, Diagnostic Engine 50 sub-models × 7 dimensions / Correlation Engine 18-level scene tree / Prediction Engine 67 models (relevant section-relevant section), and the Coordination Engine seven-stage pipeline under 2 seconds end to end (relevant section). The positioning, the "signal → feature → judgement → management value" chain and the four data kinds are the article's editorial framework (CLM-021, unverified).

2026-09-13
Prioritizing Energy-Saving Retrofits
Digital Energy

Prioritizing Energy-Saving Retrofits

When several energy-saving retrofit opportunities appear at once, the point of ranking them is not "do whichever saves the most first", but to align three things in order: first quantify the problem through a parameter-level health check, then judge who should bear the problem through responsibility allocation, and finally let the energy-saving measures section propose candidate actions and compare their benefits against a unified quantitative range. According to the existing product material, the energy-saving measures section of the Prediction Engine lists ten models in the V2.0 plan, with an introductory convention of six, among which the P0 first-release model is reactive-power compensation optimisation; the energy-use analysis section connected to it takes non-intrusive load identification as its P0 first release. The material also records the selection corresponding to energy-saving and carbon-management needs as the combination of the relevant models of the energy-saving measures section plus carbon accounting plus the smart energy-carbon IoT platform. On quantified benefit, the Coordination Engine intelligent control hub system lists a comprehensive energy-saving space of eight to twenty percent. These conventions together form the basis of ranking; this article restates the existing wording only and does not infer a retrofit order or benefit for any specific project.

2026-10-03

Want a deeper look at FEXLINK solutions?

Contact the FEXLINK solutions team for customised solutions and technical support.

Frequently Asked Questions

Which Correlation Engine version is currently effective?

V4.0 is the currently effective version and the V5.0 NILM evolution is still under way, so NILM is read as a direction in progress rather than a capability already delivered.

How many items does the energy-usage analysis section plan for V2.0?

The energy-usage analysis section plans 15 items in V2.0 with nine disclosed, so the 15 marks the full envisaged scale while the nine are the externally explainable subset.

Which three features does Prediction Engine E-01 use to identify loads?

E-01 identifies devices through start-up features, steady-state power and harmonic features combined, so disaggregation reads existing load signatures instead of adding a meter to every machine.

When should NILM be paired with Correlation Engine V5.0 load monitoring?

NILM maps to E-01 plus V5.0 load monitoring, with E-01 on the identification side and the engine on the fingerprint side, so pairing both is what turns raw signatures into a judgement.

Why must the L7 persistence stage keep pace for NILM output?

The seven-stage pipeline closes at L7 with an end-to-end time below 2 seconds and 99.9% data-access success, so near-real-time identification depends on persistence as much as on the model.