§ Energy
Diagnostics at the asset, not the control room.
Wind, grid and solar condition monitoring that sends a flag instead of a waveform.
§ Why edge AI
The case for running it on the asset.
Energy assets are remote, long-lived and expensive to visit. The useful signal — a gearbox tonal, a blade imbalance, a thermal drift — is buried in raw data nobody wants to backhaul. E1M-AEN runs the always-on side: acoustic and vibration features computed continuously, the NPU woken on an anomaly, and a flag sent instead of a waveform. E1M-X V2N adds Linux for station-side gateways, with Zephyr on its Cortex-M33 for the deterministic half.
§ The constraint
Substations and turbines don’t have spare bandwidth for raw telemetry, and protection traffic such as IEC 61850 GOOSE trips is budgeted in milliseconds (3 ms for transfer-time class TT6). Diagnostics belong next to the signal.
Cloud round trip
sensor → uplink → inference → downlink → action
On-module
sensor → on-module inference → action
Sensor
Vibration and acoustic sensor
On-module inference
E1M-AEN
Action
Anomaly flag to SCADA
01 · Sense
A nacelle microphone, accelerometers on bearings, current monitors on drives.
02 · Extract
DSP features normalised to rotor order, so the signature holds at any rotor speed.
03 · Score
An anomaly score and advisory subsystem from the Ethos-U, woken only when features move.
04 · Report
A flag and a feature summary to SCADA, with signed firmware for the fleet.
§ Where it fits
What teams build with it.
01
Wind-turbine condition monitoring
Nacelle acoustics normalised to rotor order, so drivetrain tonals and blade aero faults are scored at any rotor speed.
02
Transformer and generator health
Vibration and thermal signatures classified at the asset, sending a flag instead of a raw waveform.
03
Solar and battery sites
Energy per inference measured on the bench, so the power budget of a solar-powered node is a number, not a guess.
04
Station gateways
Linux for the data plane and Zephyr for the deterministic side, on one module, with signed firmware for the fleet.
§ Recommended modules
Pick the AI compute that fits.
E1M-AEN
Always-on asset monitoring: acoustic, vibration and current on a power budget.
- CPU
- 0 / 1 / 2× Arm Cortex-A32 @ 800 MHz (per SoC: E3/E4 none, E5/E6 single, E7/E8 dual)
- NPU
- 2× Arm Ethos-U55; E4/E6/E8 add 1× Arm Ethos-U85 (250 GOPS · E3/E5/E7, 454 GOPS · E4/E6/E8)
- Temp
- −40 °C to +75 °C
Full specs →
E1M-X V2N
Station gateways: Linux data plane plus Zephyr on the Cortex-M33.
- CPU
- Quad Cortex-A55 @ 1.8 GHz
- NPU
- DRP-AI3 4 TOPS (dense), 15 TOPS (sparse)
- Temp
- −40 °C to +85 °C
Full specs →
§ Inside the E1M-AEN
The module this vertical leads with.
- MCU (SoC)
- 2× Arm Cortex-M55: 1× 160 MHz (HE) + 1× 400 MHz (HP)
- CPU
- 0 / 1 / 2× Arm Cortex-A32 @ 800 MHz (per SoC: E3/E4 none, E5/E6 single, E7/E8 dual)
- NPU
- 2× Arm Ethos-U55; E4/E6/E8 add 1× Arm Ethos-U85 (250 GOPS · E3/E5/E7, 454 GOPS · E4/E6/E8)
- SoC family
- Alif Ensemble E3 / E4 / E5 / E6 / E7 / E8
- RAM
- 13.5 MB (E3/E5/E7) or 9.75 MB (E4/E6/E8) SRAM on-die
- ROM
- 5.5 MB MRAM on-chip
- External memory
- HyperRAM + NOR flash on one OSPI bus: E1M-AEN803 fits 512 Mbit HyperRAM + 256 Mbit NOR; E1M-AEN801 fits none; build option on E1M-AEN301 / 401 / 501 / 601 / 701
- MIPI CSI-2
- 1× 2-lane
- MIPI DSI
- 1× 2-lane
- Parallel Camera
- 1× (8-bit parallel)
- Ethernet
- 100 Mbps PHY
- USB
- USB 2.0 HS
- CAN-FD
- 1 channel (CAN PHY is an assembly option)
- Connectivity
- 2.4 GHz Wi-Fi 6 + BLE 5.4 (BDE-BW3501N, TI CC3501E)
§ Reference examples from the Alp SDK
Code you can fork, with its evidence on the label.
Every example is real C/C++ in the open-source alp-sdk repo. Clone it, change the SKU in board.yaml, build. Each card says how far it has been proven (on silicon, partly, or not yet on hardware) as recorded in the example's own README and the SDK's verification ledger.
1 Run on silicon · 4 Not yet on hardware
acoustic-anomaly-wind-turbine
Not yet on hardwareWind-turbine acoustic monitor
Nacelle microphone → spectral features normalised to rotor order (blade-pass frequency) → anomaly score for drivetrain tonals and blade aero faults, RPM-invariant.
E1M-AEN801 · An airborne mic does not detect internal blade cracks or delamination. Runs end to end on native_sim; a run on a real nacelle is bench-gated.
View on GitHub ↗
ai-anomaly-detection-vibration
Not yet on hardwareTransformer / generator health
Vibration signatures → anomaly classification on the Ethos-U NPU, at the asset.
E1M-AEN801 · Compiles; not yet run on hardware.
View on GitHub ↗
aen-inference-energy
Run on siliconEnergy per inference
Runs a model on the Ethos-U85 while sampling the EVK INA236 rail monitors, and reports the incremental energy per inference.
E1M-AEN801 · Measured on silicon. The figure is a carrier-rail delta — not NPU-only or module-isolated energy.
View on GitHub ↗
iot-fleet-ota
Not yet on hardwareFleet OTA verification
Signed firmware updates for a fleet: MCUboot plus ECDSA-P256 signature verification, declared in board.yaml.
E1M-AEN801 · Verify-only on E1M-AEN801 — the image-apply step is deferred (#1069). native_sim build verified; hardware run gates on a staged Mender server.
View on GitHub ↗
heterogeneous-offload
Not yet on hardwareDual-OS gateway pattern
The pattern behind a station gateway: Linux on the A55 offloads a 1024-point FFT to Zephyr on the M33 over RPMsg, from one board.yaml.
E1M-V2N101 · Builds against the V2N memory map; hardware bring-up not yet run.
View on GitHub ↗
Portable SDK headers for this vertical
These are the SDK's portable surfaces across E1M families; what each family supports underneath them is on the SDK page.
§ Getting started
From example to your board.
- 01First, scaffold a project from the reference example:
tan init --from-example audio/acoustic-anomaly-wind-turbine - 02Then set
som.skuin board.yaml to your module (the example ships targetingE1M-AEN801). - 03Build with the
tanCLI or the Alp IDE extension for VS Code, and run it on an E1M EVK. - 04Before relying on a result, read the status block at the top of the example README — it says what has run on hardware and what has not.
§ Proven today, and what is ahead
On silicon today: energy per inference measured on the E1M-AEN801 EVK. Still ahead: the wind-turbine and vibration monitors on real assets with trained models, and OTA image apply on E1M-AEN801 — fleet OTA verifies signatures today but does not yet apply images.
SDK verification ledger ↗§ Built for this vertical
Why E1M wins here.
- →
AEN: sub-1 mW standby, wake-on-event AI for solar trackers + smart meters
- →
MCUboot secure boot with signed images; fleet OTA signature verification (image apply on E1M-AEN801 still in progress)
- →
Outdoor substations: −40 °C to +75 °C on AEN, −40 °C to +85 °C on V2N