Page 32 - AI Ready – Analysis Towards a Standardized Readiness Framework

AI Ready – Analysis Towards a Standardized Readiness Framework

Datasets include streamed data from cameras and smoke and light detection. Post-fire analysis
using proprietary algorithms for fire prediction, detection, and fire propagation models. Pilot
deployments in India (2 reserves), Brazil, and Portugal are used to test the models.

4�6�3 Disaster Risk Management in Complex Geography

The use case [50] actors in this case are transporters, and other systemic actors such as businesses,
insurers, disaster management entities such as government agencies, and the general public.
Data analysed includes open data from disaster management agencies (risk data, real-time), and
private data for creating context country-region specific advisories (satellite images). This data is
used for training models such as generation (advisory generation) and prediction (forecasting),
including continuous improvement models. Interoperability and compatibility of cross-region
data (e.g. early warning) and generated advisories are major reasons for standards and metrics
for AI readiness in this use case.

4�6�4 Networked ASEAN Peatland Forest for Net-Zero

This use case [48] proposes a tropical peatland fire weather index (FWI) system by combining the
GWL with the DC. Pilot deployments include a LoRa-based IoT system for peatland management
and detection in RMFR in Kuala Selangor, Malaysia. Verification of data with truth values from
the Malaysian Meteorological Department (METMalaysia) are significant in this use case. GWL
is predicted and FDRS indices such as DC, Duff Moisture Code (DMC), and FWI are predicted
based on the GWL. Gateways such as weather stations, water level sensors, and soil sensors in
combination with LoRa nodes and LoRa Gateways make it an end-to-end solution.

4.7 Climate, Clean Energy

While we transition to cleaner sources of energy, AI technology can provide suggestions on
optimizing energy sources and enable a smooth transition from conventional energy sources.
AI-based forecasts are crucial for real-time power distribution and load balancing, as they
help integrate green energy into electric grids and overcome the challenges of intermittent
availability.

4�7�1 AI Boosted Interpretable Renewable Energy Forecasting

This use case [80] [81] uses AI-based methods to deliver accurate and interpretable renewable
energy forecasting, which covers all the wind plants and the rooftop photovoltaics within the
area, alongside an attribution analysis and error analysis. The predictions are used to decide
power distribution and real-time power load balancing. AI based predictions help the integration
of economic green energy to the electric grids despite the intermittent availability of alternate
sources of energy.

The use case applied convolutional neural networks and conventional tree-based models with
large-scale automatic feature augmentation to produce reliable wind power and solar power
forecasts and shapely value-based explainable AI to interpret the underlying reason for such
interpretations. Ray Tune is used for hyper-parameter tuning. Data including measured power
from turbines, solar panels, and weather predictions are privately available.

The use case has been deployed in Zhejiang China.

25