Page 207 - AI for Good-Innovate for Impact Final Report 2024
P. 207
AI for Good-Innovate for Impact
Use case – 48: AI boosted Interpretable Renewable Energy
Forecasting 48-Alibaba
Country: China
Organization: Alibaba Damo (Hangzhou) Technology Co., Ltd.
Contact person: Yi Chen, (elaine.cy@ alibaba -inc .com, (86) 13605817921)
48�1� Use case summary table
Domain Sustainable Energy; Power Systems
The problem to be Deliver interpretable accurate day ahead wind power and rooftop
addressed solar power forecasting to mitigate the intermittency and less reliability
posed by booming capacity instalment of renewable energy.
Key aspects of the AI-based methods to deliver accurate and interpretable renewable
solution energy forecasting in an Asian city. The forecasting service covers all
the wind plants and the rooftop photovoltaics within the area, alongside
with an attribution analysis and error analysis.
Technology CNNs and Conventional Tree-based models with large-scale automatic
keywords feature augmentation, XAI
Data availability Private
Metadata (type of Tabular data, including measured power from wind turbines and solar
data) panels; numerical weather predictions.
Model Training and • CNN and tree models.
fine-tuning • Temporal convolutional architecture, (time series)
• Use the standard choices for optimizers.
• We use Ray Tune for hyper-parameter tuning.
Testbeds or pilot • Electricity bureau in Chinese city.
deployments • enewable Energy Forecasting System for State Grid Zhejiang
Electric Power Co. LTD. , Jiaxing Branch
https:// doi .org/ 10 .1609/ aaai .v37i13 .26853
https:// arxiv .org/ abs/ 2402 .05823
48�2� Use Case Description
48�2�1� Description
The booming capacity instalment of renewable energy such as wind power and photovoltaic
power in the past years have posed tremendous challenges to power grid scheduling and
191
