This trial serves as a practice round for the upcoming Global AI Challenge for Building E&M Facilities 2025. Your task will be to predict hourly cooling loads for a specific building, giving you valuable hands-on experience with the competition's data format, problem complexity, and evaluation framework before the main event.

### Problem:
Modern building automation systems generate vast amounts of sub-hourly sensor data from hundreds of measurement points, yet these datasets are frequently compromised by missing values and errors, undermining their usefulness for developing reliable HVAC performance models. The challenge is further compounded for newly constructed buildings that lack sufficient historical data for model training, as well as by the inherent diversity of building types in competition datasets - each with unique occupancy patterns, energy consumption profiles, and cooling load influencers - making it difficult to create prediction models that generalize effectively across different facilities.

### Proposed solutions:
Addressing the challenges of data quality, availability, and the diversity of building types is essential for developing AI-powered cooling load forecasting solutions. Successfully overcoming these hurdles will empower facility managers to determine optimal chiller operation strategies, thereby enhancing the overall energy efficiency of their building portfolios.

### Details:
This trial challenge prepares you for the Global AI Challenge for Building E&M Facilities 2025 through a hands-on forecasting exercise. Using real-world sensor data from Building X (an animal clinic), you'll develop and refine predictive AI models that address practical challenges in building energy management. The task mirrors real-world scenarios you'll encounter in the main competition, providing valuable experience with authentic datasets and operational conditions.

You will be tasked with leveraging this data to develop AI models that can accurately predict the hourly cooling load for building X in the next 30 days. In addition to the provided datasets, you are allowed to utilise open-source libraries and additional data sources that can be accessed by the Competition Organisers.

![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F26053641%2F38b3d43055cbb97918b77e266abde5af%2F1.png?generation=1744349017044932&alt=media)