The aims of this work package are:

• To develop a framework to test and validate data analysis methods, including AI techniques, for the early detection of abnormal grid events such as frequency deviations, sub-synchronous oscillations, power quality disturbances, and grid asset faults. To quantify the robustness and reliability of methods using existing grid measurement data as well as the test datasets generated in WP2 by determining the uncertainty of the probability of the predicted events.

• To develop a framework to test and validate data analysis methods, including AI techniques, for forecasting grid congestion and system imbalance between supply and demand, including the use of other data, such as weather predictions and temperature measurements of cables and lines. To quantify the robustness and reliability of methods using existing grid measurement data and the test datasets generated in WP2 by determining the uncertainty of the forecasts. To provide a Good Practice Guide on the validation of data-driven grid applications.

Whereas many different data analysis methods have been proposed in the past, it is not clear how to thoroughly and consistently assess the performance of these methods under a range of real-world conditions. A range of different metrics and datasets have previously been reported to test these methods, but this type of testing does not enable comparability of different methods, and it does not clarify how method performance may vary in challenging practical situations such as increased measurement uncertainty.

WP3 will fill this gap by developing a framework for comprehensive testing and validation of data analysis methods for abnormal grid event detection and grid congestion/system imbalance forecasting. The framework will combine outputs from WP1 and WP2 on test data creation and preparation, with definitions of metrics, uncertainty evaluation, as well as test cases and parameters developed in WP3. The framework will be demonstrated in detail to produce case studies on the testing and validation of at least 15 data analysis methods relating to five different types of grid phenomena.

Task 3.1 will identify and obtain at least 15 data analysis methods for the early detection of abnormal grid events and for forecasting grid congestion and system imbalance. Task 3.2 will develop a framework for testing, validation and uncertainty evaluation these methods, including the usage of procedures for data augmentation, uncertainty evaluation and explainability of AI methods. In task 3.3, the framework of Task 3.2 will be applied to the data analysis methods of Task 3.1 using the real measurement data from WP1 and the simulated measurement data from WP2.