Storm surges pose a critical threat to the Dutch North Sea coast, where low-lying areas are highly vulnerable to extreme water levels. While short-term storm surge forecasting is well established, extending predictions beyond 10 days remains a challenge due to the complexity of atmospheric dynamics and limitations in numerical models, which require extensive computational resources and lack flexibility for alternative forecasting techniques. This study explores the potential of weather pattern-based classification as a method for improving mid-term storm surge prediction. Phase I evaluates a set of predefined weather patterns (Neal, et al., 2018), originally developed by the Met Office for probabilistic forecasting in the UK, to assess their applicability for surge forecasting along the Dutch coast. The results indicate that while certain patterns show associations with high-surge events, they systematically underestimate surge magnitudes and, at longer lead times, the surge distributions associated with different patterns become increasingly similar, reducing their ability to distinguish between high- and low-surge conditions. Phase II explores an alternative self-clustered classification approach using k-means clustering with Principal Component Analysis (PCA) for dimensionality reduction. Several data selection methods, including surge thresholding, Maximum Dissimilarity Algorithm (MDA), and stratified sampling, are tested to optimise clustering. While the self-clustered patterns show slight improvements over Neal’s predefined patterns, they still underestimate surge magnitudes and lack the accuracy needed for operational forecasting. A proof-of-concept evaluation using storm Pia (December 2023) and a representative SEAS5 storm reveals that the self-clustered weather patterns struggle to capture extreme surge events. Although these methods are not yet suitable for operational forecasting, this study suggest several possible refinements, such as sequential clustering and expanding the spatial domain, to be promising avenues for enhancing predictive skill.