Timely and accurate electricity price forecasting is essential for the efficient operation of competitive electricity markets. However, predicting day-ahead electricity prices remains challenging due to market volatility, anomalous price observations, and complex nonlinear relationships in electricity price dynamics. This study proposes a novel ensemble forecasting framework for predicting day-ahead electricity prices in the German electricity market. The framework integrates classical time-series models, machine-learning approaches, and hybrid models as base learners, and combines their forecasts using three weighting strategies based on validation performance, model diversity, and error minimization. To enhance forecasting robustness, six years of hourly electricity market data were preprocessed to detect and normalize anomalous observations. Experimental results demonstrate that the proposed ensemble framework consistently outperforms individual classical time-series models, standalone machine-learning models, and hybrid approaches. Among the benchmark models, hybrid models show better predictive performance than classical time-series methods but remain inferior to the proposed ensemble strategies. The findings highlight the importance of anomaly treatment and intelligent model combination for improving electricity price forecasting accuracy. The proposed framework provides valuable insights and practical forecasting tools for energy traders, grid operators, and policymakers involved in electricity market decision-making.