This paper presents a real-time forecasting procedure that utilizes multiple factors with different sampling frequencies to predict the weekly carbon price. Novel combination-MIDAS models with five weight-type schemes are proposed for evaluating the forecast accuracy. The evidence shows that combination-MIDAS models provide forecasting performance gains over traditional models, which supports the use of mixed-frequency data that consist of economic and energy indicators to forecast the weekly carbon price. It is also shown that, Coal is the best predictor for carbon price forecasting and that forecasts that are based on Crude have similar trends to actual carbon prices but are higher than the actual prices.