The invention discloses an EMD and LSTM fused urban PM2.5 concentration prediction method, and relates to the field of air quality concentration prediction. The method comprises the following steps offirstly, acquiring the time sequence data per hour, and performing data cleaning on the acquired data; then, using the EMD (empirical mode decomposition) for carrying out stationary processing on thePM2.5 concentration data to obtain a plurality of components; then, determining a sliding time window T, carrying out data sequence segment segmentation processing on each component, and normalizinga unified dimension to obtain a plurality of data sets; dividing the data set into a training set and a test set, respectively constructing an LSTM network model for training, finally predicting eachcomponent by using the trained model, and carrying out reverse normalization processing on each component to obtain a final urban PM2.5 concentration prediction result; on the basis, constructing a long-term and short-term memory neural network LSTM model and training; finally, using the trained model for prediction, carrying out reverse normalization processing on the model, and obtaining the final urban PM2.5 concentration prediction result.