A convolutional neural network hardware acceleration device, a convolutional calculation method, and a storage medium are provided. The device includes an instruction processing element (1), a hardware acceleration component (2), and an external data memory element (3). The instruction processing element (1) decodes instructions, and controls the hardware acceleration component (2) to perform operations corresponding to decoded instructions. The hardware acceleration component (2) includes: an input buffer element (21), a data calculation element (22) and an output buffer element (23). The external data memory element (3) stores the calculation result output by the output buffer element (23) and transmits the data to the input buffer element (21).