- In the non-intrusive load identification field, the accuracy of event detection, the uniqueness of feature extraction, and the identification rate of load affect the system performance. In recent years, as the sampling frequency of smart meters and instruments progreses, the technology of high sampling frequency (HSF) has been widely used in the field of non-Intrusive load monitoring (NILM). In this paper, a novel method combining grayscale images and convolutional neural networks for non-Intrusive load monitoring was proposed. The separated waveform is acquired, normalized, and compared with the sine waveform. Then, the comparison is done by calculating the Euclidean distance between the standard waveform and the separated current waveform at the first step. The second step is that the current waveform that was used to obtain the first derivative is arranged into a two-dimensional array, and then a two-dimensional array is converted into a gray-scale image. In the final step, the image is used for the input of a convolutional neural network to detect the event for the loads with normal as well as low or slow-climbing power conditions. The result shows that by applying the grayscale images for training the convolutional neural network, the operation of appliances can be detected without using a threshold.