This paper proposes the transmission and identification of energy sensing signals for IoT smart vehicles. The article proposes two research structures, including deep learning to detect abnormal sound signals of vehicle motors. The Internet of Things technology to collect vehicle sensing signals upload to the cloud system. A new approach of tunable FOPID (fractional-order proportional-integral-derivative) fuzzy controller for reduction of power losses on electric vehicle, which saving energy included as the AC generator, supercapacitors and Li-ion battery charging device etc., is proposed as shown in Figure 1. The energy consumption for power inverter, AC generator, transformer and supercapacitors battery was calculated. A modified switching pulse width-modulation (PWM) inverter, which a generator with signal AC/DC inverter FOPID controller is performed. A deep learning neural network method to solve motor unformal signal is proposed. Long short term memory (LSTM) have a higher accuracy of intercepting motor noise than conventional machine learning methods. Additionally, the disadvantage of deep learning algorithms is that they require large datasets to ensure an efficient performance. In this study, a combination of CNN and LSTM is proposed to prevent failure cause by the motor anomalous noise. This is applied to the CNN for noise image feature material. In addition, the LSTM is combined with the actual noise measurement spectrum data between 40 Hz – 20 kHz to analyze the noise of the motor. The CNN-LSTM can be effectively analyzed substantial anomalous motor noises. This approach yields a reduction in the switching losses and an improvement in the converting efficiency evidenced by the numerical data, as shown in Figure 2. Finally, experimental results also show that FOPID with fuzzy controller in saving energy can improve power loss, compared with that of conventional PID controller around 35%.