 The progress of renewable energy generation and energy storage technologies has led to the integration of more renewable sources and devices into the power system. As a result, single and multiple power quality disturbances, PQDs, have become more frequent, necessitating the need for real-time detection of these disturbances. To address this problem, we propose a novel approach combining fused convolutional neural networks, CNN, and long short-term memory, LSTM, architectures with time and frequency domain features. The frequency domain features were obtained from time series data using Fast Fourier Transform FFT. The original time domain and frequency domain features were then extracted by respective CNN LSTM structures. These features were subsequently concatenated and fed into fully connected layers to classify the PQDs. Our proposed method was trained and tested using 16 types of synthetic noise PQDs generated by mathematical models, as well as a simulation distributed power system. We compared three advanced neural network approaches, deep CNN, CNN, LSTM, and multi-fusion CNN, MFCNN. The results showed that the fused CNN LSTM model took only 0.6 for Milisecans. This article was authored by Senfeng Centre, Dong Ok Kim and Chang Iung Lim. We are article.tv, links in the description below.