Electricity consumption forecasting is a vital task for smart grid building regarding the supply and demand of electric power. Many pieces of research focused on the factors of weather, holidays, and temperatures for electricity forecasting that requires to collect those data by using kinds of sensors, which raises the cost of time and resources. Besides, most of the existing methods only focused on one or two types of forecasts, which cannot satisfy the actual needs of decision-making. This paper proposes a novel hybrid deep model for multiple forecasts by combining Convolutional Neural Networks (CNN) and Long-Short Term Memory (LSTM) algorithm without additional sensor data, and also considers the corresponding statistics. Different from the conventional stacked CNN–LSTM, in the proposed hybrid model, CNN and LSTM extracted features in parallel, which can obtain more robust features with less loss of original information. Chiefly, CNN extracts multi-scale robust features by various filters at three levels and wide convolution technology. LSTM extracts the features which think about the impact of different time-steps. The features extracted by CNN and LSTM are combined with six statistical components as comprehensive features. Therefore, comprehensive features are the fusion of multi-scale, multi-domain (time and statistic domain) and robust due to the utilization of wide convolution technology. We validate the effectiveness of the proposed method on three natural subsets associated with electricity consumption. The comparative study shows the state-of-the-art performance of the proposed hybrid deep model with good robustness for very short-term, short-term, medium-term, and long-term electricity consumption forecasting.

Short-Term Load Forecasting (STLF) is one critical assignment regarding the power supply and demand in the smart grid. Multi-step STLF provides strong evidence for decision-making to achieve consistent, quick supply and reduce direct or indirect cost. However, most of the current research only focuses on one-step STLF, which cannot satisfy the human-beings needs. Besides, short-term consumption fluctuates significantly in different periods and people, which increases the difficulty of forecasting. In this paper, we present a novel deep model named multi-channel long short-term memory (LSTM) with time location (TL-MCLSTM) in a multiple output strategy to forecast the multi-step short-term power consumption. The proposed model contains three channels: power consumption, time location, and customer behavior channels, respectively. Power consumption channel reflects the change and general trend of use; Time location channel reflects the hidden pattern of customer habits, which records the information consisting of time, day of the week, holidays. Moreover, we combine a convolution autoencoder and k-means to identify the type of behavior at the customer behavior channel. Power consumption and time location channels are trained individually through the LSTM as it has excellent memory function. Extracted features from LSTM in power consumption and time location channels are combined with customer behavior as comprehensive features to forecast. We designed, trained, and verified our proposed deep model on two nature data sets, and compared with other leading deep learning-based methods. The comparative studies have confirmed the effectiveness and priority of TL-MCLSTM for multi-step short-term consumption forecasting. INDEX TERMS Short-term Load Forecasting (STLF), smart grid, LSTM, multi-step forecasting, convolution autoencoder.

Short-term power consumption forecasting plays a critical role in the process of building the smart grid. However, it is very challenging as the power consumption series has strong randomness and volatility. In this paper, the authors proposed a novel domain fusion deep model based on convolutional neural network (CNN), long short-term memory (LSTM), and discrete wavelet transform (DWT) to deal with this task accurately. The proposed deep model has two channels: raw power consumption and DWT. The raw power consumption channel corresponding to time-domain feature extraction while the DWT channel is frequency domain. They extract time-domain and frequency-domain features individually by using CNN. CNN extracted time-domain, and frequency-domain features are merged as time-frequency fusion features, which fully reflect the changing power consumption trend. The time-frequency fusion features are fed into LSTM to mining the features which have a long-time dependency. The comprehensive features are the fusion of time-domain and frequency-domain features with a long-time dependency, which are utilized for power consumption forecasting. The proposed method is evaluated on two public nature data sets related to power consumption with multiple metrics. The comparative experimental analysis has confirmed the state-of-theart performance of the proposed method for short-term power consumption forecasting.

Fault diagnosis (FD) plays a vital role in building a smart factory regarding system reliability improvement and cost reduction. Recent deep learning-based methods have been applied for FD and have obtained excellent performance. However, most of them require sufficient historical labeled data to train the model which is difficult and sometimes not available. Moreover, the big size model increases the difficulties for real-time FD. Therefore, this article proposed a domain adaptive and lightweight framework for FD based on a one-dimension convolutional neural network (1D-CNN). Particularly, 1D-CNN is designed with a structure of autoencoder to extract the rich, robust hidden features with less noise from source and target data. The extracted features are processed by correlation alignment (CORAL) to minimize domain shifts. Thus, the proposed method could learn robust and domain-invariance features from raw signals without any historical labeled target domain data for FD. We designed, trained, and tested the proposed method on CRWU bearing data sets. The sufficient comparative analysis confirmed its effectiveness for FD.

No abstract