Under an internet of thing (IoT)-based framework, a power grid monitoring system can collect real-time transmission line data, which can then be used to predict the cascading failure risks within the grid. This study proposed a two-stage cascade failure prediction (CFP) model that integrated deep learning techniques with supervised classification algorithms to identify transmission lines susceptible to cascading failures, thereby preventing large-scale power system collapses. In the first-stage, the CFP model utilizes historical transmission line data to predict lines at risk in the next time step. In the second-stage, it assesses whether the predicted at-risk lines could trigger system instability or power outages. Three deep learning-based algorithms were employed in the first-stage, whereas three binary classification algorithms were utilized in the second-stage. The proposed CFP model was evaluated using an IEEE 39-bus system, where 2,000 cascading failure events were simulated through a real-time digital simulator (RTDS). Experimental results demonstrate that the two-stage CFP model, particularly when combining a deep neural network with a binary classifier, can accurately identify potential risk lines in the test grid with an accuracy ranging from 97.5% to 98.3%. Moreover, the CFP model exhibits strong robustness against noisy and erroneous data. These predictive results provide valuable insights for power companies in formulating preventive safety strategies against cascading failures.

The issue of air pollution in urban areas is gaining attention due to the rise of environmental and health concerns, especially for the particulate matter 2.5 (PM textsubscript 2.5), which poses the greatest health risk to humans. Accurate air quality prediction data allows government officials and the public to take preventive measures in advance. Recently, many air quality prediction studies have used machine learning techniques to identify patterns and rules in air quality data. However, these studies generally adopted under-represented background levels, and the prediction intervals were often in hours, which may not be suitable for residents who needed accurate air quality forecasts. Therefore, this study proposes a deep-learning-based multi-timestamp multi-location PM2.5 prediction system built on two acrlong RNN models: 1) long short-term memory (LSTM) and 2) gated recurrent unit (GRU). Airbox data for the Taipei metropolitan area serves as the main source of training data to develop a forecasting model that can predict changes of PM2.5 levels within the next 6–30 min in different locations. The prediction results are verified by comparing them with the PM2.5 measuring results from an Internet of Things (IoT)-based acrlong OVMS, which enables real-time data sensing and collection, and wireless transmission. The error and accuracy are$0.922 \mu \text{g}$/m 3 and 100% for the LSTM-based prediction model, and$0.940 \mu \text{g}$/m 3 and 95.7% for the GRU-based prediction model, respectively. These results can be sent out as warning messages to elderly and asthmatic patients, or serve as important information for route recommendations and policy formulation.

Bad road quality brings many problems, such as putting drivers and passengers in danger and causing vehicle suspension system wear. Maintaining high-quality roads relies on regular inspections and repairs, but this is a time-consuming and labor-intensive task. To improve road quality and increase the efficiency of road repairs, an Internet of Things based anomaly detection system (ADS) is proposed to monitor road surfaces. A machine-learning method, support vector machine (SVM), is utilized to identify and classify different types of road surface anomalies. Other five classifiers are also examined using the same testing data. The high classification accuracies obtained from the proposed SVM model can be incorporated with a Google Map, so the road surface information can be easily browsed. With the proposed ADS system, it requires manpower and time that can be greatly reduced for examining surface conditions of roads and significantly improve the efficiency of road maintenance.

For overhead power grids, unexpected serious line sagging of extra-high voltage transmission lines would easily lead to major blackouts. Different direct and indirect sag measuring methods have been proposed, but they all have their own limitations. In this study, an Internet of Things (IoT)-based sag-monitoring system is proposed, which is able to perform real-time detection of line sags. The whole system has been deployed on two 161-kV lines for long-term field testing. The effectiveness of the proposed sag-monitoring system is verified through both theoretical calculation and field measurements. In this sag-monitoring system, a sag-sensing module is equipped with an embedded triaxial accelerometer to detect line sags at different spans of a single circuit. A catenary equation that takes temperature dependency into consideration is derived, so the measured accelerometer parameters can be converted to accurate line sag values. The long-term testing results show that the proposed sag-monitoring system yields an average error of 2.09%. The average differences between the sag values coming from the proposed system and a commercial sag measuring device are relatively small (between 0.57% and 4.14%), which proves that the sag values provided by the proposed system are reliable and accurate in the long-term testing. In addition, compared to the other existing sag measuring methods, the advantages of employing the proposed system are high measurement accuracy, and enabling wide field implementation, online monitoring, long-term field operation, and real-time transmission.

To quickly isolate suspected cases to control the epidemics, this study proposes a body temperature monitoring system with a thermography based on the Internet of Things (IoT) architecture. The collected data are transmitted to a back-end platform via wireless communication. Using the analyzed data, the platform provides services, such as instant alerts for any anomalies, infectious disease outbreak prediction, and risk level assessment for a given area, and it will be a great help to epidemic prevention. The mean absolute percentage error and root mean square error of the proposed monitoring system under an extensive series of experiments are 0.04% and 0.0204°C, respectively. It shows that the body temperature measured by the thermal imaging sensor in the system can accurately represent the actual body temperature after specific calibrations that take the environmental temperature into account. It can also be expanded to a decision supporting system to help schools or government agencies to make proper decisions to stop the spread of infectious diseases.

Dynamic thermal rating (DTR) is a technique that can effectively reduce the complexity of the decision-making processes for a smart grid. Internet-of-Things-based DTR monitoring systems can be used to achieve reliable and low-cost remote monitoring of power grids, but this method is heavily reliant on collecting accurate real-time meteorological data by sensors deployed on the power lines. However, deploying sensors on each span of the line may not be feasible due to the high cost of such sensors. Thus, this article proposes a modified binary particle swarm optimization (MBPSO) strategy to solve multiobjective combinational decision problems. The proposed method is able to determine the minimum number of sensors required to achieve nearly ideal performance. A 161-kV line located between Xizhi and Minquan, part of the Taiwan Power Company transmission system, was selected as the experimental target. The MBPSO algorithm was solved based upon hourly meteorological data provided by Taiwan's Central Weather Bureau. The results obtained with the proposed method show that only one sensor needs to be deployed on the third span of the line to effectively monitor more than 89.6% of high conductor temperature events, and the root mean square error on the reconstructed conductor temperature distribution is less than 0.8 °C.