Search

Flooding is one of the most common natural hazards that have extremely detrimental consequences. Understanding which areas are vulnerable to flooding is crucial to addressing these effects. In this work, we use machine learning models and Automated machine learning (AutoML) systems for flood susceptibility mapping in Kerala, India. In particular, we used a three-dimensional convolutional neural network (CNN) architecture for this purpose. The CNN model was assisted with hyperparameter optimization techniques that combine Bayesian optimization with evolutionary algorithms like differential evolution and covariance matrix adaptation evolutionary strategies. The performances of all models are compared in terms of cross-entropy loss, accuracy, precision, recall, area under the curve (AUC) and kappa score. The CNN model shows better performance than the AutoML models. Evolutionary algorithm-assisted hyperparameter optimization methods improved the efficiency of the CNN model by 4 and 9 percent in terms of accuracy and by 0.0265 and 0.0497 with reference to the AUC score. [Display omitted] • A 3D CNN model is proposed to assess flood susceptibility. • Model performance is compared with state-of-the-art machine learning and AutoML models. • A novel hyperparameter optimization model designed for the 3D CNN model. • The model is further improvised by using the evolutionary algorithms-assisted Bayesian optimization technique. [ABSTRACT FROM AUTHOR]

• A novel approach is proposed to detect the non-stationarity in rainfall time series. • Developed a probabilistic time-dependent bivariate intensity-duration model. • Proposed approach checks for any signature of non-stationarity in the probabilistic distribution parameters and prunes off the insignificant (stationary) probabilistic distribution parameters, thus reducing the complexity and improving the computational efficiency of the model. • The present approach is superior to the stationary and traditional non-stationary approaches in avoiding any possible under/over-estimation of the return periods. Dynamic behavior of extreme rainfall characteristics heightened by the abrupt warming of the environment has affected the sustainability of the existing water resources systems and infrastructure, which were designed employing the traditional 'stationary' assumption. Here, we propose a realistic and efficient framework to detect non-stationarity in the observed hydrologic variables, to overcome a few limitations suffered by the traditional non-stationary approaches. The methodology is demonstrated over the short period rainfall series of four major metropolitan cities of India, where the intensity of the rainfall is reportedly increasing sharply accompanied by the changes in the pattern of rainfall. Since direct runoff is influenced by the intensity and duration of rainfall, it is important to study the joint characteristics of intensity and duration in the context of non-stationarity, especially in urban regions where the relationship is more distinct. Hence, we estimated the time-varying joint return period/return level of its extremes, utilizing a dynamic Bayesian copula. We have implemented time-varying multivariate probability frequency analysis to derive the time-varying Intensity-Duration relationship. Here we employed a Bayesian approach through Differential Evolution Markov Chain (DE–MC) algorithm to estimate the uncertainty bound of the time-varying return level. The results emphasize that the probabilistic distribution parameters vary both temporally and spatially, and recommend the incorporation of non-stationarity in the extreme event modeling, only if there is a change in the probabilistic distribution parameters. This non-stationary model can be seamlessly employed to compute return levels with better accuracy and reliability than traditional stationary/non-stationary methods. We observe that the short duration return level increases at a faster rate than the longer ones, with the credibility interval larger than that of long duration return levels. The results also highlight the importance of adopting event based non-stationary IDF curves for the design of water resource systems and to devise long-term decision-making strategies to address the changing climate. However, we do not draw any conclusion concerning climate change, in general, from the short duration rainfall records used in this study. [ABSTRACT FROM AUTHOR]

In India, power shortage is a major issue for economical growth. According to the data provided by National Load Despatch Centre, peak power shortage in Tamilnadu during the year 2014 varies between 3000 MW and 4000 MW. This power shortage can be reduced by increasing installed capacity of conventional and non-conventional energy sources. But the construction of new generation plants is cost-effective and also power generation is not assured throughout the year. This power shortage can also be minimized by implementing load management in the consumer side. This paper focuses on the reduction of peak demand by the proper operating schedule of major equipments. For this analysis, three rice industries have been considered. The major operating sections in the rice industries are pre-cleaning, soaking, pre-milling and milling. In this proposed work, to reduce the peak demand, the operating time of pre-milling section is rescheduled using the optimization techniques, DE (Differential Evolution), PSO (Particle Swarm Optimization) and ABC (Artificial Bee Colony). The rescheduled results given by DE, PSO and ABC algorithms reduce the peak demand cost of the energy consumed in three rice industries. However, the optimum scheduling obtained by ABC reduces the feeder power flow than the DE and PSO scheduling. [ABSTRACT FROM AUTHOR]

Coal mills are bottleneck in coal–fired power generation process due to difficulty in developing efficient controls and faults occurring inside the mills. In this paper, a dynamic coal mill model is developed using conservation laws and empirical relations. The physical processes occurring inside the mill such as grinding, pneumatic transport, drying, heat transfer, and classification are modeled. Differential Evolution is used to estimate the model parameters. The model is validated using data obtained from a 500 MW plant in India. An application of the proposed model for online estimation of moisture in the raw coal is presented. [ABSTRACT FROM AUTHOR]