30 March 2020 in IEEE Transactions on Sustainable Computing
In this paper, a novel method is developed for day-ahead daily river-flow forecasting based on convolutional neural network (CNN). The proposed method incorporates both spatial and temporal information to improve the forecasting performance. A CNN model is usually trained by minimizing the mean squared error which is, however, sensitive to few particularly large errors. This character of squared error loss function will result in a poor estimator. To tackle the problem, a robust loss function is proposed to train the CNN. To facilitate the training of CNNs for multiple sites forecasting, transfer learning is also applied in this study. With transfer learning, a new CNN inherits the structure and partial learnable parameters from a well-trained CNN to reduce the training complexity. The forecasting performance of the proposed method is validated with real data of four rivers by comparing with widely used benchmarking models including the autoregressive model, multilayer perception network, kernel ridge regression, radial basis function neural network, and generic CNN. Numerical results show that the proposed method performs best in terms of the root mean squared error, mean absolute error, and mean absolute percentage error. The two-sample Kolmogorov-Smirnov test is further applied to assess the confidence on the conclusion.
20 January 2020 in IEEE Access
Parameter estimation of Lorenz chaotic system using a novel hybrid Jaya-Powell algorithm is proposed in this paper. Since the nonlinear dynamic system is complex with multi-dimension parameters, estimating parameters of the system can be considered as a muti-objective optimization task. The proposed Jaya-Powell algorithm combines the Jaya and Powell algorithm to search for the relatively global optimum and local optimum respectively, which offers a more accurate and effective estimation. The searching strategy of the proposed algorithm facilitates the balance of the exploration and exploitation in the optimization procedure. Due to no algorithm-specific parameters are required in the Jaya and Powell algorithm, the proposed Jaya-Powell can avoid deliberate fine-tuning of corresponding parameters. To validate the accuracy and robustness of the proposed algorithm in parameter estimation, the simulation of Lorenz chaotic system and comparative experiments are conducted. Seven algorithms, including Jaya algorithm, Powell algorithm, Teaching-learning-based optimization (TLBO) algorithm, particle swarm optimization (PSO), genetic algorithm (GA), covariance matrix adaptation evolution strategy (CMA-ES), and cluster-chaotic-optimization algorithm (CCO), are considered as benchmarking algorithms in the comparison. The proposed hybrid Jaya-Powell algorithm outperforms seven benchmarking algorithms with the more accurate estimation and the relatively faster convergence. Based on the embedded system Raspberry pi 3, the proposed algorithm achieves the similar performance by comparing with the experiments conducted on the computer. The successful implementation via Raspberry pi 3 facilitates the application of the proposed algorithm in edge computing.