Information about future streamflow is important for hydropower production planning, especially for damless hydro-power plants. The river flow is a reflection of various hydrological, hydrogeological, and meteorological factors, which increases the direct modeling difficulty, and favors the use of data-driven methods. In this paper, we propose the use of one-dimensional convolutional neural networks (1d-CNN) for multi-day ahead river flow forecasting and we present a multi-input model using correlated-input time-series. The proposed model was applied at the Madeira River, the Amazon’s largest and most important tributary, near the Santo Antônio damless hydro-power plant. We compared the proposed correlated-input 1d-CNN to a single-input 1d-CNN model and some baseline models. Furthermore, we conclude that 1d-CNN performed better than all baseline models and that the correlated-input forecasting model is 5 times smaller than the single-input equivalent with accuracy improvements.
This work introduces a new method for long-period fiber grating (LPG) sensors interrogation. This proposal uses a fiber Bragg grating (FBG) array to extract spectral information of the LPG sensor and an Artificial Neural Network to process this information. The information is processed to estimate the LPG resonant wavelength, without prior knowledge on the LPG spectrum. Therefore, the interrogator is LPG-insensitive, can be easily manufactured by optical fiber sensors laboratories at low-cost and is suitable for in-field applications. We demonstrated the filter array and Multilayer Perceptron (MLP) design, which are the proposed interrogator core. Furthermore, we analyzed the interrogation performance by the Mean Squared Error (MSE), the Mean Absolute Error (MAE), and the distribution of the residuals. The results showed our proposal can estimate the LPG resonant wavelength with 2.82 nm uncertainty, considering a 95% confidence interval, over 75 nm dynamic range for several LPGs, with different spectral characteristics. Moreover, our proposal can be easily tailored for different dynamic ranges and resolutions with proper adjustments on the FBG array and MLP.
In this work, we report on a new interrogation method for long-period fiber-grating (LPFG) sensors. This type of sensor has been gaining increasing attention due its several interesting practical and metrological advantages. However, the LPFG optical signal is complex and could be difficult to process. We present a novel approach to the LPFG resonant wavelength detection and tracking using a multifilter LPFG interrogator. Multifilter LPFG interrogators present a great solution for LPFG sensor interrogation because there is no need for a priori knowledge concerning the LPFG sensor spectrum. We present a novel source-compensating preprocessing and a fuzzy inference system (FIS) to track the LPFG resonant wavelength. Our proposal was compared to two baseline models: a simple linear regression and an artificial neural network. Our results showed the new approach could reduce resonant wavelength uncertainty by almost three times comparing to results found in the literature.
To ensure the safety and durability of structures, the constant monitoring of their structural health has been used by engineers to manage these structures. Optical fiber sensors have the advantage of long transducer-to-processing distances, the capacity to work in harsh environments, immunity to electromagnetic noise, and reliability. Therefore, optical fiber sensors are great candidates to monitor structures in-service. Here, we propose the use of a long-period fiber grating (LPG) array to monitor the loading condition of a structure. The identification method was tested in a simply-supported beam model and the estimated loading conditions were compared to the loading conditions applied to the beam. By analyzing different combinations of force intensity and position along the beam length, the results showed the force position could be estimated with 0.901% accuracy and force intensity with 2.51% accuracy using a four-LPG strain sensor array.