Search

Spatiotemporal assessment of solar potential is one of the most promising solutions to find suitable locations for future photovoltaic systems' placement. However, accurate assessment of solar potential on large areas can be challenging due to missing data or computational complexity. In this paper, a fully convolutional neural network based method for high-resolution spatiotemporal assessment of solar potential by using remote sensing data is presented. The method is trained and validated on the area of 32 km2 of the Maribor city, Slovenia, and tested on the 6 different locations in Slovenia and Germany. The proposed method was tested against the simulation algorithm, which utilized isotropic Liu Jordan diffuse model or anisotropic Perez model with sky view factor based shading. On average, a normalized root mean square error (NRMSE) of 6.06% and mean absolute percentage error (MAPE) of 4.29% was achieved in test locations against the simulation based on Liu-Jordan model. When training the fully convolutional neural network models against the ground truth, generated with more advanced Perez diffuse model, the average NRMSE was 8.37% and MAPE of 10.90% was achieved across all test locations. Additionally, it was shown that the proposed method can assess solar potential in high-resolution from the input in lower resolution with higher accuracy than the simulation algorithm, while being up to 1500-times faster. [ABSTRACT FROM AUTHOR]

Spatiotemporal assessment of solar potential is one of the most promising solutions to find suitable locations for future photovoltaic systems' placement. However, accurate assessment of solar potential on large areas can be challenging due to missing data or computational complexity. In this paper, a fully convolutional neural network based method for high-resolution spatiotemporal assessment of solar potential by using remote sensing data is presented. The method is trained and validated on the area of 32 km2 of the Maribor city, Slovenia, and tested on the 6 different locations in Slovenia and Germany. The proposed method was tested against the simulation algorithm, which utilized isotropic Liu Jordan diffuse model or anisotropic Perez model with sky view factor based shading. On average, a normalized root mean square error (NRMSE) of 6.06% and mean absolute percentage error (MAPE) of 4.29% was achieved in test locations against the simulation based on Liu-Jordan model. When training the fully convolutional neural network models against the ground truth, generated with more advanced Perez diffuse model, the average NRMSE was 8.37% and MAPE of 10.90% was achieved across all test locations. Additionally, it was shown that the proposed method can assess solar potential in high-resolution from the input in lower resolution with higher accuracy than the simulation algorithm, while being up to 1500-times faster. [ABSTRACT FROM AUTHOR]

The research presented in this paper is based on the hypothesis that the machine learning approach improves the accuracy of soil properties prediction. The correlations obtained in this research are important for understanding the overall strategy for soil properties prediction using optical spectroscopy sensors. Several research results have been stated and investigated. A comparison is made between six commonly used techniques: Random Forest, Decision Tree, Naïve Bayes, Support Vector Machine, Least-Square Support Vector Machine and Artificial Neural Network, showing that the best prediction accuracy cannot always be achieved by the most common and complicated method. The influence of the chosen category for nutrient characterization was investigated, indicating better prediction when a multi-component strategy was used. In contrast, the prediction of single-component soil properties was less accurate. In addition, the influence of category levels was not as significant as expected when choosing between 3-level, 5-level or 13-level nutrient characterization for some nutrients, which can be used for a more precise nutrient characterization strategy. A comparative analysis was performed between soil from a local farm with similar texture and soils collected from different locations in Slovenia, which gave a better prediction for a local farm. Finally, the influence of principal component analysis was validated using 5, 10, 20 and 50 first principal components, indicating the better performance of machine learning when using the 50 principal components. [ABSTRACT FROM AUTHOR]