Search

Dissolved gas analysis (DGA) is an effective online fault diagnosis technique for large oil-immersed transformers. However, due to the limited number of DGA data, most deep learning models will be overfitted and the classification accuracy cannot be guaranteed. Therefore, this paper has introduced the idea of deep neural networks into the multi-grained cascade forest (gcForest), which is a tree-based deep learning model, and proposed an improved gcForest that can be accelerated by GPU. Firstly, in order to extract features more effectively and reduce memory consumption, the multi-grained scanning of gcForest is replaced by convolutional neural networks. Secondly, the cascade forest (CasForest) is replaced by cascade eXtreme gradient boosting (CasXGBoost) to improve the classification ability. Finally, 235 DGA samples are used to train and evaluate the proposed model. The average fault diagnosis accuracy of the improved gcForest is 88.08 %, while the average recall, precision, and Fl-score are 0.89, 0.90, 0.89, respectively. Moreover, the proposed method still has high fault diagnosis accuracy for datasets of different sizes.

Abstract Ecosystem services (ES) are benefits people receive from nature. To sustain these benefits, we need to spatially connect communities benefitting from specific ES with landscape features that generate the ES. A variety of process‐based models support ES assessments by estimating the biophysical supply of ES that comes from landscapes. However, less attention has been given to how ES flow from landscapes to beneficiary groups. A ‘serviceshed’ is defined as the spatial area that provides an ecosystem service to beneficiaries at a specific location—thus connecting people to the landscapes and seascapes upon which they depend through ES flows. In this article, we propose a general framework to empirically define serviceshed boundaries. Using publicly available data, we apply this framework to two provisioning and two regulating services (1) agricultural and fishing livelihoods and (2) pollination and coastal flood control, respectively, in Canada. We find that agricultural fields of different types and sizes contribute livelihood value to 85% of the communities in the agricultural landscape, and fishing grounds of different sizes contribute to 24% of the communities in the fishing study area. On average, communities with a lower proportion of agricultural labour are associated with larger fields, whilst larger fishing grounds were associated with communities with a greater percentage of their population in fisheries, showing how different ES can have varied relationships with beneficiary communities. For regulating services, we find 66% of pollinator‐supplying areas are within the serviceshed of farming communities. Natural habitats and agricultural land account for 72% and 28% of this serviceshed area, respectively. Our models suggest also that 26% of saltmarshes are within the serviceshed of flood‐prone communities and most coastal communities in our study area benefit from saltmarshes, especially from those without dykes. We demonstrate how serviceshed mapping, when integrating social and ecological information, can be useful in multiple decision contexts. Servicesheds can help planners and managers better design zoning restrictions, restoration activities to benefit communities, or subsidy programmes to replace the value of ES lost due to climate change or land use development. Read the free Plain Language Summary for this article on the Journal blog.