Engineering topics

Chen, Yilin

Publications and source records attributed to Chen, Yilin.

2 records

DOE OSTI2024

Fertilizer management for global ammonia emission reduction

Crop production is a large source of atmospheric ammonia (NH 3 ), which poses risks to air quality, human health and ecosystems. However, estimating global NH 3 emissions from croplands is subject to uncertainties because of data limitations, thereby limiting the accurate identification of mitigation options and efficacy. In this report we develop a machine learning model for generating crop-specific and spatially explicit NH 3 emission factors globally (5-arcmin resolution) based on a compiled dataset of field observations. We show that global NH 3 emissions from rice, wheat and maize fields in 2018 were 4.3 ± 1.0 Tg N yr -1 , lower than previous estimates that did not fully consider fertilizer management practices. Furthermore, spatially optimizing fertilizer management, as guided by the machine learning model, has the potential to reduce the NH3 emissions by about 38% (1.6 ± 0.4 Tg N yr -1 ) without altering total fertilizer nitrogen inputs. Specifically, we estimate potential NH3 emissions reductions of 47% (44–56%) for rice, 27% (24–28%) for maize and 26% (20–28%) for wheat cultivation, respectively. Under future climate change scenarios, we estimate that NH3 emissions could increase by 4.0 ± 2.7% under SSP1–2.6 and 5.5 ± 5.7% under SSP5–8.5 by 2030–2060. However, targeted fertilizer management has the potential to mitigate these increases.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2022

Enhanced Oblique Decision Tree Enabled Policy Extraction for Deep Reinforcement Learning in Power System Emergency Control

Deep reinforcement learning (DRL) algorithms have successfully solved many challenging problems in various power system control scenarios. However, their decision-making process is usually regarded as black-boxes. Furthermore, how DRL models interact with human intelligence remains an open problem. Thus, this paper proposes a policy extraction framework to extract a complex DRL model into an explainable policy. This framework includes three parts: 1) DRL training and data generation. We train an agent for a specific control task and generate data, which contains the control policy of the agent. 2) Policy extraction. We propose an information gain rate based weighted oblique decision tree (IGR-WODT) for DRL policy extraction. 3) Policy evaluation. We define three metrics to evaluate the performance of the proposed approach. A case study for the under-voltage load shedding problem shows that the IGR-WODT presents a performance enhancement compared with DRL, weighted oblique decision tree, and univariate decision tree. The proposed policy extraction method could provide an intuitive explanation of the neural network decision-making process to the dispatchers when making final decisions on power grid operation. Also, the resulted rule-based controller could replace the deep neural network-based controller in many field edge devices with limited computing resources, providing comparable performance.

deep reinforcement learning↗