Agriculture is becoming increasingly reliant upon accurate data from sensor arrays, with localization an emerging application in the livestock industry. Ground-based Time Difference of Arrival (TDoA) radiolocation methods have the advantage of being lightweight and exhibit higher energy effciency than methods reliant upon Global Navigation Satellite System (GNSS). Such methods can employ small primary cell batteries, rather than rechargeable cells, and still deliver a multi-year deployment. In this paper, we present a novel deep learning algorithm adapted from a one dimensional U-Net like a convolutional neural network (CNN) model, originally developed for the task of semantic segmentation. This model both converts TDoA sequences directly to positions and reduces positional errors introduced by sources such as multipathing. We have evaluated the model by using simulated animal movements in the form of TDoA position sequences in combination with known distributions of TDoA error. When errors with a standard deviation of 50 m and 100 m are added to simulated TDoA transmissions the model is able to reduce this error to 22 m and 27 m (RMSE) respectively. Without correction, the standard deviation of these errors is on the order of 90 and 200 m. Accordingly, the model can reduce the error by greater than 80 m (> 80%), demonstrating the effectiveness of this novel 1D CNN U-Net like encoder/decoder for error correction of TDoA position estimates.