Forecasting the piezometric level can be used to regulate water consumption, avoid floods and optimize water exploitation. In this context, the French geological survey (BRGM) has launched a challenge at the Knowledge Extraction and Management conference (EGC 2022) to propose models for predicting the future values of the piezometric levels in France. In this paper, we use three types of neural networks (convolutional, recurrent, and residual), which collaborate to forecast the piezometric level from the 15 of October 2021 to the 15 of January 2022. The source code and the results of our forecasting system are publicly available on GitHub 4.