pH reactor dataset

Dataset of noisy input and output measurements collected from a simulator of a pH reactor. The system is single input single output.

The dataset consists of the following files:

'PH_U_Train.csv' and 'PH_Y_Train.csv': training dataset. The training dataset consists of 15 experiments, each one with 2000 (u, y) samples. The i-th column corresponds to the input (or output) variable of the i-th experiment.

'PH_U_Val.csv' and 'PH_Y_Val.csv': validation dataset. The validation dataset consists of 4 experiments, each one with 2000 (u, y) samples. The i-th column corresponds to the input (or output) variable of the i-th experiment.

'PH_U_Test.csv' and 'PH_Y_Test.csv': test dataset. The test dataset consists of 1 experiment  of 2000 (u, y) samples.

'PH_U.csv' and 'PH_Y.csv': concatenation of the training, validation, and test datasets. 

Moreover, the parameters of an Incrementally Input-to-State Stable LSTM neural network trained to learn the dynamical system are reported in the file 'LSTM_parameters.mat'.

If you use this data please cite the related paper

Terzi, E., Bonassi, F., Farina, M., & Scattolini, R. (2021). Learning model predictive control with long short‐term memory networks. International Journal of Robust and Nonlinear Control.