Optimising the prediction of hourly heat demand using an effective Artificial Neural Network method

This is the revised version (2026) of the code and network, as used in the revised manuscript; it replaces the 2024 version. The attached data includes the following files:</p><ol><li>Trained neural network for Model 3 (.mat file), 49 input delays and 2 hidden neurons</li><li>Matlab code for running Model 3 (.m file), including the spike repair rule</li><li>Heat demand data (2010 test series), to be used with the above two files as an example</li><li>rsquare.m, a helper function called by the code</li></ol><p dir="ltr"><br></p>