Differentiable physics: pooling across substations

The summary of this theme is in progress. The differentiable-physics series builds models from physical terms (demand, heating, solar) whose parameters are fitted by gradient descent. This theme covers whether one such model fitted across many Primary substations can match a model per substation, and beat it for a new substation or one with a short history. The Experiments page lists every experiment tried, with its verdict and a take-home message, and the Research plan lists what is planned next. Experiment 112 found that fitting the pooled model exactly did not remove its gap to per-substation models: with the interaction network on, the exact fit forecast far too high on four days.