João Machado de Freitas

Avoiding Domain Drift and Constant Predictions with Diffusion Enhanced Vector-Quantized Autoencoders for Temperature Predictions

Accurate temperature prediction in rotary cement kilns is crucial for process stability and equipment longevity. However, the repeated application of the predicted values creates an error accumulation over the length of the forecast, causing a domain drift of the predictions. This issue is exacerbat…