A per‐unit curve rotated decoupling method for CNN‐TCN based day‐ahead load forecasting

The existing load forecasting method based on the per-unit curve static decoupling (PCSD) would easily lead to deviation and translation of results. To tackle this challenge, a rotated (PCRD) is proposed for day-ahead with convolutional neural network temporal framework. PCRD decomposes into three parts: curve, 0 AM load, daily average load. shape feature extracted by CNN, features are TCN. rotation operation rotate at midpoint until first point aligned same point, in order improve similarity curves alleviate deflection can verify accuracy which alleviates Several experimental results show that has higher stability than PCSD method. After repeated experiments multiple data sets, generalization ability model also verified.