Abstract:

To address the issue that the traditional load forecasting algorithm is insufficient in handling the complex nonlinear relationship and uncertain factors in the grid-connected environment of new energy,a new intelligent load forecasting model for power system based on a sparse encoder and an improved ant colony algorithm is proposed.Firstly,a sparse encoder with regularization constraints is constructed to effectively extract the deep features of load data and enhance the generalization ability of the encoder.Then,the dynamic domain search mechanism is introduced to form an improved ant colony algorithm,which significantly improves the global search capability and convergence efficiency in complex scenarios.Finally,the model is trained and compared based on the historical load data.Experimental results show that the load prediction accuracy of the proposed algorithm can reach 97.9%,which is about 20% higher than that of traditional methods, providing effective technical support for smart grid scheduling and sustainable development of energy Internet.

Key words: load forecasting, sparse encoder, improve ant colony algorithm, historical data, experimental comparison