A Study of the BiLSTM Model Based on WOA Optimized Attention Mechanism for Power Load Forecasting
- Resource Type
- Conference
- Authors
- Li, Xinze; Zheng, Xinyue
- Source
- 2023 IEEE International Conference on Sensors, Electronics and Computer Engineering (ICSECE) Sensors, Electronics and Computer Engineering (ICSECE), 2023 IEEE International Conference on. :237-242 Aug, 2023
- Subject
- Communication, Networking and Broadcast Technologies
Computing and Processing
Robotics and Control Systems
Analytical models
Load forecasting
Computational modeling
Bidirectional control
Predictive models
Prediction algorithms
Whale optimization algorithms
electricity load forecasting
WOA (whale optimization algorithm)
BiLSTM (bi-directional long and short-term memory network)
BiLSTM-Attention model
- Language
The foundation for keeping a balance between power supply and demand is power load forecasting. The power load is crucial data for substations and power plants to plan daily power generation and choose the power system's mode of operation. It is crucial to increase the power load prediction accuracy in order to better plan the schedule and ensure the smooth operation of the power grid while enhancing system efficiency. To address the non-linear and periodic nature of the time-series changes in power load data, a hybrid model power load forecasting approach using a whale algorithm to optimize a bi-directional long and short-term memory network (WOA-BiLSTM-Attention) is presented. It initially fed a lot of historical electricity load data. The WOA portion moves in accordance with the fitness value to update the population and arrive at the global optimal solution after the data is trained and predicted by the network. The mean squared error between the actual output value and the desired output value is calculated. Some of the data sets from the 10th Teddy Cup Data Mining Challenge B were chosen for prediction in this work (https://aistudio.baidu.com/aistudio/datasetdetail/140138/0). The experimental results demonstrate that the suggested method's prediction accuracy is superior to that of the traditional LSTM network, the BiLSTM network, and the BP network and that the R 2 value can reach 0.9743.