Feature recognition of feeder load composition based on deep learning network based on feature linear modulation

With the increasing scale of power grid, the structure is more and more complex, and the requirement of power supply reliability is higher and higher, it is very difficult to obtain the dynamic characteristics of power load through field test. Based on the idea of statistical synthesis method, this paper uses the power system simulation software to obtain the dynamic characteristics of the corresponding power load through simulation calculation. In order to obtain the load database required for the identification of feeder load composition proportion, the power grid model including variable frequency air conditioning load, distributed photovoltaic load, induction motor load and static load is established as shown in the figure. By setting voltage drop, the feeder power data under different load conditions are obtained, and the feeder load database is established, The improved deep learning network method based on feature linear modulation is used to mine the characteristics of load node or feeder power change,and identify the load composition. Following training, the results were compared to those of the traditional BP neural network, revealing a recognition accuracy surpassing 99%. Additionally, certain loads exhibited accuracy exceeding 99.6%, thus indicating a significantly elevated level of recognition precision.