In the context of the construction of new power systems

with the large-scale grid connection of new energy and the widespread application of nonlinear equipment

the harmonic problems in the power grid exhibit stronger randomness and volatility characteristics. Building a general model for uncertainty of harmonic sources is the foundation for evaluating the impact of harmonics and optimizing governance strategies. However

existing models have high computational complexity and are difficult to fully leverage the advantages of large sample data

resulting in limited model accuracy and difficulty in meeting practical engineering requirements. To solve the above problems

a method for establishing a general model with uncertain harmonic source based on XGBoost-KDE was proposed. Firstly

the mapping relationship between the physical quantities related to the harmonic current and harmonic emission characteristics of the harmonic source was established through the eXtreme gradient boosting (XGBoost) algorithm. Secondly

the kernel density estimation (KDE) method was used to obtain the probability density function and probability distribution function of the prediction error and to construct the prediction error interval. Then

based on the point prediction results

the interval prediction results of harmonic current at a certain confidence level were solved. Finally

a model online updating method based on adaptive dual strategy was proposed

which enabled the model to more accurately track the dynamic changes of harmonic characteristics of harmonic sources. Through actual testing and simulation data verification

compared with other modeling methods

the proposed modeling method has high accuracy and strong universality in single harmonic source modeling and complex multi harmonic source scenarios. The proposed online model updating method can use more historical sample data for dynamic model updates

thereby improving the applicability of the model.