Research On Frequency Stability Evaluation Of New Energy Power System Based On LightGBM

The new energy generation technology represented by wind power and photovoltaic has developed rapidly in recent years,and the proportion of new energy installed and generated in the power system is constantly increasing.Wind power and photovoltaic power generation are connected to the system through inverters,but their inertia response ability is poor and output uncertainty is strong,which occupies the output space of conventional units,leading to faster frequency change speed and changes in frequency response characteristics of the system.After disturbance,the risk of frequency instability increases,and the accuracy and speed of power system frequency stability evaluation methods are required to be higher.The continuously increasing proportion of new energy and the uncertainty of new energy output also pose higher requirements for the adaptability of frequency stability assessment methods.In response to the impact of new energy integration into the power system on frequency stability,this article first analyzes the impact of new energy integration on various stages of frequency response and its dominant factors,establishes wind power and photovoltaic power output models,and analyzes the output characteristics of wind power and photovoltaic power.The specific effects of wind speed and sunlight intensity on frequency response curves were analyzed in simulation examples;Analyzed the effect of new energy participation in frequency modulation on system frequency stability and the impact of changes in the proportion of new energy on system frequency response characteristics,laying the foundation for subsequent research.In view of the accuracy and rapidity requirements of the frequency stability assessment method,considering the round-off error caused by the setting of the classification threshold by the conventional classification algorithm,this paper uses the lightweight gradient boosting machine(lightGBM)with the characteristics of rapidity to establish three frequency index regression models with the frequency response index as the output and the dominant factor of the frequency response of the disturbed system as the input,Comprehensive judgment of frequency stability results.Combining feature processing with evaluation algorithms,analyzing the contribution value of input features to the model’s evaluation ability,removing features with smaller contributions,and reducing evaluation time.The proposed method has been verified to have high accuracy and fast calculation speed through simulation examples.Considering the uncertainty of new energy output and the impact of increased proportion of new energy on frequency response characteristics,a frequency stability evaluation method based on sampling of new energy output prediction error is proposed to describe the uncertainty of wind and solar power output using the distribution of new energy output prediction error,The superposition of the predicted values of new energy output at different times and the expected values of multiple sampling results of error distribution at corresponding output levels is used as the highest probability output scenario for frequency stability evaluation,further improving the accuracy of the evaluation.In response to the issue of significant changes in frequency response characteristics after the increase in the proportion of new energy,the applicability of the original evaluation model to new samples was analyzed,and the differences in simulation samples before and after the increase in the proportion of new energy were compared.Taking the actual simulation samples of the new system as input,a large number of similar samples are generated using the generated adversarial network and combined with some actual simulation samples to form a training set to train the new evaluation model.The effectiveness of the proposed method is verified in the remaining actual simulation samples.