Research On Medium-term Power Generation Planning Considering Wind Power Output Characteristics

Vigorously developing and using renewable energy is a national strategy to achieve low-carbon energy transformation.Wind power,as the main power source of renewable energy,has maintained a rapid growth trend for many years(cumulative installed capacity of 210 million kW at the end of 2019).Large-scale wind power integration faces severe accommodation problems due to its strong volatility.In the Three-north areas of China,where renewable energy is abundant,the peak-shaving capacity of thermal power units is becoming increasingly saturated.As the penetration rate of renewable energy increases,Start-stop Peak shaving becomes inevitable.In China,there are many large-scale thermal power generation units with long start-stop periods.In the medium-term power generation plan,the credible capacity of wind power should be reasonably included in order to activate the start-stop peaking ability of large-scale units.This thesis focuses on the research of medium-scale wind power characteristics,mid-term-across-scale generation plans considering wind power,wind power scenarios generation and robustness of unit commitment models.First,the fluctuation characteristics of medium-term wind power are studied.The medium-term wind power process is proposed to quantify the credible capacity of wind power,the sensitivity analysis of its key parameters is done later.Based on the spatial and seasonal statistics of the medium-term wind power characteristics,the availability and scope of the mid-term wind power process is analyzed.Based on the definition of the credible capacity and its confidence level,the mid-term wind power equivalent unit model is established,and a method for selecting the credible capacity coefficient based on sensitivity analysis is proposed.The proposed equivalent unit model can improve the credible capacity of the wind power in the medium-term scale with low risk.Then based on the unit commitment model of the traditional power generation plan,the correlation principle of the proposed mid-day-ahead double-layer unit commitment model and the differences between the proposed model and the traditional model at each stage were analyzed.The CPLEX optimization program was written to solve optimization model.Design an example system with a high wind power penetration,and compare the specific performance of the proposed method with the traditional mid-term/day unit commitment in terms of safety and economy.The results show that the proposed model can integrate the advantages of mid-term information integrity and the accuracy of the day-ahead information,improve wind power consumption,and ensure the reliability of system operation.Finally,compared with the traditional scenario generation method based on sampling,the advantages of the novel deep learning method in characterizing the uncertainty of wind power is discussed.The general framework of wind power scenario generation based on the Generative Adversarial Networks algorithm is given.In view of the improvement of the mid-scale and convolutional layer defects,a wind-power scenario generation model based on the improved conditional WGAN-GP algorithm was established.A robust analysis method based on robust decision-making for the cross-scale unit commitment model was proposed.The performance is tested in the case study.Result proves the robust advantages of the proposed mid-day-across-scale unit commitment model in terms of mid-term wind power forecast error.