基于气象特征自适应重构算法与机器学习的超短期发电功率预测方法

Ultra Short Term Power Generation Prediction Method Based on Adaptive Reconstruction Algorithm of Meteorological Features and Machine Learning

风力发电与光伏发电的技术已经较为成熟,因可实现性强的特征使其得到可广泛应用,但由于风电和光伏发电具体波动性和不可控性,当其接入电力系统进行调度时,也对电力系统的安全运行造成了严峻挑战。对风电场和光伏系统的超短期发电功率预测具有重要意义,只有对超短期发电功率实现准确预测,才能保证电力系统的发电稳定性,研究研究基于气象特征自适应重构算法与机器学习的超短期发电功率预测方法。气象特征自适应重构算法定量功率与特征相关性;长短期记忆网络确定超短期发电功率预测函数;基于机器学习对偶转化函数预测超短期发电功率,完成方法设计。实验以风电场并入电力系统作为测试对象,将连续4日的发电功率数据作为测试样本,新方法能够实现精准预测,无论在突变时刻还是正常时刻,其预测值均与实际发电功率的曲线变化保持一致,具有应用价值。

The technology of wind and photovoltaic power generation has become relatively mature,and its strong feasibility has made it widely applicable.However,due to the specific volatility and uncontrollability of wind and photovoltaic power generation,when connected to the power system for scheduling,it also poses a serious challenge to the safe operation of the power system.The prediction of ultra-short term power generation in wind farms and photovoltaic systems is of great significance.Only by accurately predicting the ultra-short term power generation can the stability of the power system be guaranteed.Research and development of ultra-short term power generation prediction methods based on meteorological feature adaptive reconstruction algorithms and machine learning.The adaptive reconstruction algorithm for meteorological features quantifies the correlation between power and features;Determine the ultra short term power generation prediction function using short-term and short-term memory networks;Based on machine learning dual transformation function to predict ultra short term power generation and complete method design.The experiment takes the integration of wind farms into the power system as the test object,and takes 4 consecutive days of power generation data as the test sample.The new method can achieve accurate prediction,and its predicted values are consistent with the curve changes of actual power generation at both sudden and normal times,which has practical value.