摘要
电力系统负荷具有时变性和不确定性,且历史负荷数据的稀疏性也会导致预测模型因训练不足而产生较大偏差。为此,提出基于数字孪生与灰色预测算法的电力系统中长期负荷预测方法。首先,根据电力系统的负荷类型,通过数字孪生模型,将电力系统的物理实体与虚拟模型相结合,实时获取多维数据,弥补历史负荷数据的不足。利用快速傅里叶方法提取负荷的时域和频域瞬态分量,并建立负荷衍生特征。结合数字孪生体中负荷与影响因素间的关系,利用最大信息系数对其进行耦合特性分析,提取强相关性负荷数据。依据强关联性负荷数据的测量协方差矩阵,利用高斯隶属函数提取负荷的尺度特征;然后,利用局部加权方法对负荷数据的多维时域特征进行时间序列分解,得到周期分量、趋势分量以及残差分量;最后,引入灰色系统理论构建负荷预测模型,实现对电力系统中长期负荷的预测。灰色系统理论适用于小样本数据,能够在中长期负荷数据稀疏的情况下有效处理负荷数据中的不确定性和随机性,提高预测精度。实验结果表明,应用该方法得到的预测结果与实际结果的相关指数高于0.8,负荷波动捕捉率保持始终低于1%,预测结果保留率保持在93.3%~95.6%之间,说明该方法具有较高的预测精度。
Electric load in power systems exhibits time-varying and uncertain characteristics,and the sparsity of historical load data can also lead to significant bias in prediction models due to insufficient model training.To address this,a medium-and long-term load forecasting method for power systems based on digital twin and grey prediction algorithm was proposed.First,based on the load type of the power system,a digital twin model was used to combine the physical entity of the power system with the virtual model,and real-time multidimensional data were obtained to compensate for the lack of historical load data.The time-domain and frequency-domain transient components of the load were extracted using the fast Fourier method,and load-derived features were established.Based on this,combined with the relationship between load and influencing factors in the digital twin,the maximum information coefficient was used to analyze its coupling characteristics and extract strongly correlated load data.Based on the measurement covariance matrix of strongly correlated load data,the scale features of the load were extracted using Gaussian membership functions.Finally,the grey system theory was introduced to construct a load forecasting model,achieving the prediction of medium-and long-term loads in the power systems.The grey system theory is applicable to small sample data and can effectively handle the uncertainty and randomness in load data in the case of sparse medium-and long-term load data,improving prediction accuracy.The experimental results showed that the correlation index between the predicted results and the actual results obtained by applying this method is higher than 0.8,and the load fluctuation capture rate remains below 1%.The retention rate of the predicted results remains between 93.3%and 95.6%,indicating that this method has high prediction accuracy.
作者
薛涛
张丁月
程爱勇
傅谦晶
刘广华
王鹏飞
XUE Tao;ZHANG Dingyue;CHENG Aiyong;FU Qianjing;LIU Guanghua;WANG Pengfei(CHN Energy Zhejiang Zhoushan Power Generation Co.,Ltd.,Zhoushan 316012,China;CHN Energy I&C Technology Co.,Ltd.,Beijing 100088,China)
出处
《国外电子测量技术》
2025年第2期128-134,共7页
Foreign Electronic Measurement Technology
关键词
数字孪生
灰色预测
电力系统
中长期负荷
预测模型
digital twin
grey prediction
power system
medium and long-term load
predictive model
