当前位置: X-MOL 学术Appl. Energy › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Designing a short-term load forecasting model in the urban smart grid system
Applied Energy ( IF 10.1 ) Pub Date : 2020-03-25 , DOI: 10.1016/j.apenergy.2020.114850
Chen Li

The transition of the energy system from fossil fuel towards renewable energy (RE) is rising sharply, which provides a cleaner energy source to the urban smart grid system. However, owing to the volatility and intermittency of RE, it is challenging to design an accurate and reliable short-term load forecasting model. Recently, machine learning (ML) based forecasting models have been applied for short-term load forecasting whereas most of them ignore the importance of characteristics mining, parameters fine-tuning, and forecasting stability. To dissolve the above issues, a short-term load forecasting model is proposed that incorporates thorough data mining and multi-step rolling forecasting. To alleviate the chaos of short-term load, a de-noising method based on decomposition and reconstruction is used. Then, a phase space reconstruction (PSR) method is employed to dynamically determine the train-test ratios and neurons settings of the artificial neural network (ANN). Further, a multi-objective grasshopper optimization algorithm (MOGOA) is applied to optimize the parameters of ANNs. Case studies are conducted in the urban smart grid systems of Victoria and New South Wales in Australia. Simulation results show that the proposed model can forecast short-term load well with various measurement metrics. Multiple criterion and statistical evaluation also show the good performance of the proposed forecasting model in terms of accuracy and stability. To conclude, the proposed model achieves high accuracy and robustness, which will provide references to RE transitions and smart grid optimization, and offer guidance to sustainable city development.



中文翻译:

在城市智能电网系统中设计短期负荷预测模型

能源系统从化石燃料向可再生能源(RE)的过渡急剧增加,这为城市智能电网系统提供了更清洁的能源。但是,由于可再生能源的波动性和间歇性,设计准确,可靠的短期负荷预测模型具有挑战性。最近,基于机器学习(ML)的预测模型已用于短期负荷预测,而其中大多数忽略了特征挖掘,参数微调和预测稳定性的重要性。为了解决上述问题,提出了一种短期负荷预测模型,该模型结合了全面的数据挖掘和多步滚动预测。为了减轻短期负荷的混乱,使用了一种基于分解和重构的降噪方法。然后,相空间重构(PSR)方法用于动态确定人工神经网络(ANN)的训练测试比率和神经元设置。此外,多目标蚱optimization优化算法(MOGOA)被应用于优化人工神经网络的参数。案例研究在澳大利亚维多利亚和新南威尔士州的城市智能电网系统中进行。仿真结果表明,该模型可以较好地预测各种测量指标下的短期负荷。多准则和统计评估在准确性和稳定性方面也显示了所提出的预测模型的良好性能。总而言之,该模型具有很高的准确性和鲁棒性,将为可再生能源转换和智能电网优化提供参考,并为可持续城市发展提供指导。

更新日期:2020-03-26
down
wechat
bug