基于RNN神经网络的甘肃电力现货价格有关损失函数的优化分析
Optimization Analysis of Loss Function Related to Spot Price of Gansu Power Based on RNN
摘要: 针对甘肃高比例新能源双边现货市场价格数据的非线性特征以及主流损失函数的缺陷，本文主要研究如何使用Huber损失函数与神经网络精确模拟日前电力现货价格的时间序列，旨在系统解决甘肃日前电力现货价格预测时反向传播损失函数的选择。首先分析了甘肃电力现货市场的运行特点及问题，建立了其电力现货价格的神经网络模型；为了模型的优化建立了反向传播算法中损失函数应该满足的期望属性，并筛选出满足属性的Huber损失函数用于甘肃电力市场现货价格预测，还对市场因素的相关性进行了分析；通过实证研究，在MSE、MAE和Huber上训练了RNN神经网络，发现利用2024年1月1日至2024年6月30日期间甘肃电力现货价格数据在Huber上训练的模型比在MSE和MAE上训练的模型提供了更准确的日前电价预测，Huber函数是训练神经网络预测甘肃日前电力现货价格的最佳选择。而且Huber函数的回溯结果和稳定性比MSE和MAE都好，即利用Huber损失函数预测数据的精准性和稳定性更强。
Abstract: In response to the nonlinear characteristics of the bilateral spot market price data for high-proportion new energy in Gansu and the shortcomings of mainstream loss functions, this paper primarily investigates how to use the Huber loss function with neural networks to accurately model the time series of day-ahead electricity spot prices, aiming to systematically address the selection of the backpropagation loss function for day-ahead electricity spot price forecasting in Gansu. Firstly, the operational characteristics and issues of the Gansu electricity spot market are analyzed, and a neural network model for its electricity spot price is established; to optimize the model, the expected properties that the loss function should meet in the backpropagation algorithm are established, and the Huber loss function, which meets the properties, is selected for forecasting the spot price of the Gansu electricity market, and the correlation of market factors is analyzed; through empirical research, Recurrent neural networks (RNN) are trained on MSE, MAE, and Huber, and it is found that using the Gansu electricity spot price data from January 1, 2024, to June 30, 2024, the model trained on Huber provides more accurate day-ahead electricity price forecasts than the models trained on MSE and MAE. The Huber function is the best choice for training neural networks to predict day-ahead electricity spot prices in Gansu. Moreover, the retrospective results and stability of the Huber function are better than those of MSE and MAE; that is, the precision and stability of predicting data using the Huber loss function are stronger.
文章引用：马乐, 马寅, 吴锋, 刘丹丹, 崔剑, 魏博, 冯文韬, 韩凯莉. 基于RNN神经网络的甘肃电力现货价格有关损失函数的优化分析[J]. 计算机科学与应用, 2024, 14(10): 110-126. https://doi.org/10.12677/csa.2024.1410207

参考文献

[1] Xu, B. and Li, Y. (2024) Prescribed-Time Fully Distributed Nash Equilibrium Seeking of Nonlinear Multi-Agent Systems over Unbalanced Digraphs. Automatica, 169, Article ID: 111847. [Google Scholar] [CrossRef] 
[2] Kapoor, G. and Wichitaksorn, N. (2023) Electricity Price Forecasting in New Zealand: A Comparative Analysis of Statistical and Machine Learning Models with Feature Selection. Applied Energy, 347, Article ID: 121446. [Google Scholar] [CrossRef] 
[3] Aggarwal, S.K., Saini, L.M. and Kumar, A. (2009) Electricity Price Forecasting in Deregulated Markets: A Review and Evaluation. International Journal of Electrical Power & Energy Systems, 31, 13-22. [Google Scholar] [CrossRef] 
[4] Cinar, Y.G., Mirisaee, H., Goswami, P., Gaussier, E., Aït-Bachir, A. and Strijov, V. (2017) Position-Based Content Attention for Time Series Forecasting with Sequence-to-Sequence RNNs. 24th International Conference, ICONIP 2017, Guangzhou, 14-18 November 2017, 533-544. [Google Scholar] [CrossRef] 
[5] Anbazhagan, S. and Kumarappan, N. (2014) Day-Ahead Deregulated Electricity Market Price Forecasting Using Neural Network Input Featured by DCT. Energy Conversion and Management, 78, 711-719. [Google Scholar] [CrossRef] 
[6] Beigaite, R., Krilavicius, T. and Man, K.L. (2018) Electricity Price Forecasting for Nord Pool Data. 2018 International Conference on Platform Technology and Service (PlatCon), Jeju, 29-31 January 2018, 1-6. [Google Scholar] [CrossRef] 
[7] Guen, V. and Thome, N. (2019) Shape and Time Distortion Loss for Training Deep Time Series Forecasting Models.
[8] Fan, C., Zhang, Y., Pan, Y., Li, X., Zhang, C., Yuan, R., et al. (2019) Multi-Horizon Time Series Forecasting with Temporal Attention Learning. Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, Anchorage, 4-8 August 2019, 2527-2535. [Google Scholar] [CrossRef] 
[9] Lim, B. and Zohren, S. (2021) Time-Series Forecasting with Deep Learning: A Survey. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 379, Article ID: 20200209. [Google Scholar] [CrossRef] [PubMed]
[10] De Livera, A.M., Hyndman, R.J. and Snyder, R.D. (2011) Forecasting Time Series with Complex Seasonal Patterns Using Exponential Smoothing. Journal of the American Statistical Association, 106, 1513-1527. [Google Scholar] [CrossRef] 
[11] Sánchez-Sánchez, P.A., García-Gonzalez, J.R. and Coronell, L.H.P. (2019) Encountered Problems of Time Series with Neural Networks: Models and Architectures. Proceedings of the International Conference on Recent Trends in Artificial Neural Networks-Training and Prediction. [Google Scholar] [CrossRef] 
[12] Qi, M. and Zhang, G.P. (2001) An Investigation of Model Selection Criteria for Neural Network Time Series Forecasting. European Journal of Operational Research, 132, 666-680. [Google Scholar] [CrossRef] 
[13] Franses, P.H. and van Dijk, D. (2000). Nonlinear Time Series Models in Empirical Finance. Cambridge University Press.[CrossRef] 
[14] Egmont-Petersen, M., de Ridder, D. and Handels, H. (2002) Image Processing with Neural Networks—A Review. Pattern Recognition, 35, 2279-2301. [Google Scholar] [CrossRef] 
[15] Goldberg, Y. (2017) Neural Network Methods for Natural Language Processing. Synthesis Lectures on Human Language Technologies, 10, 1-309.
[16] 陈振寰, 杨春祥, 张柏林, 韩杰, 杨迎, 张天宇, 陈雨果. 甘肃电力现货市场双边交易机制设计[J]. 全球能源互联网, 2020, 3(5): 441-450.
[17] 王春玮. 甘肃省新能源现货市场应对策略研究[J]. 节能与环保, 2019(4): 84-85.
[18] 杨春祥, 张天宇, 张晓斌, 等. 适应高比例新能源电网的甘肃双边现货市场机制设计与运行分析[J]. 电网技术, 2022, 46(1): 63-69.
[19] Pan, J., Zhang, Z., Chu, S., Zhang, S. and Wu, J.M. (2024) A Parallel Compact Marine Predators Algorithm Applied in Time Series Prediction of Backpropagation Neural Network (BNN) and Engineering Optimization. Mathematics and Computers in Simulation, 220, 65-88. [Google Scholar] [CrossRef] 
[20] Bliemel, F. (1973) Theil’s Forecast Accuracy Coefficient: A Clarification. Journal of Marketing Research, 10, 444-446. [Google Scholar] [CrossRef] 
[21] Wang, Q., Ma, Y., Zhao, K. and Tian, Y. (2020) A Comprehensive Survey of Loss Functions in Machine Learning. Annals of Data Science, 9, 187-212. [Google Scholar] [CrossRef]