摘要：

大型换流站调相机外冷水系统腐蚀结垢直接影响冷却器的热力性能，开展冷却水结垢机理研究和结垢预测研究对保障电网安全运行具有重要意义。对调相机外冷水结垢机理和现有结垢预测模型进行探讨，在此基础上提出以深度学习算法进一步建立精确的垢生长数学模型。首先对某大型特高压换流站冷却水系统运行数据进行预处理，得到3 250组有效数据样本，之后以该数据集进行训练，采用反向传播（BP）神经网络和机器学习算法建立循环冷却水系统的结垢预测模型，并对该模型的预测精准度进行评价。结果表明，训练后的模型能够有效预测结垢量，总体平均相对百分比误差（MAPR）在7.53%以下，决定系数（ R ²）为0.985，具备良好的拟合效果以及泛化能力。

关键词: 循环冷却水系统, 阻垢模型, 结垢机理, BP神经网络, 机器学习

Abstract:

The corrosion and scaling of the external cooling water system of large converter station phase modulator direcbly affect the thermal performance of the cooler. Research on the scaling mechanism of cooling water and scaling prediction is of great significance to ensure the safe operation of the power grid. This paper discussed the scaling mechanism of the external cooling water system of the phase modulator and the existing scaling prediction models, and proposed to further establish an accurate scaling growth mathematical model by deep learning algorithms. Firstly, operational datas from the cooling water system of a large ultra-high voltage converter station were preprocessed, resulting in 3 250 valid data samples. Based on this dataset, a scale prediction model of the circulating cooling water system was trained using backpropagation(BP) neural networks and machine learning algorithms, and the accuracy of the model was evaluated. The results showed that the trained model effectively predicted the amount of scaling. The mean absolute percentage error(MAPE) and the coefficient of determination( R²) were 7.53% and 0.985 respectively, indicating good fitting performance and generalization ability of the model.

Key words: circulating cooling water system, scale inhibition model, scaling mechanism, BP neural network, machine-learning

中图分类号: 

• X52