AI技术下复杂发电站运行风险视觉特征库构建

Construction of visual feature library for operational risk of complex power plants under AI technology

传统发电站运行风险特征库构建方法存在冗余特征量过多的问题,导致风险评估结果可信度不足,因此提出基于AI技术的复杂发电站运行风险视觉特征库构建方法。该方法整合文本条件向量与噪声向量,将整合结果作为风险观察的输入数据。引入AI技术进行图像特征映射和自主学习,并设计视觉可视化合成策略。计算风险估测值序列的算术平均数,采用相邻比较法确定变点的时刻和位置,收集总风险特征。将先验概率转换为后验概率,对独立风险特征进行分类处理,识别未知风险并贴上标签,从而完成风险视觉特征库的构建。实验结果表明,该方法能够提取与实际视觉特征库中数据一致的雷电流和电压幅值变化范围,为综合评估发电站运行状态提供了数据支持。

The traditional method for constructing a risk feature library for power plant operation has the problem of excessive redundant features,which leads to insufficient credibility of risk assessment results.Therefore,a visual feature library construction method for complex power plant operation risks based on AI technology is proposed.This method integrates text condition vectors and noise vectors,and uses the integration results as input data for risk observation.Introduce AI technology for image feature mapping and self-learning,and design visual visualization synthesis strategies.Calculate the arithmetic mean of the risk estimation value sequence,use the adjacent comparison method to determine the time and position of the change point,and collect the total risk characteristics.Convert prior probability to posterior probability,classify independent risk features,identify unknown risks and label them,thereby completing the construction of a risk visual feature library.The experimental results show that this method can extract the range of lightning current and voltage amplitude changes that are consistent with the actual visual feature library data,providing data support for the comprehensive evaluation of the operating status of power plants.