基于深度学习的火力发电炉膛监视自动化控制系统优化

Optimization of Automatic Control System for Furnace Monitoring in Thermal Power Plant Based on Deep Learning

摘要

火电锅炉炉膛处于高温含尘与强辐射环境，现场通常以炉膛火焰电视配合火检信号，由运行人员在DCS上综合判读着火、燃烧偏斜、结焦苗头与灭火风险，并联动给煤、二次风、燃尽风和油枪投退。煤质掺烧与深度调峰使火焰形态更易突变，传统阈值与人工经验在遮挡、反光与光照突变下容易误报或漏报。本文从图像获取、模型训练、报警判据和控制接口等提出可实施的深度学习优化路径，并强调模型在长期运行中的校核与冗余配置。

Abstract

Thermal power boiler furnaces operate in high-temperature, dust-laden, and radiation-intensive environments. On-site operators typically use flame television monitoring combined with fire detection signals to comprehensively assess ignition risks, combustion deviations, coking precursors, and fire suppression hazards on the Distributed Control System (DCS). This enables synchronized adjustments to coal feeding, secondary air supply, exhaust air, and oil gun deployment. Coal blending and deep peak shaving often cause abrupt flame morphological changes, while traditional threshold-based methods and human experience are prone to false alarms or missed detections under conditions of obstructions, reflections, or sudden light variations. This paper proposes actionable deep learning optimization strategies covering image acquisition, model training, alarm criteria, and control interfaces, with emphasis on long-term operational validation and redundant system configurations.

关键词

深度学习模型 / 火力发电 / 炉膛监视 / 自动化控制系统 / 优化

Key words

deep learning model / thermal power generation / furnace monitoring / automated control system / optimization

参考文献

[1] 赵焱. 基于深度学习的火电厂锅炉燃烧优化与氮氧化物减排协同控制研究[J]. 行车指南, 2023(5):0098-0099.

[2] 胡显辉. 基于深度学习的固废焚烧过程关键变量多步预测[D]. 大连海洋大学, 2024.

[3] 王化鲁, 谢发焕, 杜宇博. 基于深度学习的链条炉火床燃烧工况分析[J]. 数码设计, 2024(23):108-110.

[4] 张元正. 火力发电企业燃料智能化管理系统研究与应用[J]. 河南科技, 2022, 41(24):28-32.

[5] 吴进勇. 基于遗传算法的火电厂锅炉炉膛燃烧二次风挡板调节多目标优化策略[J]. 进展, 2025(15):137-139.