百度智能云AI助力电厂节煤降耗，年碳排减少一万吨

Baidu AI Cloud AI helps power plants save coal and reduce consumption, reducing annual carbon emissions by 10000 tons

山西省吕梁市汾阳市三泉镇，晋能集团旗下山西国峰煤电有限责任公司的两台300MW循环流化床直接空冷机组正在运行，燃煤通过传送带进入锅炉燃烧，将水加热成高温高压蒸汽，用以推动汽轮机拖动发电机旋转发电，支撑工业生产、点亮万家灯火。

In Sanquan Town, Fenyang City, Luliang City, Shanxi Province, two 300MW circulating fluidized bed direct air cooling units of Shanxi Guofeng Coal Power Co., Ltd., a subsidiary of Jinneng Group, are in operation. Coal is fed into the boiler through a conveyor belt for combustion, heating the water into high-temperature and high-pressure steam, which is used to drive the steam turbine to drive the generator to rotate and generate electricity, supporting industrial production and lighting up the lights of thousands of households.

“十四五”是碳达峰的关键期、窗口期，燃煤火电的碳减排，对实现电力行业碳排放达峰以及全国双碳目标具有重要意义，亟需通过实施改造升级进一步提升机组高效清洁、低碳灵活和智能化水平，助力煤电产业高质量转型发展。

The 14th Five Year Plan is a critical and window period for carbon peaking. The carbon emission reduction of coal-fired power plants is of great significance for achieving the peak carbon emissions in the power industry and the national dual carbon goals. It is urgent to further improve the efficiency, cleanliness, low-carbon flexibility, and intelligent water level of units through implementation of transformation and upgrading, and help the high-quality transformation and development of the coal power industry.

山西国峰煤电有限责任公司携手百度智能云，共同探索AI智慧运行应用场景，通过智能算法实现机组冷端经济运行与脱硫运行优化，供电煤耗综合降低可达2.9g/kWh，折算减少超过10600吨二氧化碳排放，降低电厂运行成本600万/年，成为智慧电厂和能耗优化示范标杆。

Shanxi Guofeng Coal Power Co., Ltd. joined hands with Baidu AI Cloud to jointly explore the application scenario of AI smart operation, and optimized the cold end economic operation and desulfurization operation of units through intelligent algorithms. The comprehensive reduction of coal consumption for power supply can reach 2.9 g/kWh, the reduction of more than 10600 tons of carbon dioxide emissions by conversion, and the reduction of power plant operation cost of 6 million yuan/year, becoming a model for smart power plants and energy consumption optimization.

AI降碳新高度

New heights of AI carbon reduction

我国发电量的60.8%来自于燃煤火电。火电厂生产工艺流程复杂且需要实时调控，是典型的多参数耦合多变的场景，传统的控制方式对于运行人员能力要求很高，很难做到最优运行，而人工智能技术在处理复杂运算和优化控制方面有很强的优势。

60.8% of China's electricity generation comes from coal-fired power plants. The production process of thermal power plants is complex and requires real-time regulation, which is a typical scenario of multi-parameter coupling and variability. Traditional control methods require high levels of operator capability and are difficult to achieve optimal operation. However, artificial intelligence technology has strong advantages in handling complex calculations and optimizing control.

冷端优化对于提高汽轮机组冷端性能，提升发电机组热经济性，从而达到节能降耗的目标具有重要意义。冷端系统的热量损失在热力系统中占比最大，是节能潜力最大的部分。直接空冷是北方缺水地区普遍采用的汽轮机排汽冷却方式，但也存在着背压经济运行缺乏数据支撑、空冷风机节能缺乏实时对比分析、空冷岛散热片脏污程度缺乏动态监视和可视化运行维护指导等问题。

Cold end optimization is of great significance for improving the cold end performance of steam turbine units, enhancing the thermal economy of generator units, and achieving the goal of energy conservation and consumption reduction. The heat loss of the cold end system accounts for the largest proportion in the thermal system and is the part with the greatest energy-saving potential. Direct air cooling is a commonly used steam turbine exhaust cooling method in water scarce areas in the north, but there are also problems such as a lack of data support for backpressure economic operation, a lack of real-time comparative analysis for energy saving of air cooling fans, and a lack of dynamic monitoring and visual operation and maintenance guidance for the dirty degree of air cooling island fins.

百度智能云以机组最大化净出力为目标，以机理建模为骨架，融合人工智能技术对冷端系统精确建模，并利用深度学习算法预测未来机组工况及外部环境，综合考虑冬季防冻、风机运行等安全边界条件，通过DCS闭环控制实时调节风机转速等参数，优化冷端系统整体运行效率，从而提高机组经济性、减少碳排放。

Baidu AI Cloud aims to maximize the net output of the unit, takes mechanism modeling as the skeleton, integrates artificial intelligence technology to accurately model the cold end system, and uses deep learning algorithm to predict the future unit working conditions and external environment, comprehensively considers the safety boundary conditions such as winter antifreeze, fan operation, etc., and optimizes the overall operating efficiency of the cold end system by real-time adjusting the fan speed and other parameters through DCS closed-loop control, so as to improve the unit economy Reduce carbon emissions.

国峰电厂空冷岛智能优化系统上线后，核算供电煤耗平均降低1.6g/kWh,单台机组每年可节约超过200万的燃料成本，减少超过5300吨的二氧化碳排放。

After the launch of the intelligent optimization system for the air cooling island of Guofeng Power Plant, the average coal consumption for power supply has been reduced by 1.6g/kWh. A single unit can save over 2 million yuan in fuel costs annually and reduce carbon dioxide emissions by over 5300 tons.

注：空冷优化系统供电煤耗节约量估算采用历史相似工况对比的方法，基于2021年7月至2022年7月的纯凝发电工况分钟级运行数据，以环境温度及机组负荷为工况指标，选择算法上线后的实时工况冷端净出力与历史相似工况对比，由此计算得到以上供电煤耗降低量。

Note: The estimation of coal consumption savings for power supply in the air cooling optimization system adopts the method of comparing historical similar operating conditions. Based on the minute level operating data of pure condensing power generation from July 2021 to July 2022, and using environmental temperature and unit load as operating indicators, the real-time net output of the cold end after the algorithm is selected to compare with historical similar operating conditions. Therefore, the above reduction in coal consumption for power supply is calculated.

热电联产提效能

Improving efficiency of cogeneration

热电联产是指在同一电厂中将供热和发电联合在一起，既生产电能，又利用汽轮发电机做过功的蒸汽对用户供热，可以显著提高燃料利用率，具有良好的经济和社会效益，是实现循环经济的重要技术手段。

Combined heat and power generation refers to the combination of heating and power generation in the same power plant, which not only produces electricity but also uses steam from steam turbines to heat users. It can significantly improve fuel utilization and has good economic and social benefits. It is an important technical means to achieve circular economy.

国峰电厂通过高背压供热改造提升了城市供热能力。然而机组高背压运行需要综合考虑机组负荷、供水温度、供水流量、低压缸及小汽机末级超温保护、精处理凝结水温度等因素，工况变化复杂，影响参数众多，难以确定最优抽汽比例与最优背压，造成供热能耗的大量损失。

Guofeng Power Plant has improved urban heating capacity through high back pressure heating renovation. However, high back pressure operation of the unit requires comprehensive consideration of factors such as unit load, water supply temperature, water supply flow rate, low pressure cylinder and small steam turbine final stage overtemperature protection, and fine treatment condensate temperature. The changes in operating conditions are complex, and there are many influencing parameters, making it difficult to determine the optimal steam extraction ratio and optimal back pressure, resulting in a significant loss of heating energy consumption.

百度智能云综合利用机器学习及机理建模方式，搭建了兼顾鲁棒性、可靠性、准确性的供热系统模型，利用智能优化算法确定安全经济背压并给与运行指导。在保证各设备安全约束的情况下，供热季平均背压提升1～2kPa，核算供电煤耗降低 1.3g/kWh，提升高背压凝汽器综合换热利用效率，降低抽汽量，从而提升供热经济性。

Baidu AI Cloud has comprehensively used machine learning and mechanism modeling methods to build a heating system model that takes into account robustness, reliability and accuracy. It uses intelligent optimization algorithms to determine the safety and economic backpressure and give operation guidance. Under the premise of ensuring the safety constraints of various equipment, the average back pressure during the heating season is increased by 1-2 kPa, and the calculated coal consumption for power supply is reduced by 1.3 g/kWh. This improves the comprehensive heat exchange efficiency of the high back pressure condenser, reduces the steam extraction volume, and thus improves the heating economy.

注：供热优化系统供电煤耗节约量同样采用历史相似工况对比的方法，基于2021年11月至2022年3月抽凝供热工况的分钟级历史运行数据，选取影响高背压凝汽器运行效率的核心指标，如供水温度、回水温度、供水流量等作为工况选择指标，算法自2022年11月供热季上线后，实时对比历史相似工况与机组实时优化运行的乏汽利用效率，基于实际采集数据统计了2022年供热季算法上线4个月的抽汽供热节约热量，以此计算得到了以上标煤节约量及年平均供电煤耗降低量。

Note: The coal consumption savings of the heating optimization system also use the method of comparing historical similar operating conditions. Based on the minute level historical operating data of the condensing and heating conditions from November 2021 to March 2022, the core indicators that affect the operating efficiency of high back pressure condensers are selected, such as water supply temperature, return water temperature, and water supply flow rate, as the operating condition selection indicators. The algorithm has been in operation since the heating season in November 2022, Real time comparison of the exhaust steam utilization efficiency between historical similar operating conditions and real-time optimized operation of the unit was conducted. Based on actual collected data, the heat savings of the extraction heating algorithm in the 2022 heating season for four months were calculated, and the above standard coal savings and annual average reduction in power supply coal consumption were calculated.

智能减排新突破

New breakthroughs in intelligent emission reduction

节能减排不仅是降低碳排放，对于火电厂，SO2等污染物治理仍是重点，政策对排污指标要求越来越严格，对燃煤电厂的环保工程发展提出严格要求。燃煤电厂的脱硫、脱硝和除尘已成为中国环保关注的主要领域之一。

Energy conservation and emission reduction are not only aimed at reducing carbon emissions, but also at controlling pollutants such as SO2 in thermal power plants. Policies have increasingly strict requirements for emission indicators, and strict requirements have been put forward for the development of environmental protection projects in coal-fired power plants. The desulfurization, denitrification, and dust removal of coal-fired power plants have become one of the main areas of environmental protection concern in China.

国峰电厂循环流化床机组采用炉内石灰石脱硫和炉后半干法脱硫两种工艺。由于锅炉燃烧与脱硫反应的大时延、煤质多变且工艺间耦合关系强等原因，导致脱硫剂用量难以准确判断，无法实现自动化运行，影响环保指标达成。

The circulating fluidized bed unit of Guofeng Power Plant adopts two processes: limestone desulfurization inside the furnace and semi dry desulfurization after the furnace. Due to the large delay in boiler combustion and desulfurization reaction, the variability of coal quality, and strong coupling between processes, it is difficult to accurately determine the amount of desulfurization agent and achieve automated operation, which affects the achievement of environmental protection indicators.

百度智能云深入电厂生产一线，基于机理+数据驱动融合建模方式，定期对煤质硫分和延迟时间进行估计，指导深度调峰/调频下机组的SO2的生成与脱除过程预测；其次，自适应控制按照最优比例分配炉内外的脱硫剂消耗量，并进行秒级精准调节；此外，还实现了脱硫剂输送过程的故障智能预警，修正控制策略，确保控制算法执行的可靠性。从而最大限度减少脱硫剂用量，保障系统运行的经济性和可靠性。

Baidu AI Cloud goes deep into the production line of the power plant, regularly estimates the sulfur content of coal and delay time based on the mechanism+data driven fusion modeling method, and guides the prediction of SO2 generation and removal process of units under deep peak shaving/frequency modulation; Secondly, adaptive control distributes the consumption of desulfurizer inside and outside the furnace according to the optimal proportion, and makes precise adjustments in seconds; In addition, it has also achieved intelligent fault warning for the desulfurization agent transportation process, corrected control strategies, and ensured the reliability of control algorithm execution. To minimize the amount of desulfurizer used and ensure the economy and reliability of system operation.

脱硫控制算法学习数据来源于电厂2022年全年的秒级运行数据，在2023年3月开始投运上线，实现自动投运率达到96%以上，无需人工干预，脱硫岛入口烟气SO2小时平均值（炉内石灰石脱硫）与设定值平均偏差小于30mg/Nm3，脱硫岛出口SO2小时平均值（脱硫岛半干法脱硫，实际环保指标）在20±10mg/Nm3内；选择全年相似工况对比，并将煤质硫分折算为统一标准，单位发电量的脱硫剂平均可节省8%，折合约4万吨脱硫剂，每台机组的物耗节省可达200万元，仅使用AI算法进行脱硫控制，大幅减少了电厂对人工运维的依赖，并且在满足环保指标（35mg/m3）排放要求的同时，有效地提升了烟气处理的经济性。

The learning data of desulfurization control algorithm comes from the second level operation data of the power plant for the whole year of 2022. It was put into operation in March 2023 and achieved an automatic operation rate of over 96% without manual intervention. The hourly average value of SO2 in the flue gas at the inlet of the desulfurization island (limestone desulfurization in the furnace) deviates from the set value by less than 30mg/Nm3, and the hourly average value of SO2 at the outlet of the desulfurization island (semi dry desulfurization in the desulfurization island, actual environmental indicators) is within 20 ± 10mg/Nm3; By comparing similar operating conditions throughout the year and converting the sulfur content of coal into a unified standard, the desulfurization agent per unit of power generation can save an average of 8%, equivalent to about 40000 tons of desulfurization agent. The material consumption of each unit can be saved up to 2 million yuan. Only using AI algorithm for desulfurization control greatly reduces the dependence of power plants on manual operation and maintenance, and effectively improves the economy of flue gas treatment while meeting the emission requirements of environmental protection indicators (35mg/m3).

下沉控制一区，AI算法完成闭环控制

Sinking control zone 1, AI algorithm completes closed-loop control

传统上做工艺优化，通常只是做参数推荐，系统投运率低，效果不明显；工艺优化的系统数据采集与数据反控，通常通过SIS系统等部署在二区的系统来进行交互，数据闭环延时大，算法控制难度大，精度低。

Traditionally, process optimization is only done through parameter recommendations, resulting in low system operation rates and ineffective results; The system data collection and data reverse control for process optimization are usually interactive through systems deployed in Zone 2, such as SIS systems. The data closed-loop delay is large, the algorithm control difficulty is high, and the accuracy is low.

本项目将AI算法下沉到控制一区，数据从采集到反控，闭环周期降低至秒级，大大降低了算法控制难度，提高了控制精度，同时考虑到控制一区对稳定性、安全性的较高要求，我们在硬件、软件、策略上均做了相应防护，通过搭建服务器集群、容器化服务部署、服务间状态的主、被动检测以及与原有控制系统的控制逻辑进行配合，建立冗余自动切换的控制逻辑，实现系统在硬件、软件、控制策略上的多级冗余机制，确保系统的稳定性与安全性。

This project sinks the AI algorithm into the control zone 1, reducing the closed-loop cycle from data collection to reverse control to seconds, greatly reducing the difficulty of algorithm control and improving control accuracy. At the same time, considering the high requirements for stability and security in the control zone 1, we have taken corresponding measures in hardware, software, and strategy, by building server clusters, container service deployment, and controlling the state of the service room Passive detection and coordination with the control logic of the original control system to establish a redundant automatic switching control logic, achieving multi-level redundancy mechanisms in hardware, software, and control strategies, ensuring the stability and security of the system.

“双碳”战略下，以新能源为主体的新型电力系统的构建正加速传统火电企业转型升级，技术革新变得更加紧迫，亟需通过实施改造升级进一步提升机组高效清洁、低碳灵活和智能化水平，助力煤电产业高质量转型发展。

Under the "dual carbon" strategy, the construction of a new type of power system with new energy as the main body is accelerating the transformation and upgrading of traditional thermal power enterprises. Technological innovation has become more urgent, and it is urgent to further improve the efficiency, cleanliness, low-carbon flexibility, and intelligence level of units through the implementation of transformation and upgrading, to help the high-quality transformation and development of the coal power industry.

“我们和百度智能云合作，基于电厂运行机理与运行规范，采用机理模型与人工智能模型融合驱动的方法，实时计算各工况下影响机组优化运行的参数，并通过提供指导建议和自动闭环控制的方式，帮助电厂优化运行与节能降碳，创新发展火力发电技术，解决煤炭清洁高效利用问题。”国峰煤电工程师表示。

"We cooperated with Baidu AI Cloud, based on the operation mechanism and operation specification of the power plant, and adopted the method of integration and driving of mechanism model and artificial intelligence model, to calculate the parameters that affect the optimal operation of units under various working conditions in real time, and help the power plant to optimize operation, save energy and reduce carbon, innovate and develop thermal power generation technology, and solve the problem of clean and efficient utilization of coal by providing guidance suggestions and automatic closed-loop control." Guofeng Coal Electric Engineer stated.