In order to optimize the control effect of the boiler fans in cogeneration power plants and improve the combustion efficiency and energy efficiency level，a frequency prediction model suitable for boiler fans was developed. Two prediction models were constructed using the fan performance method and the boiler operating condition method， and the two models were verified based on the actual data. The results show that the boiler operating condition method has higher accuracy in fan frequency prediction compared with the fan performance method. The frequency prediction relative errors of the primary fan and the secondary fan have decreased to 0.03% and 0.39%，respectively. Therefore， the prediction model established by the boiler operating condition method can provide more reliable fan frequency prediction results for the operators of cogeneration power plants，which is conducive to achieving the optimal control of boiler fans operation，reducing fan energy consumption，and improving the overall production efficiency of cogeneration power plants.

Aiming at the problems of low initial discharge field strength and easy occurrence of corona in stator coil bars and windings under high-altitude conditions，the research on the anti-corona performance of stator coil bars and windings was conducted. Firstly，the impact of different altitudes on the initial discharge field strength at the end of stator coil bars was analyzed. Secondly，the key parameter ranges of anti-corona materials，the basic requirements for the optimization design of anti-corona structures，the field strength distribution at the corners of stator coil bar conductors，and the equipotential layer treatment technology in the slot section of conductors were introduced. Finally， anti-corona technologies for the end of stator windings were presented，and the influence of the ground clearance at the end of stator windings on end discharge was analyzed. The research results indicate that the anti-corona technologies for individual stator coil bars and stator windings have laid a solid foundation for enhancing the anti-corona performance of stator coil bars and windings in high-altitude impulse hydrogenerator units.

With the development of thermal power units towards higher parameters and larger capacities，the amount of data that needs to be monitored during the operation of turbo-generator set is increasing. The widely used expert diagnosis systems have severely impacted fault discrimination accuracy. To address this problem，a new steam turbine fault diagnosis method based on deep learning technology was proposed. Firstly，the original signals were reconstructed and output by using the autoencoder network. Secondly，the convolutional neural network（CNN）was employed to reduce the number of input parameters in the fully connected layers through convolution and pooling operations，while extracting data features and combining them into deeper，higher-dimensional data. Finally，the fully connected layers were used for output prediction and classification. Engineering application results demonstrate that this method can adaptively extract features of different fault locations and types in steam turbine systems under various operating conditions，accurately identifying equipment health status. Its discrimination accuracy is significantly higher than that of the currently widely used expert diagnosis systems.

Establishing an accurate and efficient integrated energy system model on the source side of thermal power plants is crucial for optimizing the system management process and improving the safety and economy of operation，and is an important reference for the planning and transformation of thermal power plants. A digital twin operation and maintenance model of the integrated energy system on the source side of thermal power plants based on feedforward neural networks was proposed. This system took the thermal power unit as the core，coupled three types of renewable energy，namely biomass gasification，waste gasification and dried sludge，at the input end，and supplied four types of energy products，namely cold，heat，steam and electricity，at the output end. The feedforward neural network was trained using the computational data provided by the physical model，and its optimal hyperparameters were calculated. The corresponding neural network model was constructed，and finally a digital twin model was formed. Through comparison and verification with the physical model，it is found that this digital twin model has good calculation accuracy and high calculation efficiency，and can complete the prediction within milliseconds.

In order to achieve the economic evaluation and decision-making guidance of peak regulation in the operation mode of cogeneration units coupled with electric boilers，a peak regulation decision-making model for the unit was established. Through big data tools such as data mining technology and back propagation （BP） neural network，a prediction model for energy consumption indicators of cogeneration units was constructed to achieve precise calculation of various cost and revenue indicators of the unit revenue model. Meanwhile，an economic evaluation method for peak regulation of thermal power units was proposed. It comprehensively considered the coupled operation modes such as peak regulation，steam extraction heating，and electric boilers. By calculating the revenue of the full-load operation in real time，it provided guidance for peak regulation decisions based on the optimal operating income. The results show that when the heat load is higher than the critical value，the operation mode of giving priority to starting the electric boiler of the unit is more economical. The critical heat load of the unit is determined by the ratio of the unit price of standard coal to the heat price. During the heating period，to maximize the revenue of the electric boiler，the unit should operate below the critical heat load. The critical peak regulation compensation correction coefficient of the unit is mainly affected by the heat load and the non-peak regulation allocation electricity price. Referring to this coefficient to guide the peak regulation decision-making of the cogeneration units can maximize the operating income of the unit.

Under the dual carbon policy framework， the new power system introduces new flexibility requirements for coal-fired power units，including rapid start-stop operations and deep peak regulation. During deep peak regulation，the rapid load fluctuations of units lead to substantial variations in water-steam flow rates and steam parameters. This significantly complicates the maintenance of water-steam quality control and increases the risks of stress corrosion fatigue on thermal equipment，thereby threatening the safe and stable operation of the units. A systematic analysis was conducted on how the deep peak regulation process affects steam-water parameters and quality control，as well as corrosion and scaling in thermal equipment，and the underlying causes of these issues were examined. Six major problems were summarized，including inadequate feedwater control and increased steam carryover，corrosion and scaling of the furnace heating surfaces and hydrodynamic instability，abnormal fluctuations in steam temperature，intensified water erosion of the final stage blades of the steam turbine，excessive dissolved oxygen in condensate water and feedwater，and lagging regulation of the chemical dosing system. The proposed multi-dimensional risk prevention measures provide theoretical guidance and practical reference for the optimal control of water-steam systems in coal-fired power enterprises during deep peak regulation operations.

To address the hydrodynamic instability and combustion disturbances of a 600 MW supercritical oncethrough boiler at Ligang power plant during deep peak regulation down to 20% rated load，a hydrodynamic calculation model for the water wall and a coupled simulation platform for the combustion system were established based on the actual boiler structure and operational parameters. Through the combination of numerical simulations and field tests， the mass velocity distribution characteristics，wall temperature deviation，and pulsation risks under low-load conditions were quantitatively analyzed. A dynamic control strategy based on the optimization of coal mill combinations was proposed. The results show that adopting the mid-upper coal mill combination（B+C+D）with a staged air distribution strategy can reduce the water wall mass velocity deviation to 18.7%. The research findings provide a theoretical basis for the deep peak regulation of supercritical units.

In order to solve the problem of reheat steam temperature deviation during the comprehensive upgrading and transformation for a unit，the inter-tube panel flow distribution curve of the high-temperature reheater before the transformation was obtained through mathematical modeling. Combined with the measured values of the operating wall temperature，the heat exchange deviation on the flue gas side was calculated. Under the new transformation boundary，based on the known deviation of the flue gas side，the structural form of the header was renovated and optimized，and the optimization plan was determined. Results show that a high heat-transfer zone exists slightly right of the furnace center because of the residual swirl at the furnace exit and the size of the imaginary circle. By reducing the temperature rise amplitude of the high-temperature reheaters and adjusting the positions of the radial inlet and outlet tees，the temperature deviation level of the reheat steam can be significantly reduced，achieving the transformation goal.

The CAP1000 unit is a representative of advanced pressurized water reactor nuclear power technology independently developed in China. The overspeed protection system of its steam turbine is the core to ensure the safe operation of the unit. To identify an optimal solution，the traditional "mechanical and electrical overspeed" scheme and the innovative "dual electrical overspeed" scheme were compared，and an analysis was conducted from the aspects of technical compliance，safety redundancy，operation and maintenance efficiency，and economy，supported by a case study. The results demonstrate that upgrading from mechanical overspeed protection to independent electrical overspeed protection conforms to the technical trends and relevant standards of the nuclear power industry. It enhances protection response speed，simplifies operation and maintenance，and delivers both safety and economic benefits. The research findings can serve as a reference for the design and retrofit of CAP1000 units.

In order to enhance the competitive advantage of industrial steam turbines in the thermoelectric field， increasing the pressure in the turbine impeller chamber can effectively improve the internal efficiency of the entire machine. However，the increase of impeller chamber pressure poses higher requirements for the strength and low cycle fatigue life of the high pressure inner cylinder. Therefore，the finite element method and the theory of low cycle fatigue life were adopted to analyze the strength and low cycle fatigue life of the high pressure inner cylinder. The research results show that stress concentrations are prone to appearance at the connection points of the inner cylinder structure and the areas with sudden shape changes，leading to fatigue damage. The research results can provide a theoretical basis for the design and optimization of the inner cylinder of industrial steam turbines.

In photovoltaic power stations，high-power string inverters often employ forced air cooling as a heat dissipation method to control the operating temperature of the equipment. During actual operation，the air inlets of these inverters are prone to being blocked by dust， willow catkins and other impurities in the surrounding environment，which can lead to thermal failure. To address this issue，a numerical simulation method was adopted for analysis and research. A numerical model with engineering accuracy was constructed，and based on the simulation results，a self-cleaning solution that takes into account the heat dissipation performance was designed. The results show that this solution not only ensures the heat dissipation performance of the inverter under normal operation but also has strong self-cleaning ability for the air inlets. Field test results have verified the feasibility and actual operation effect of the solution，providing an efficient solution for the problem of air inlet blockage in in-service inverters in photovoltaic power stations and having significant application reference value.

Experimental study was carried out to investigate the effect of H₂O on the mercury removal performance of modified rice straw coke under conventional combustion conditions. Experiments were conducted to get rice straw coke particles prepared from rice straw in Jiangsu region and modified by chemical impregnation of NH₄Cl combined with HNO₃. A fixed-bed experimental platform was utilized to simulate the flue gas environment of coalfired power plants，through which the mercury removal effect of the modified rice straw coke was tested. The results show that the addition of H₂O inhibits the mercury removal effect of modified rice straw coke under conventional combustion atmosphere（composed of N₂ and 6% O₂）. Further analysis reveals that the increase of H₂O concentration leads to a gradual decrease of mercury adsorption，and the mercury adsorption forms on the surface of coke samples are mainly Hg⁰ and HgO under the H₂O-containing atmosphere，without changing the main adsorption forms. However，H₂O dissociates the oxidizable OH，weakens the Hg⁰ competitive adsorption，and provides additional electrons for the reduction of Hg²⁺. Excessive H₂O concentration will exacerbate the segregation of Hg⁰ and CO₂， leading to the decomposition of unstable oxidation products such as HgCl₂，thus inhibiting the overall Hg removal effect of the adsorbent.