Supercritical unit low-load deep peak shaving system and method

Abstract

The invention provides a supercritical unit low-load deep peak regulation system and method, and relates to the field of thermal generator sets, and the supercritical unit low-load deep peak regulation system comprises an instruction analysis and pre-control module, an energy redistribution module, a combustion steady-state module, a parameter cooperation module and a state backtracking and self-gain module. According to the application, through the arranged instruction parsing and pre-control module, the power grid load instruction is parsed and converted into a set of accurate equipment action time sequence list, the instruction feature recognition unit in the instruction parsing and pre-control module performs mode recognition on the load instruction, and the strategy generation unit calls the preset strategy template according to the mode, so that the power grid load instruction is analyzed and converted into the accurate equipment action time sequence list. Action instructions which face all execution units of the whole system and are arranged in a wrong sequence on a time axis are generated, and conversion from passive response to active pre-control is achieved.

Description

A deep peak-shaving system and method for low-load supercritical units Technical Field

[0001] This invention relates to the field of thermal power generating units, and more specifically, to a deep peak-shaving system and method for supercritical units under low load. Background Technology

[0002] Against the backdrop of energy structure transformation, thermal power units need to undertake deep peak shaving tasks to enhance the grid's ability to absorb new energy sources. However, supercritical units face severe challenges under low-load conditions, such as difficulties in stable boiler combustion, instability of key parameters (such as steam temperature and steam pressure), and a prominent contradiction between rapid load response and equipment safety. Traditional control methods typically use relatively isolated subsystems to address local issues such as combustion, steam and water supply, and heating, lacking a global coordination strategy. This results in limited peak shaving depth, slow response rate, and often requires oil injection for combustion assistance under extreme loads, increasing operating costs and risks.

[0003] Therefore, we have made improvements to this and proposed a deep peak-shaving system and method for low-load supercritical units. Summary of the Invention

[0004] The purpose of this invention is to solve the problem that current control methods often lack a global collaborative strategy.

[0005] To achieve the above-mentioned objectives, the present invention provides the following deep peak-shaving system and method for supercritical units at low loads, in order to improve the aforementioned problems.

[0006] This application specifically describes a deep peak-shaving system for supercritical units under low load, comprising: an instruction parsing and pre-control module for parsing load instructions into a list of equipment action sequences; an energy redistribution module for executing instructions in the equipment action sequence list and rapidly adjusting power generation by regulating the distribution of steam energy flow; a combustion steady-state module for maintaining furnace combustion stability under low load through optical sensing and active air distribution; a parameter coordination module for evaluating the overall system stability margin and dynamically adjusting the control priority of different parameters; and a state backtracking and self-gain module for storing peak-shaving cases, matching historical strategies, and optimizing control parameters; wherein the instruction parsing and pre-control module is communicatively connected to the energy redistribution module, the combustion steady-state module, the parameter coordination module, and the state backtracking and self-gain module to collaboratively achieve the deep peak-shaving process.

[0007] As a preferred technical solution of this application, the instruction parsing and pre-control module includes: an instruction feature recognition unit, used to classify instructions into different peak shaving modes by calculating the rate of change of load instructions and comparing it with a preset threshold; a strategy generation unit, used to call the corresponding strategy template according to the peak shaving mode, and generate a list of equipment action sequence containing specific action timing and parameters based on the current load and the target load; and an efficiency evaluation unit, used to compare the actual operating data with the expected target of the equipment action sequence list after the peak shaving task is completed, and generate fine-tuning suggestions for optimizing the strategy template.

[0008] As a preferred technical solution of this application, the energy redistribution module includes: a path control unit, used to calculate and output a path adjustment factor Kp according to the target power generation value, so as to continuously adjust the steam distribution ratio to the low-pressure cylinder and the heating network, wherein Kp=(P_current-P_target) / P_range, and the value range of Kp is [-1,1]; a heating network coupling unit, used to calculate the additional heat load capacity that the heating network can accept in real time, and coordinate with the path control unit when it operates, and send an early warning signal to the command parsing and pre-control module when the capacity is close to saturation; wherein, the adjustment rate of the path control unit is limited by the thermal stress index output by the parameter coordination module.

[0009] As a preferred technical solution of this application, the combustion steady-state module includes: a spectral sensing unit, used to calculate a combustion stability index R by analyzing the spectral intensity and fluctuation characteristics of specific chemical free radicals in the furnace, and to perform local instability positioning, where R=α*(I_OH / I_avg)+β*(1-σ_CH / μ_CH); and an air distribution optimization unit, used to receive the combustion stability index R and local instability positioning information, and to activate a global combustion stability strategy when the R value is lower than a safety threshold, and to activate a local intervention strategy for upstream adjacent burners when a weak combustion stability zone exists; wherein, the air distribution optimization unit receives a feedforward signal from the energy redistribution module and acts in advance before the energy flow changes.

[0010] As a preferred technical solution of this application, the parameter coordination module includes: a state evaluation unit, used to receive and integrate the combustion stability index R from the combustion steady-state module, the path adjustment factor Kp from the energy redistribution module, and the main steam pressure and temperature deviation, to calculate a system stability margin S, where S=w1*f(R)+w2*(1-|ΔP| / P_set)+w3*(1-|ΔT| / T_set)-w4*|Kp|, and to identify the dominant conflicting parameter; and a priority decision unit, used to dynamically switch between different control strategy templates according to the system stability margin S and the dominant conflicting parameter, so as to adjust the control dead zone and computational resource allocation of different parameters.

[0011] A method for deep peak shaving at low load in supercritical units includes the following steps: Instruction deep parsing and strategy pre-generation step: parsing AGC instruction characteristics and generating a list of equipment action timings; Energy flow forward-looking transfer step: executing the list and allocating power generation and heating energy by adjusting the steam path; Combustion state perception and combustion stabilization step: evaluating and actively maintaining combustion stability based on spectral signals; System stability margin decision step: fusing multi-source parameters to evaluate system stability and dynamically adjusting control priorities; Case backtracking and self-gain step: storing peak shaving process data and using historical cases to optimize subsequent strategies.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: In the solution of this application: 1. By setting up an instruction parsing and pre-control module, the power grid load instruction is parsed and transformed into a precise equipment action sequence list. The instruction feature recognition unit in the instruction parsing and pre-control module performs pattern recognition on the load instruction, and the strategy generation unit calls the preset strategy template according to the pattern to generate action instructions arranged in a staggered order on the time axis for all execution units of the entire system, realizing the transformation from passive response to active pre-control, and solving the problem that the control method in the prior art usually lacks a global collaborative strategy; 2. By setting up an energy redistribution module, the rapid and flexible decoupling and redistribution of power generation and heating energy is realized. The path control unit continuously adjusts the path to low-voltage areas by dynamically calculating and adjusting the "path adjustment factor Kp". The ratio of steam flow between the pressure cylinder and the heating network, instead of the traditional rigid cut-off, and the heat network coupling unit evaluates and utilizes the heat storage capacity of the heat network as a buffer, solves the problems of the impact of the thermoelectric decoupling process on the steam turbine in the existing technology, and the inability to fully utilize the capacity due to the limitation of the heating system during rapid peak shaving; 3. Through the set combustion steady-state module, active combustion stabilization control based on real-time diagnosis of combustion status is realized. The spectral sensing unit calculates the "combustion stabilization index R" that can quantify the degree of combustion stability by analyzing the free radical spectrum in the furnace and locates the local weak combustion zone. The air distribution optimization unit performs "differential air volume compensation" accordingly to accurately intervene in the weak combustion zone, which solves the problems of the existing technology relying on simple flame detection signals, being unable to predict combustion trends, and having a single and ineffective means of stabilizing combustion under extremely low loads, which has to rely on oil injection for combustion assistance. Attached Figure Description

[0013] Figure 1 is a system flowchart of the supercritical unit low-load deep peak shaving system provided in this application; Figure 2 is a system flowchart of the command parsing and pre-control module and the state backtracking and self-gain module in the supercritical unit low-load deep peak shaving system provided in this application; Figure 3 is a system flowchart of the energy redistribution module in the supercritical unit low-load deep peak shaving system provided in this application; Figure 4 is a system flowchart of the combustion steady-state module in the supercritical unit low-load deep peak shaving system provided in this application; Figure 5 is a system flowchart of the parameter coordination module in the supercritical unit low-load deep peak shaving system provided in this application; Figure 6 is a system flowchart of the state backtracking and self-gain module in the supercritical unit low-load deep peak shaving system provided in this application. Detailed Implementation

[0014] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0015] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0016] It should be noted that, unless otherwise specified, the embodiments and features and technical solutions in the embodiments of the present invention can be combined with each other.

[0017] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0018] Example 1 (Refer to Figures 1, 2, 3, 4, 5, and 6) illustrates a deep peak-shaving system for a supercritical unit under low load. This system includes: an instruction parsing and pre-control module, whose core function is to transform abstract load instructions issued by the upper-level dispatching system into a specific, executable sequence of equipment actions with minimal impact on the unit, and to issue precise instructions to other modules to perform specified actions at designated times, thereby optimizing the peak-shaving process; and an energy redistribution module, whose core function is to respond to the strategies issued by the instruction parsing and pre-control module, quickly and accurately adjust the power generation by changing the energy flow distribution within the unit, while ensuring stable heating and ensuring that this conversion process has no impact on the main equipment. The combustion steady-state module ensures the safe and stable operation of the unit under low-load conditions. Through sensing technology and proactive intervention strategies, it actively shapes and maintains a superior combustion environment over a wide load range, thereby overcoming the technical bottleneck of stable combustion without combustion at 30% ECR. The parameter coordination module can assess the unit's status and determine its operational behavior at specific times, thereby resolving the inherent contradictions between different control objectives and ensuring global optimization of the peak-shaving process. The status backtracking and self-gain module systematically collects, stores, and analyzes each deep peak-shaving process to form reusable "knowledge" and uses this knowledge to optimize future operations, enabling the entire system to learn from experience and continuously improve its performance.

[0019] Furthermore, as shown in Figures 1 and 2, the instruction parsing and pre-control module includes: an instruction feature recognition unit, which has a built-in sliding time window (e.g., 10 seconds in length) to calculate the first derivative (rate of change, unit: %Pe / min) and second derivative (acceleration of change) of the load instruction within the window in real time; simultaneously, it records the final target load value of the instruction; based on preset thresholds, the instruction is classified into one of the following typical modes: emergency load reduction mode: rate of change < -2.5%Pe / min; slow load reduction mode: -0.8%Pe / min ≥ rate of change ≥ -2.5%Pe / min; steady-state maintenance mode: absolute value of rate of change < 0.8%Pe / min; load increase mode: rate of change > 0.8%Pe / min; data output: the instruction feature recognition unit outputs a structure containing... It includes three key segments: command mode, target load, and rate of change. The strategy generation unit stores multiple predefined strategy templates for different "command modes." These templates are essentially blank timelines defining the start time and sequence of different equipment actions. The strategy generation unit's job is to fill these templates with the specific parameters (target load, rate of change) of the current peak-shaving task. Taking "emergency load reduction to 30%" as an example, the strategy generation process is as follows: Template selection: Match the "emergency load reduction mode" template; Timeline filling: T+0 seconds: Send a command to the path control unit of the energy redistribution module: Activate "steam path flexible guidance," reducing the target power generation by 60%; T+0.2 seconds: Send a command to the air distribution optimization unit of the combustion steady-state module: Activate "heat storage buffer" mode, offsetting the furnace pressure setpoint by +X. Pa; T+45 seconds: Send an instruction to the priority decision unit of the parameter coordination module: switch to the "maintain pressure - stabilize combustion" priority strategy; T+60 seconds: Send an instruction to the mill management unit of the fuel optimization module: execute the B-level coal mill shutdown preparation; T+240 seconds: Send an instruction to the mill management unit of the fuel optimization module: execute the B-level coal mill shutdown; Quantitative calculation of the strategy: The specific parameters in the list (such as "target power generation reduction of 60%)" are calculated in real time through a simple weighted scoring model; This model comprehensively considers the rate of change, the current unit status (such as the boiler heat storage level), and the target load; For example, the power reduction target = (current load - target load) * response coefficient K; where the response coefficient K is a function of the rate of change, and when the rate is extremely fast, K can be 1.2. The initial power reduction target is slightly higher than the final target to reserve space for a smooth transition later; Data output: This unit outputs a complete "equipment action sequence list", which clearly lists the action commands and their parameters to be triggered at absolute or relative time points, and specifies the specific units of the downstream modules that receive these commands; The strategy generation unit directly sends action commands to the path control unit, realizing precise start-up and shutdown and intensity control of the "electrothermal decoupling" action, which is the first step in quickly responding to load commands; While issuing the rapid load reduction command, the strategy generation unit will immediately notify the air distribution optimization unit to advance the action. Entering a "heat storage buffer" state avoids combustion instability caused by rapid energy transfer. The strategy generation unit commands the priority decision unit to switch control strategies at different stages of the peak shaving process. For example, it commands the unit to "maintain pressure" at the beginning of load reduction and to "maintain temperature" when approaching the target load. This allows parameter control to adapt to the main contradictions at different stages. The performance evaluation unit is activated after each peak shaving task. The performance evaluation unit retrieves the actual operating data of this task from the DCS historical database (such as the actual load change curve, main steam pressure fluctuation range, maximum furnace negative pressure, etc.). The system compares the data with the expected targets in the "Equipment Action Sequence List" issued by the strategy generation unit. The performance evaluation unit first collaborates with the case library storage unit of the state backtracking and self-gain module to package the complete dataset of this peak-shaving process (including command characteristics, generated strategies, and actual effects) into a new case for storage. The performance evaluation unit quantifies the execution effect of the strategy; for example, calculating indicators such as "the degree of agreement between the actual load decrease rate and the target rate" and "main steam pressure overshoot." If the timing or parameters of a certain action command are found to be consistently (continuously)... If performance indicators deviate from ideal values ​​(e.g., excessive main steam pressure disturbances occur every time the coal mill stops), the efficiency evaluation unit will generate a fine-tuning suggestion. This suggestion will be fed back to the strategy generation unit to optimize its internal "strategy template." For example, the original template's "stop mill at T+60 seconds" might be revised to "stop mill at T+75 seconds," resulting in a smoother transition in similar future operating conditions. When initializing a new strategy, the strategy generation unit will also prioritize querying the case library, requesting similar successful cases from the case matching and calling unit as reference templates, thereby improving the initial quality of the strategy.

[0020] Furthermore, as shown in Figures 1 and 3, the energy redistribution module includes: a path control unit, which continuously monitors the steam flow to the low-pressure cylinder and the extraction steam flow to the heating network. Its core control target is a "power generation target value" defined by upstream instructions. Internally, based on the difference between the current actual power and the target power, the unit continuously adjusts the steam distribution ratio of the two channels through a core variable called "path adjustment factor Kp": Kp=(P_current-P_target) / P_range; where P_current is the current power generation, P_target is the commanded target power, and P_range is the power adjustment range under the current operating conditions. The value range of Kp is limited to [-1, 1]. Control actions: When Kp > 0 (power reduction is required), the unit proportionally closes the virtual passage to the low-pressure cylinder (physically achieved by adjusting the hydraulic-electric dual-control butterfly valve), and simultaneously proportionally opens the virtual passage to the heating network. The larger the Kp value, the greater the adjustment amplitude and rate according to a predetermined rule. When Kp < 0 (power increase is required), the opposite operation is performed. When the power approaches the target value, Kp tends to 0, and the unit enters a holding state. The path control unit is associated with the "thermal stress index" provided by the state evaluation unit of the parameter coordination module. When the thermal stress index is high, the system will automatically reduce the upper limit of the rate of change of Kp to achieve "flexible" adjustment and protect the turbine body. When the path control unit rapidly transfers steam energy, it causes a sharp change in the energy supply and demand on the boiler side. At the moment of initiation, the path control unit sends a feedforward signal indicating an impending change in energy flow to the air distribution optimization unit of the combustion steady-state module. Based on this signal, the air distribution optimization unit can pre-adjust the combustion state to prepare for subsequent changes in boiler energy balance, rather than reacting passively, thus significantly reducing fluctuations in furnace pressure. The path control unit receives the "thermal stress index" and "vibration risk index" from the state assessment unit of the parameter coordination module and limits its adjustment rate accordingly, which is crucial for ensuring equipment safety. Simultaneously, the steam flow distribution data generated by the path control unit is sent to the parameter coordination module in real time. The priority decision unit of the coordination module uses this data to predict the changing trends of key parameters of the turbine and boiler in the next tens of seconds, so as to switch control priorities earlier and more intelligently (for example, to intervene in advance if the reheat steam temperature is predicted to drop rapidly). After each peak shaving task, the adjustment process curve of the path control unit (Kp changes over time) and the heat storage state change of the heat network coupling unit are recorded by the case library storage unit of the state backtracking and self-gain module. When a similar operating condition is encountered again, the case matching and calling unit can provide this historical data to the instruction parsing and pre-control module to generate a better initial strategy, thereby realizing the continuous evolution of the entire system in the energy redistribution stage.The heat network coupling unit incorporates a "heat network heat storage state estimation" algorithm. By monitoring parameters such as the supply and return water temperatures, flow rates, and water levels in the storage tanks, it calculates in real-time the additional heat load capacity the heat network can currently accommodate. When the path control unit needs to transfer a large amount of steam to the heating side for rapid load reduction, the heat network coupling unit assesses its own capacity in real time: if the capacity is sufficient, it fully accepts the load and feeds this signal back to the path control unit, allowing it to make rapid adjustments; if the capacity is close to saturation, the heat network coupling unit sends a "heat network buffer capacity limited" warning signal to the strategy generation unit of the command parsing and pre-control module. Upon receiving this signal, the strategy generation unit... Subsequently, the system dynamically adjusts its issued "equipment action sequence list," for example, appropriately reducing the initial load reduction rate target or triggering the intervention of the combustion steady-state module in advance, thereby avoiding regulation failure or equipment risks caused by heat network constraints from the source. The heat network coupling unit has a "thermal inertia compensation" function: in the initial stage of regulation, it temporarily raises the heating parameters (such as temperature) to slightly higher than the rated value within the allowable range. This aims to actively utilize the huge thermal inertia of the heat network pipeline to store more instantaneous heat energy, providing the unit with a larger "power regulation pool" that can last for tens of seconds to several minutes. This significantly enhances the system's ability to respond to rapid AGC commands.

[0021] Furthermore, as shown in Figures 1 and 4, the combustion steady-state module includes: a spectral sensing unit, which receives signals from optical sensors arranged at different heights in the furnace and focuses on analyzing the emission spectral intensity and fluctuation characteristics of specific chemical free radicals (such as OH and CH) in the ultraviolet to visible light band; its calculation is a comprehensive index called "combustion stability index R": R = α*(I_OH / I_avg) + β*(1-σ_CH / μ_CH), where I_OH is the instantaneous spectral intensity of OH free radicals, I_avg is its moving average, σ_CH and μ_CH are the standard deviation and mean of the spectral intensity of CH free radicals, respectively, and α and β are... Weighting coefficients; a higher R value (closer to 1) indicates more intense and stable combustion; a lower R value (closer to 0) indicates combustion tends towards instability; the spectral sensing unit can not only assess the overall combustion state, but also perform "local instability positioning" by comparing the R values ​​of different burner corner regions; when the R value of a certain corner is significantly lower than that of other corners (e.g., the difference exceeds 0.15), the area is marked as a "weakly stable combustion zone," and this positioning information is output along with the overall R value; the spectral sensing unit sends the R value to the status assessment unit of the parameter coordination module in real time; the status assessment unit compares the R value with other key boiler parameters (such as main steam pressure, furnace negative pressure)... The system performs integrated calculations to generate a more comprehensive "overall boiler stability" index. The priority decision-making unit then uses this comprehensive index to determine the resource allocation of the control system. For example, when the "overall boiler stability" is low, the priority decision-making unit will order the coordination control system to temporarily relax the requirements for the accuracy of the main steam pressure, allowing it to fluctuate within a certain range. This creates a more relaxed control environment for the stable combustion of the air distribution optimization unit, avoiding the vicious cycle of "loss of combustion" due to "pressure maintenance". During each deep peak shaving process, the R-value change curve recorded by the spectral sensing unit and the damper operation sequence executed by the air distribution optimization unit will be completely recorded by the case library storage unit. These data are directly related to specific operations and combustion state responses; by analyzing historical cases, the system can learn the optimal air volume allocation mode for specific coal quality and specific load stages; when the next case matching and calling unit matches a similar working condition, it can directly provide the instruction parsing and pre-control module and the air distribution optimization unit of this module with a set of practically verified, near-optimal initial damper opening offset values, significantly improving combustion stability in the early stage of peak shaving; the air distribution optimization unit is used for "differential air volume compensation"; the air distribution optimization unit receives the global R value and "weak stable combustion zone" positioning information sent by the spectral sensing unit: when the global R value is lower than the safety threshold (e.g., 0).6) When the unit is in a "weakly stable combustion zone", it will activate a global combustion stabilization strategy. For example, it will moderately increase the differential pressure setting between the secondary air box and the furnace to enhance the rigidity of the entire tangential flame. When a "weakly stable combustion zone" exists, the unit will activate a local intervention strategy: finely adjust the opening of the secondary air damper of the adjacent burner located upstream of the weakly stable combustion zone. For example, if corner D is a weakly stable combustion zone, the auxiliary air damper of corner C will be opened appropriately to utilize the enhanced jet kinetic energy of corner C to entrain the high-temperature flue gas from corner D downstream to the root of the burner in corner D, thereby achieving stable combustion in corner D. This cooperative flow field intervention is more effective than simply adjusting the air volume of corner D itself. During the load stabilization phase, the air distribution optimization unit will use the dual objectives of maximizing the R value and minimizing NOx emissions to conduct small, slow oscillations near the current total air volume, automatically finding and locking the optimal air-coal ratio. The setpoint ensures the boiler always operates in a highly efficient and clean condition. The real-time R-value calculated by the spectral sensing unit is continuously sent to the strategy generation unit. When formulating strategies involving significant load reduction, the strategy generation unit refers to the current R-value: if the R-value is already low, a more conservative load reduction rate will be chosen, or backup measures such as oil injection will be arranged in advance to mitigate risks at the source. When the path control unit begins to rapidly transfer energy, it sends a feedforward signal to the air distribution optimization unit. Upon receiving this signal, the air distribution optimization unit does not wait for changes in furnace parameters but acts immediately based on preset rules; for example, at the moment of load reduction, the secondary air differential pressure setpoint is increased by 3-5% in advance to proactively counteract the potential decrease in furnace temperature and combustion fluctuations caused by reduced steam heat absorption.

[0022] Furthermore, as shown in Figures 1 and 5, the parameter coordination module includes: a state assessment unit, which continuously receives key signals from the entire system, including but not limited to: the combustion stability index R from the spectral sensing unit of the combustion steady-state module, the main steam pressure and its setpoint deviation, the main steam temperature deviation, the furnace negative pressure, and the steam path adjustment factor Kp provided by the path control unit of the energy redistribution module; and calculates a core index—"system stability margin S"—through a weighted algorithm: S=w1*f(R)+w2*(1-|ΔP| / P_set)+w3*(1-|ΔT| / T_set)-w4*|Kp| where f(R) is a function that maps the combustion stability index to a 0-1 value. |ΔP| / P_set and |ΔT| / T_set are the relative deviations of pressure and temperature, and w1~w4 are weighting coefficients. A higher S value (closer to 1) indicates a more stable system with stronger resistance to disturbances; a lower S value indicates a more fragile system. By comparing the deterioration of the above sub-indicators, the 1-2 parameters that currently have the greatest drag on the S value are identified (e.g., currently mainly due to excessive main steam pressure deviation or excessively low fuel stability index R), and this identifier is output to provide precise direction for subsequent priority decisions. The state assessment unit receives the Kp value from the path control unit and uses it as an important negative factor in assessing the system's stability margin S. Simultaneously, the priority decision unit generates control... Strategies (such as the "pressure protection" mode) directly affect the execution effect of the path control unit. More importantly, the priority decision unit sends a "maximum allowable adjustment rate" command to the path control unit. This rate dynamically changes based on the real-time calculated S value. When the S value is low, this rate limit automatically tightens, forcing a smoother energy redistribution process and protecting equipment safety. The status assessment unit uses the combustion stability index R provided by the spectral sensing unit as one of the core inputs for calculating the S value. When a decrease in the R value leads to a decrease in the S value, and the "dominant contradiction indicator" points to a combustion problem, the priority decision unit immediately switches to a control mode that is conducive to combustion stability. It sends a command to the boiler's main controller to temporarily relax the control mode. The precise control requirements for main steam pressure create a more relaxed environment for the air distribution optimization unit to operate more freely and achieve stable combustion; it is an important guarantee for achieving stable combustion without oil injection under extreme peak shaving; during each peak shaving process, the S-value change curve of the state assessment unit and the strategy switching sequence of the priority decision unit are recorded in detail by the case library storage unit; by analyzing historical data, the weight coefficients w1~w4 and the trigger thresholds of various strategy templates can be optimized; when the next case matching and calling unit matches a similar operating condition, it can not only recommend the equipment action sequence, but also directly recommend a set of verified and optimal initial control strategy parameters for the parameter coordination module, so that the system runs in the best state from the beginning;The priority decision-making unit internally stores several sets of "control strategy templates," such as the "pressure maintenance-stable combustion" template, the "temperature maintenance-environmental protection" template, and the "rapid response" template. Its decision-making logic is based on two inputs: first, the "system stability margin S" and "dominant contradiction identifier" provided by the state assessment unit; second, the phased control intent issued by the strategy generation unit in the instruction parsing and pre-control module. When the S value is higher than the safety threshold (e.g., 0.7), the system adopts a "balanced control" mode, finely adjusting all parameters; when the S value decreases (e.g., between 0.4 and 0.7), or a rapid load change instruction is received from upstream... When the dominant contradiction is identified, the unit immediately switches to the strategy template corresponding to the "dominant contradiction identifier." For example, if the dominant contradiction is "large pressure deviation and rapid load reduction," the "maintain pressure - stabilize combustion" template is activated. This template issues the following instructions: temporarily relax the control dead zone for main steam temperature (e.g., from ±5℃ to ±8℃) and allocate more control computing resources to feedwater control and combustion control to prioritize stabilizing pressure and combustion. When the S value is below the danger threshold (e.g., 0.4), the priority decision unit triggers the "safety priority" mode, taking stronger measures such as significantly slowing down the rate of load change or forcibly intervening to fix certain regulating mechanisms.

[0023] Furthermore, as shown in Figures 1, 2, and 6, the state backtracking and self-gain module includes: a case library storage unit, which defines a standard data structure called "peak shaving case"; when a peak shaving task (such as reducing load from 50% to 30%) is completed, this unit will initiate the data acquisition and encapsulation process; a complete "peak shaving case" includes: operating condition labels: initial load, target load, coal quality information, whether heating is provided, etc.; strategy input: a complete "equipment action sequence list" from the instruction parsing and pre-control module; process data: key time series data from all modules of the system, including the path adjustment factor Kp of the energy redistribution module, the combustion steady-state index R of the combustion steady-state module, the system stability margin S of the parameter coordination module and its dominant contradiction identifier, etc.; performance indicators: key performance indicators calculated after the task is completed, such as total completion time, load instruction tracking error integral, maximum deviation of main steam pressure, whether oil is injected, etc.; before storage, the case library storage unit will automatically verify the integrity of the data. Regarding rationality, for cases of data loss due to sensor anomalies or communication interruptions, they will be marked and their subsequent priority will be reduced to ensure the cleanliness and reliability of the case library. The case library storage unit records the evolution history of the system stability margin S and the dominant contradiction identifier, providing a panoramic view of the control difficulty and bottlenecks of each peak shaving operation. By analyzing this data, the strategy optimization unit can optimize the weight coefficients (w1~w4) of the state evaluation unit in the parameter coordination module, or suggest adjusting the strategy switching threshold of the priority decision unit, making the judgment of parameter coordination more accurate and the decision more timely. The case matching and invocation unit is triggered when the instruction parsing and pre-control module starts processing a new peak shaving task. The case matching and invocation unit receives the initial conditions (operating condition labels) of the new task and searches for similar historical cases in the case library. Cases are used to calculate a case similarity score: Similarity score = W1 * Load similarity + W2 * Coal quality similarity + W3 * Heating mode similarity. Load similarity primarily considers the proximity of the initial and target loads. Upon successful matching, this unit returns a complete data package of one or more matching cases. The case matching and invocation unit analyzes the "equipment action sequence list" and key operation nodes (such as the timing of pulverizer shutdown, peak values ​​of path adjustment factors, etc.) in the best-matched case, extracting a set of "recommended strategy parameters," such as "it is recommended to shut down pulverizer C when the load drops to 38%," and provides this to the upstream module in a more concise form to improve information utilization efficiency. The strategy optimization unit runs automatically on a regular basis (e.g., weekly) or after accumulating a sufficient number of new cases. It employs a "comparative analysis method" to focus on analyzing multiple cases targeting the same or similar operating conditions. The strategy optimization unit seeks two patterns: commonalities of successful patterns: it selects all cases with excellent performance indicators and analyzes the commonalities in their strategies, such as whether they all use a specific air-coal ratio curve or perform a certain operation at a specific load point.Root cause of failure mode: By comparing the operational differences between successful and failed cases (such as combustion flashing or severe parameter overruns), the key operational instructions or parameter settings leading to the failure are identified. The strategy optimization unit, after analyzing and confirming that a certain adjustment can improve performance, generates a "strategy fine-tuning suggestion" and interacts directly with the strategy generation unit in the instruction parsing and pre-control module. For example, analysis reveals that when the volatile matter content of coal is below 20%, advancing the "activation of heat storage buffer mode" time point in the strategy template from T+0.2 seconds to T+0 seconds significantly improves the initial stable combustion effect. The strategy optimization unit modifies the corresponding template parameters within the strategy generation unit accordingly. This allows the system optimization to be completed automatically without manual intervention. By analyzing these historical curves, the strategy optimization unit can optimize the best Kp change pattern for different load reduction rates and correct the strategy accordingly, thereby guiding the future energy redistribution process.

[0024] Example 2 further optimizes the supercritical unit low-load deep peak shaving system provided in Example 1. Specifically, as shown in Figures 1, 2, 3, 4, 5, and 6, a supercritical unit low-load deep peak shaving method includes the following stages: First stage, instruction deep parsing and strategy pre-generation: A. Feature extraction and intent recognition: After capturing the AGC instruction, a 10-second sliding time window is initiated to calculate the first derivative (load change rate, %Pe / min) and second derivative (acceleration of change) of the instruction in real time; this is a deep interpretation of grid dispatch; for example, a rapidly decreasing instruction (< A load reduction of -2.5% Pe / min is identified as an "emergency load reduction mode," indicating a potential power surge or deficit in the grid, requiring the generating units to "respond quickly at all costs." A slow load reduction command, on the other hand, is identified as a "gradual load reduction mode," indicating routine grid adjustments, requiring the generating units to "prioritize a smooth transition." This pattern recognition provides the fundamental basis for subsequently selecting drastically different control strategies. B. Variable parameter filling in the strategy blueprint: Based on the identified pattern, the corresponding "strategy template" is invoked—a pre-designed framework containing multiple blank action sequences. The white parameter is "real-time personalized filling"; for example, in the template of "emergency load reduction to 30%", there is an instruction "T+0 seconds: activate steam path flexible guidance, target power generation decreases by X%"; this X is not a fixed value (such as 60%), but is dynamically obtained through a lightweight calculation model: X = (current load - target load) * response coefficient K; where, the response coefficient K will be fine-tuned according to the urgency of this instruction (rate of change) and the current heat storage level of the boiler; if the boiler heat storage is sufficient, the K value can be slightly greater than 1, so that the initial energy transfer is more intense and the boiler's energy is fully utilized. Rapid response; if heat storage is insufficient, the K value is set to 1 or slightly less than 1 to avoid overdraft leading to subsequent instability; this ensures that each generated "Equipment Action Sequence List" is customized for the current unit status; C. In the list, not only is the energy redistribution module specified to act at T+0 seconds, but the combustion steady-state module is also commanded in advance to enter the "heat storage buffer" state at T+0.2 seconds; this 0.2-second advance is a "feedforward" coordination based on the physical law that "energy transfer inevitably leads to changes in boiler operating conditions"; allowing the boiler side to act like a receiver who has received a warning, preparing in advance to meet the upcoming energy impact.

[0025] The second stage involves proactive energy transfer and heat network coordination: A) Flexible guidance and self-protection of the steam path: The path control unit continuously adjusts the path adjustment factor Kp to achieve stepless and linear conversion between power generation and heating; its innovation lies in "flexibility": the rate of change of Kp is linked in real time with the "thermal stress index" provided by the parameter coordination module; B) Proactive empowerment and boundary feedback of the heat network system: In the initial stage of adjustment, the heat network coupling unit proactively raises the heating temperature temporarily by 1-2°C within the safe upper limit; the heat network coupling unit calculates the remaining buffer capacity of the heat network in real time, and once it is predicted that the buffer capacity is about to be saturated, the heat network coupling unit will immediately send a high-level warning signal to the command parsing and pre-control module; this forces a reassessment and adjustment of the strategy before the situation deteriorates, such as slowing down the load reduction rate in advance; this achieves a fundamental shift from "remediation after problems occur" to "anticipating and avoiding problems".

[0026] The third stage involves spectral sensing of combustion status and active combustion stabilization: A. Quantitative and spatial location diagnosis of combustion stability: The combustion stability index R provided by the spectral sensing unit enables a "numerical CT scan" of the combustion status; it not only provides a conclusion of good combustion but also accurately locates the "current risk of flameout in the D-angle burner area" (i.e., the "weak combustion stability zone"); this diagnostic capability from the overall to the local is a prerequisite for implementing subsequent precise interventions; B. Collaborative combustion stabilization intervention based on flow field optimization: After receiving the location of the "weak combustion stability zone," the air distribution optimization unit fine-tunes the auxiliary damper of its upstream C-angle burner; this aims to enhance the kinetic energy of the C-angle jet and utilize the aerodynamics of tangential combustion. The force characteristics forcefully entrain and guide the high-temperature flue gas to the root of corner D, fundamentally improving the environment of the weak stable combustion zone by optimizing the aerodynamic field of the entire furnace. The effect is more lasting and stable than simply adjusting itself, while also avoiding the temperature drop and NOx increase caused by blindly increasing the air volume; C. Closed-loop coordination with the previous stage: The stable combustion index R generated in this stage is sent to the parameter coordination module in real time to calculate the system stability margin S; If the S value continues to decrease due to combustion deterioration, the priority decision unit will make a decisive decision: temporarily relax the control precision of the main steam pressure; This measure ensures that the unit can prioritize survival in extreme operating conditions.

[0027] Fourth stage, system stability margin assessment and dynamic priority decision-making: A. The state assessment unit continuously operates, performs multi-source information fusion, and combines the four key pieces of information, namely the stable combustion index R (combustion state) from the combustion steady-state module, the main steam pressure / temperature deviation (steam-side state), and the path adjustment factor Kp (energy transfer intensity) from the energy redistribution module, into a single, comprehensive quantitative index - the system stability margin S through a weighted algorithm. The calculation formula for the S value is S = w1*f(R)+... - w4*|Kp|. Taking the intensity of energy transfer (|Kp|) as a negative factor intuitively reflects the physical essence that "rapid energy transfer will impact system stability". At the same time, by comparing the deterioration degrees of each item, it can output a "dominant contradiction identifier", such as clearly indicating whether the primary problem of the current system is "pressure out of control" or "combustion on the verge of instability", providing a basis for subsequent precise decision-making. B. Dynamic resource allocation and target trade-off based on the margin: When the system stability margin S is relatively high (e.g., S > 0.7), the system is in the "comfort zone" and performs refined control on all parameters. When the S value decreases (e.g., 0.4 < S < 0.7), or at the initial stage of rapid load change, the priority decision-making unit will switch to the corresponding "survival mode" according to the "dominant contradiction identifier". For example, if the dominant contradiction is "rapid pressure drop and unstable combustion", immediately activate the "pressure-maintaining - combustion-stabilizing" template. This template will execute: actively and temporarily relax the control dead zone of the main steam temperature (e.g., from ±5°C to ±8°C). This means that the system allows the temperature to fluctuate within a larger range, thus liberating the computing power of the control system and the action frequency of the actuator from temperature control and allocating more to feedwater control and combustion control to go all out to stabilize the pressure and the flame, ensuring core safety. C. Mandatory intervention in the previous stage: In this stage, an instruction of "maximum allowable adjustment rate" is sent to the path control unit. When the S value is low, this rate limit will be tightened sharply, forcing the energy transfer process to change from fast to slow, reducing the impact on the system from the source.

[0028] Phase 5: Full-Process Case Retrospective and Strategy Self-Enhancement: A. Holographic Archive Encapsulation of the Peak Shaving Process: After each peak shaving task is completed, the case library storage unit automatically starts, encapsulating the entire digital footprint of this task—including the initial "Equipment Action Sequence List" (strategy), the key parameter curves such as Kp, R, and S throughout the process (execution process), and the final performance indicators (results)—into a structured "peak shaving case." This case contains complete content of the antecedent (operating conditions), decision (strategy), process (data), and consequences (performance), providing a unique and reliable data foundation for subsequent in-depth analysis; B. Experience Preloading Based on Similarity Matching: When a new peak shaving task is started, the case matching and invocation unit will work first. It calculates similarity scores to find similar cases from a historical case database; once found, it intelligently refines the data and provides the strategy generation unit with a set of highly condensed "recommended strategy parameters," such as: "Based on historical experience, under your current coal quality and target load, it is recommended to delay the shutdown of the B coal mill from 38% load to 35%," thus improving the quality and success rate of the initial strategy; C. Cross-case in-depth analysis and self-iteration of strategy templates: The strategy optimization unit specifically seeks commonalities among successful cases and the root causes of failures. For example, through analysis, it may be found that when the volatile matter content of the coal is below 20%, all successful cases activate the heat storage buffer at the moment the load reduction begins (T+0 seconds), while most failure cases follow the default T+0.2 seconds.

[0029] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection, an electrical connection, or a connection that allows communication between them; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components, unless otherwise explicitly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0030] Obviously, the embodiments described above are merely some embodiments of the present invention, not all embodiments. The accompanying drawings show preferred embodiments of the present invention, but do not limit the patent scope of the present invention. The present invention can be implemented in many different forms; rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the patent protection scope of this invention.

Claims

1. A deep peak-shaving system for a supercritical unit at low load, characterized in that, include: The instruction parsing and pre-control module is used to parse load instructions into a list of equipment action sequences; the energy redistribution module is used to execute the instructions in the list of equipment action sequences and to achieve rapid adjustment of power generation by adjusting the distribution of steam energy flow; the combustion steady-state module is used to maintain furnace combustion stability under low load through optical sensing and active air distribution. The parameter coordination module is used to evaluate the overall stability margin of the system and dynamically adjust the control priority of different parameters; the state backtracking and self-gain module is used to store peak shaving cases, match historical strategies, and optimize control parameters; wherein, the instruction parsing and pre-control module is communicatively connected to the energy redistribution module, the combustion steady-state module, the parameter coordination module, and the state backtracking and self-gain module to collaboratively realize the deep peak shaving process.

2. The deep peak-shaving system for a supercritical unit at low load according to claim 1, characterized in that, The instruction parsing and pre-control module includes: an instruction feature recognition unit, used to classify instructions into different peak shaving modes by calculating the rate of change of load instructions and comparing it with a preset threshold; a strategy generation unit, used to call the corresponding strategy template according to the peak shaving mode, and generate a list of equipment action sequences containing specific action timings and parameters based on the current load and the target load; and an efficiency evaluation unit, used to compare the actual operating data with the expected target of the equipment action sequence list after the peak shaving task is completed, and generate fine-tuning suggestions for optimizing the strategy template.

3. A deep peak-shaving system for a supercritical unit at low load according to claim 2, characterized in that, The energy redistribution module includes: a path control unit, used to calculate and output a path adjustment factor Kp based on the target power generation value, so as to continuously adjust the steam distribution ratio to the low-pressure cylinder and the heating network, where Kp=(P_current-P_target) / P_range, and the value range of Kp is [-1,1]; a heating network coupling unit, used to calculate the additional heat load capacity that the heating network can accept in real time, and coordinate with the path control unit when it operates, and send an early warning signal to the command parsing and pre-control module when the capacity is close to saturation; wherein, the adjustment rate of the path control unit is limited by the thermal stress index output by the parameter coordination module.

4. A deep peak-shaving system for a supercritical unit at low load according to claim 3, characterized in that, The combustion steady-state module includes: a spectral sensing unit, used to calculate a combustion stability index R by analyzing the spectral intensity and fluctuation characteristics of specific chemical free radicals in the furnace, and to perform local instability assessment, where R = α*(I_OH / I_avg) + β*(1-σ_CH / μ_CH); and an air distribution optimization unit, used to receive the combustion stability index R and local instability assessment information, and to activate a global combustion stability strategy when the R value is lower than a safety threshold, and to activate a local intervention strategy targeting upstream adjacent burners when a weak combustion stability zone exists; wherein, the air distribution optimization unit receives a feedforward signal from the energy redistribution module and acts in advance before energy flow changes.

5. A deep peak-shaving system for a supercritical unit at low load according to claim 4, characterized in that, The parameter coordination module includes: a state assessment unit, used to receive and integrate the combustion stability index R from the combustion steady-state module, the path adjustment factor Kp from the energy redistribution module, and the main steam pressure and temperature deviation, to calculate a system stability margin S, where S=w1*f(R)+w2*(1-|ΔP| / P_set)+w3*(1-|ΔT| / T_set)-w4*|Kp|, and to identify the dominant conflicting parameters; and a priority decision unit, used to dynamically switch between different control strategy templates based on the system stability margin S and the dominant conflicting parameters, so as to adjust the control dead zone and computational resource allocation of different parameters.

6. A deep peak-shaving system for a supercritical unit at low load according to claim 5, characterized in that, The state backtracking and self-gain module includes: a case library storage unit, used to store peak shaving cases in a standard data structure, wherein the cases include operating condition labels, strategy inputs, process data, and performance indicators; a case matching and recall unit, used to calculate the case similarity score based on the operating condition labels when a new peak shaving task is started, and to match and return recommended strategy parameters of similar cases from the case library; and a strategy optimization unit, used to periodically analyze successful and failed cases in the case library, locate the key operations that cause performance differences, and generate direct correction instructions for the strategy templates in the instruction parsing and pre-control module.

7. A deep peak-shaving system for a supercritical unit at low load according to claim 6, characterized in that, The device action sequence list generated by the instruction parsing and pre-control module precisely specifies the specific unit in the downstream module that receives the instruction in its timeline; and when the path control unit of the energy redistribution module starts to operate, it sends an energy flow change feedforward signal to the air distribution optimization unit of the combustion steady state module; at the same time, the priority decision unit of the parameter coordination module sends the maximum allowable adjustment rate instruction based on the real-time system stability margin S dynamically calculated to the path control unit of the energy redistribution module.

8. A method for deep peak shaving at low load in a supercritical unit, using the deep peak shaving system for a supercritical unit as described in claim 7, characterized in that, Includes the following steps: Instruction deep analysis and strategy pre-generation steps: Analyze AGC instruction characteristics and generate a list of device action timings; Energy flow forward transfer steps: Execute the aforementioned list to distribute power generation and heating energy by adjusting the steam path; Combustion state sensing and stabilization steps: Assess and actively maintain combustion stability based on spectral signals; System stability margin decision-making steps: Integrate multi-source parameters to evaluate system stability and dynamically adjust control priorities; Case backtesting and self-gain steps: Store peak-shaving process data and use historical cases to optimize subsequent strategies.

9. A method for deep peak shaving at low load in a supercritical unit according to claim 8, characterized in that, The instruction deep analysis and strategy pre-generation steps include: classifying AGC instructions into different instruction modes according to their load change rate; calling the corresponding strategy template based on the instruction mode, and filling the template with calculated dynamic parameters to generate the equipment action sequence list; wherein, the dynamic parameters include the target power generation reduction ratio X, which is calculated using the formula X = (current load - target load) * response coefficient K, and the response coefficient K is dynamically adjusted according to the load change rate and the current boiler heat storage level.

10. A method for deep peak shaving at low load in a supercritical unit according to claim 8, characterized in that, The combustion state sensing and stabilization steps and the system stability margin decision-making steps include: calculating the stabilization index R by analyzing the spectral signal of chemical free radicals in the furnace to achieve quantitative assessment of combustion stability and location of the weak stabilization zone; calculating the system stability margin S based on the stabilization index R, main steam parameter deviation and energy transfer intensity; and dynamically switching the control strategy when the system stability margin S is lower than a set threshold, including temporarily relaxing the control precision of the main steam temperature to prioritize the stability of boiler pressure and combustion.

Images