A method for on-line monitoring and regulation of equal smoke speed and low abrasion of a biomass boiler

Abstract

The present application relates to the technical field of boiler monitoring and regulation, and particularly relates to an equal-smoke-velocity low-abrasion online monitoring and regulation method for a biomass boiler. The method comprises: obtaining a smoke velocity spatial dispersion diffusion index and a particle flow bias index; based on the smoke velocity spatial dispersion diffusion index and the particle flow bias index, screening an optimal induced draft fan control sequence from all to-be-analyzed control sequences; and based on the optimal induced draft fan control sequence, controlling the induced draft fan of the biomass boiler at a current monitoring time. The present application can make the boiler operation tend to a state of most uniform smoke flow field and most dispersed particle scouring by directly embedding the smoke velocity spatial dispersion diffusion index and the particle flow bias index into the regulation mechanism of the model prediction control framework, so as to inhibit the occurrence and development of local abrasion, and further improve the stability and reliability of long-term operation of the biomass boiler.

Description

Technical Field

[0001] This invention relates to the field of boiler monitoring and control technology, specifically to a method for online monitoring and control of low wear at constant flue gas velocity in biomass boilers. Background Technology

[0002] In the operation of biomass boilers, in order to ensure the stability of boiler combustion, flue gas delivery, and the operational safety of the tail heating surface, it is necessary to achieve constant flue gas velocity and low wear operation of biomass boilers. Currently, Model Predictive Control (MPC) is commonly used to adjust the operating parameters of induced draft fans online in order to achieve constant flue gas velocity and low wear operation of biomass boilers, or to monitor and control the constant flue gas velocity and low wear operation of biomass boilers.

[0003] In current technical practice, the implementation process of MPC is specifically manifested as follows: real-time online monitoring of key operating parameters such as furnace negative pressure and flue gas velocity, and based on these monitoring data, a state vector representing the boiler operating state is constructed; in each control cycle, a set of candidate induced draft fan control sequences is generated based on the current state vector representing the boiler operating state; the boiler operating state within the future prediction time window is recursively predicted using a state space prediction model, and a cost function reflecting the overall deviation between the predicted state and the expected state is constructed; by solving a quadratic programming problem, the optimal induced draft fan control sequence that minimizes the value of the cost function is found; and based on the obtained optimal induced draft fan control sequence, online monitoring and control of the biomass boiler with constant flue gas velocity and low wear is achieved.

[0004] However, the flue gas flow inside the boiler exhibits significant time-varying and local disturbance characteristics. In actual operation, the local flow field state is easily affected by various factors such as dynamic changes in particle transport paths, transient disturbances in the flue gas, and non-uniform changes in local swirl, leading to strong spatial and temporal uncertainties in the local flow state. Furthermore, the traditional MPC (Multi-Process Control) method, which uses quadratic programming to optimize the minimum cost function and find the optimal induced draft fan control sequence, essentially controls the induced draft system based on minimizing the overall operating state deviation within a predicted time window. This approach can lead to problems even when the global cost function reaches its minimum (i.e., the overall operating state deviation is minimized). Even with the minimum deviation in the flow state, the local flow field state in different spatial regions inside the boiler may still be undergoing unfavorable changes. That is, even if the global cost function reaches its minimum value, there may still be sudden changes in local flue gas velocity, frequent flue gas fluctuations, or concentrated particle transport paths, which may lead to a high scouring risk state in local areas. During the actual operation of the boiler, the stability of the flue gas field decreases. Therefore, how to construct an adaptive and collaborative online monitoring and control mechanism to optimize the overall operating performance and local flow field structure in order to improve the stability and reliability of long-term operation of biomass boilers or improve the stability and reliability of low-fly-velocity operation of biomass boilers has become an urgent problem to be solved. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides a method for online monitoring and control of low wear at constant flue gas velocity in biomass boilers. The specific technical solution adopted is as follows:

[0006] One embodiment of the present invention provides a method for online monitoring and control of low wear at constant flue gas velocity in a biomass boiler, comprising the following steps:

[0007] By monitoring the operating parameters of the monitoring points and induced draft fans in the biomass boiler online, the current actual state vector of the biomass boiler at the current monitoring time is obtained. Based on the current actual state vector, MPC, and the state space prediction model of MPC, a set of candidate induced draft fan control sequences at the current monitoring time, the cost function value of each candidate induced draft fan control sequence in the set of candidate induced draft fan control sequences, and the future state vector corresponding to each future monitoring time in the preset future prediction time window are obtained. The actual state vector includes parameters in the dimensions of flue gas velocity and particulate matter concentration.

[0008] Based on the overall degree to which the parameters in the future state vector exceed the preset safe operation constraint range, the constraint violation amount of each candidate induced draft fan control sequence is obtained, and the control sequence to be analyzed is selected based on the constraint violation amount and the cost function value.

[0009] Based on the dispersion of flue gas velocity at future monitoring times and the diffusion degree over time, the flue gas velocity spatial dispersion diffusion index of each control sequence to be analyzed is obtained. Based on the deviation of the particulate flow intensity from the uniform flow state at future monitoring times, the particulate flow bias index of each control sequence to be analyzed is obtained. Based on the flue gas velocity spatial dispersion diffusion index and the particulate flow bias index, the optimal induced draft fan control sequence is selected from all control sequences to be analyzed.

[0010] The induced draft fan of the biomass boiler is controlled based on the optimal induced draft fan control sequence at the current monitoring time.

[0011] Beneficial Effects: This invention first monitors the operating parameters of monitoring points and induced draft fans in a biomass boiler online to obtain the current actual state vector of the biomass boiler at the current monitoring moment. Based on the current actual state vector, MPC (Multi-Process Control) and the state-space prediction model of MPC, it obtains a set of candidate induced draft fan control sequences at the current monitoring moment, the cost function value of each candidate induced draft fan control sequence in the candidate induced draft fan control sequence set, and the future state vector corresponding to each future monitoring moment within a preset future prediction time window. Then, based on the overall degree to which the parameters in the future state vector exceed the preset safe operation constraint range, it obtains the constraint violation amount of each candidate induced draft fan control sequence. The control sequences to be analyzed are selected based on constraint violation and cost function values. Then, based on the dispersion of flue gas velocity at future monitoring times and its diffusion over time, the spatial dispersion diffusion index of each control sequence is obtained. Based on the deviation of particulate flow intensity from a uniform flow state at future monitoring times, the particle flow bias index of each control sequence is obtained. Based on the spatial dispersion diffusion index and the particle flow bias index, the optimal induced draft fan control sequence is selected from all control sequences. Finally, the induced draft fan of the biomass boiler is controlled based on the optimal induced draft fan control sequence at the current monitoring time. Furthermore, this invention, by directly embedding the spatial dispersion diffusion index and the particle flow bias index into the model predictive control framework's regulation mechanism, enables the boiler to operate in a state with the most uniform flue gas field and the most dispersed particle scouring, thereby suppressing the occurrence and development of localized wear and improving the long-term stability and reliability of the biomass boiler. Attached Figure Description

[0012] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart of an online monitoring and control method for low wear at constant flue gas velocity in a biomass boiler, according to the present invention. Detailed Implementation

[0014] 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. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the protection scope of the embodiments of the present invention.

[0015] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art.

[0016] This embodiment provides a method for online monitoring and control of low wear at constant flue gas velocity in a biomass boiler, detailed as follows:

[0017] like Figure 1 As shown, the method for online monitoring and control of low wear at constant flue gas velocity in this biomass boiler includes the following steps:

[0018] Step S001: By monitoring the operating parameters of the monitoring points and induced draft fans in the biomass boiler online, the current actual state vector of the biomass boiler at the current monitoring time is obtained. Based on the current actual state vector, MPC, and the state space prediction model of MPC, the candidate induced draft fan control sequence set at the current monitoring time, the cost function value of each candidate induced draft fan control sequence in the candidate induced draft fan control sequence set, and the future state vector corresponding to each future monitoring time in the preset future prediction time window are obtained.

[0019] Traditional MPC (Multi-Process Control) methods, which use quadratic programming to optimize the cost function and find the optimal induced draft fan control sequence, essentially aim to control the induced draft system by minimizing the overall operating state deviation within the predicted time window. However, the flue gas flow inside the boiler exhibits significant time-varying and local disturbance characteristics. In actual operation, the local flow field is easily affected by various factors such as dynamic changes in particle transport paths, transient disturbances in the flue gas, and non-uniform changes in local swirls. This leads to strong uncertainties in the local flow state both spatially and temporally. Consequently, even if the global cost function reaches its minimum (i.e., the overall operating state deviation is minimized), the local flow field state in different spatial regions within the boiler may still be undergoing unfavorable changes, or there may still be sudden changes in local flue gas velocity, frequent flue gas fluctuations, or concentrated particle transport paths. This can result in high scouring risk in local areas, reducing the stability of the flue gas field during actual boiler operation and affecting the long-term stability of the biomass boiler. To improve the stability and reliability of biomass boilers operating under low-velocity and low-wear conditions, this embodiment will construct a monitoring and control mechanism capable of real-time sensing and quantifying the spatial non-uniformity of the flue gas flow field and particle transport bias characteristics. This involves quantifying the flue gas velocity spatial dispersion-diffusion index and particle flow bias index of the selected control sequence. These indices reflect local flow field characteristics and are directly embedded into the model predictive control framework. This monitoring and control mechanism ensures that while maintaining the overall boiler performance, it also achieves synergistic optimization of the flue gas flow field spatial uniformity and particle transport balance, thereby improving the long-term stability and reliability of the biomass boiler or its operation under low-velocity and low-wear conditions.

[0020] This embodiment first utilizes data monitoring and acquisition equipment (flue gas velocity sensor, pressure sensor, etc.) to monitor the operating parameters of each monitoring point and the induced draft fan in any operating biomass boiler online. This yields multi-dimensional operating parameters for each monitoring point and the induced draft fan at different monitoring times, all recorded as raw operating parameters. In other words, the parameters directly monitored and acquired by the acquisition equipment are the raw operating parameters. Thus, the multi-dimensional raw operating parameters of each monitoring point and the induced draft fan at different monitoring times are obtained. The multi-dimensional raw operating parameters of any monitoring point are the parameters belonging to that monitoring point. The operating parameters of the machine are all the operating parameters of the biomass boiler at the corresponding monitoring time, and the acquisition process of the operating parameters of the biomass boiler is a known technology. In this embodiment, the dimensions of the operating parameters of the monitoring points monitored and collected at the monitoring time include, but are not limited to, flue gas velocity, flue pressure, O2 concentration, particulate matter concentration, etc., and the dimensions of the operating parameters of the induced draft fan monitored and collected at the monitoring time include, but are not limited to, the frequency, current, power and damper opening of the induced draft fan, etc.; then, the raw operating parameters of the biomass boiler collected are preprocessed, and the raw operating parameters after data preprocessing are recorded as the processed parameters; the data preprocessing process includes, but is not limited to, missing value handling, outlier detection and correction, data standardization or data normalization, etc., and the data preprocessing process is a known technology. The location and number of monitoring points in a biomass boiler are usually determined based on technical specifications for continuous flue gas emission monitoring or engineering experience. In this embodiment, the monitoring points can be arranged based on the boiler's structural characteristics and flue gas flow path. For example, in this embodiment, the monitoring points may be located in areas that are unknown or have not been identified, such as the furnace outlet area, flue bend area, inlet and outlet areas of the heating surface, and key flow change locations such as the induced draft fan inlet.

[0021] Then, the vector formed by the multi-dimensional processed parameters of each monitoring point in the biomass boiler and the multi-dimensional processed parameters of the induced draft fan at the same monitoring time is recorded as the actual state vector of the biomass boiler at the corresponding monitoring time. In this embodiment, the actual state vector of the biomass boiler at the current monitoring time is recorded as the current actual state vector of the biomass boiler at the current monitoring time. In this embodiment, the operating parameters of the monitoring points and the induced draft fan are collected synchronously. That is, in this embodiment, the length of the actual state vector of the biomass boiler at different times and the number of parameter types in the actual state vector are consistent. In addition, this embodiment requires that the actual state vector of the biomass boiler at different times should include the operating parameters of different monitoring points in the biomass boiler. Furthermore, it requires that the actual state vector of the biomass boiler at each monitoring time should at least include the parameters of the flue gas velocity dimension and the particulate matter concentration dimension of each monitoring point in the biomass boiler. That is, it requires that the actual state vector of the biomass boiler at each monitoring time should at least include the processed flue gas velocity and the processed particulate matter concentration of each monitoring point in the biomass boiler.

[0022] After obtaining the actual state vector, an MPC (Multi-Process Control) is constructed based on it to regulate or control the biomass boiler. The MPC includes a state-space prediction model, and the construction and training process of the MPC is a well-known technique. Then, based on the current actual state vector, the MPC, and its state-space prediction model, a set of candidate induced draft fan control sequences for the current monitoring time, the cost function values ​​of each candidate induced draft fan control sequence in the set, and the future state vectors corresponding to each future monitoring time within a preset future prediction time window are obtained. Specifically, the MPC generates a set of candidate induced draft fan control sequences based on the current actual state vector and control input constraints using a quadratic programming optimizer. The control input constraints need to be set by the implementer according to the actual situation, such as constraining the induced draft fan frequency adjustment range and the maximum rate of change per unit time. Each candidate induced draft fan control sequence represents a possible induced draft fan regulation strategy. Then, each candidate induced draft fan control sequence is input into the state-space prediction model, and the discrete-time state equations are used to generate the control sequence. The system recursively predicts the state trajectory or state evolution vector for a preset number of future monitoring times. The state evolution vector predicted for each candidate draft fan control sequence at each future monitoring time is recorded as the future state vector corresponding to the candidate draft fan control sequence at the corresponding future monitoring time. The types, order, and number of parameters included in the predicted future state vector must be completely consistent with the current actual state vector input to the MPC. Then, for each candidate draft fan control sequence and its corresponding future state vector, a cost function is calculated, and the cost function value for each candidate draft fan control sequence is obtained. The preset number of future monitoring times constitutes a time domain or time window, which is the preset future prediction time window. In practical applications, the control period and the preset future prediction time window are usually set according to the system's dynamic characteristics, control objectives, and computing resources. However, in this embodiment, both are set to empirical values, such as setting the control period to 5 seconds and the preset future prediction time window to 100 seconds. In this embodiment, the time interval between adjacent monitoring times is consistent with the control period. It should also be noted that the difference between MPC in this embodiment and traditional MPC is that only the decision-making process of the optimal induced draft fan control sequence is changed, without changing other processes. Therefore, the process of obtaining the candidate induced draft fan control sequence set, cost function value and future state vector is a well-known technology.

[0023] Therefore, this embodiment can obtain the current actual state vector of the biomass boiler at the current monitoring time, the candidate induced draft fan control sequence set at the current monitoring time, the cost function value of each candidate induced draft fan control sequence in the candidate induced draft fan control sequence set, and the future state vector corresponding to each future monitoring time in the preset future prediction time window through the above process.

[0024] Step S002: Based on the overall degree to which the parameters in the future state vector exceed the preset safe operation constraint range, obtain the constraint violation amount of each candidate induced draft fan control sequence, and filter out the control sequence to be analyzed based on the constraint violation amount and the cost function value.

[0025] Among the numerous candidate induced draft fan control sequences generated by MPC, some sequences, while achieving lower operating costs in optimization calculations (i.e., smaller overall prediction errors), may have future predicted state trajectories that violate the rigid constraints of boiler safe operation (e.g., flue gas velocity exceeding the upper limit, abnormal fluctuations in furnace negative pressure, and oxygen concentration deviating from the safe range). Therefore, to avoid control sequences causing certain key operating parameters (such as flue gas velocity, furnace negative pressure, and oxygen concentration) to exceed the safe operating boundary, thereby triggering local wear or operational risks, this embodiment will next examine the future state vector of each candidate induced draft fan control sequence time-by-time and parameter-by-parameter, calculate the degree of deviation of its predicted value from the safe operating upper or lower limit, and accumulate the values ​​to quantify the extent to which the control sequence causes the monitored parameters to exceed the safe operating upper or lower limit. The severity of the safe operating range, that is, the overall degree to which the parameters in the future state vectors corresponding to each candidate induced draft fan control sequence exceed the preset safe operating constraint range, is used to obtain the constraint violation amount of each candidate induced draft fan control sequence. Based on the quantified constraint violation amount, the candidate induced draft fan control sequences are screened, and only control strategies that theoretically will not cause safety problems enter the next stage of evaluation and decision-making. The specific process of obtaining the constraint violation amount of each candidate induced draft fan control sequence based on the overall degree to which the parameters in the future state vectors corresponding to each candidate induced draft fan control sequence exceed the preset safe operating constraint range is as follows: Taking the i-th candidate induced draft fan control sequence in the set of candidate induced draft fan control sequences as an example, the specific process of obtaining the constraint violation amount of the i-th candidate induced draft fan control sequence is as follows:

[0026] Calculate the constraint violation degree of each parameter in the future state vector corresponding to the i-th candidate induced draft fan control sequence at t future monitoring times. The constraint violation degree of the j-th parameter in the future state vector corresponding to the i-th candidate induced draft fan control sequence at t future monitoring times is: The sum of the constraint violation degrees of all parameters in all future state vectors corresponding to the i-th candidate induced draft fan control sequence is denoted as the constraint violation amount of the i-th candidate induced draft fan control sequence; the expression for the constraint violation amount of the i-th candidate induced draft fan control sequence is:

[0027]

[0028] in, Let be the constraint violation amount of the i-th candidate induced draft fan control sequence, N be the total number of future monitoring moments in the preset future prediction time window, R be the total number of parameters in the state vector, and max() be the maximum value function. Let j be the value of the j-th parameter in the future state vector corresponding to the t future monitoring times within the preset future prediction time window for the i-th candidate induced draft fan control sequence. This represents the maximum value within the preset safe operating constraint range for the parameter type to which the j-th parameter belongs. This is the minimum value of the preset safe operation constraint range for the parameter type to which the j-th parameter belongs. In specific applications, implementers need to set the preset safe operation constraint range for each type of operating parameter based on boiler design parameters, process operating parameters, induced draft fan parameter limitations, etc. The preset safe operation constraint range of the operating parameter refers to the engineering operating boundary that the boiler is allowed to operate safely.

[0029] This indicates the overall deviation of the i-th candidate induced draft fan control sequence from the boiler operating constraints within the future prediction time window, or the degree to which the predicted boiler operating state within the future prediction time window violates the preset safe operating constraints. The larger the value, the more severe the extent to which the monitored operating parameters exceed the safe operating range caused by the i-th candidate induced draft fan control sequence, and the higher the operating risk corresponding to the i-th candidate induced draft fan control sequence. A value of 0 indicates that the future state vector corresponding to the i-th candidate induced draft fan control sequence is always within the safe operation constraints of each operating parameter, and meets the basic operational feasibility requirements under the current operating conditions. Characterization The degree to which the preset safe operation constraint range is exceeded is 0 when the limit is not exceeded; Characterization The degree to which the value is below the preset lower limit of the safe operation constraint range is 0 when it is not below the lower limit; the higher the value is above the preset upper limit of the safe operation constraint range or below the preset lower limit of the safe operation constraint range, the more significant the impact. The larger the value, the more severe the extent to which the i-th candidate induced draft fan control sequence causes the monitored operating parameters to exceed the safe operating range.

[0030] Since the constraint violation amount of the candidate induced draft fan control sequence is equal to 0, it means that the future state vector corresponding to the candidate induced draft fan control sequence is always within the safe operation constraints of each operating parameter and meets the basic operational feasibility requirements under the current operating conditions. Therefore, in this embodiment, all candidate induced draft fan control sequences with constraint violation amounts equal to 0 are obtained next. In the cost function value set composed of the cost function values ​​of all candidate induced draft fan control sequences with constraint violation amounts equal to 0, the largest cost function value is selected as the basic operating cost value. Since under the current boiler operating conditions, as long as the cost function value does not exceed the basic operating cost value, it indicates that this control scheme at least meets the basic operating requirements. Therefore, all candidate induced draft fan control sequences with cost function values ​​less than the basic operating cost value are obtained from the candidate induced draft fan control sequence set and are all recorded as control sequences to be analyzed. That is, the cost function values ​​of the control sequences to be analyzed are all less than the basic operating cost value.

[0031] Therefore, this embodiment can obtain the control sequence to be analyzed through the above process. Subsequently, the control sequence to be analyzed will be further analyzed to obtain the control sequence that minimizes the spatial dispersion and diffusion degree of boiler flue gas flow field and the degree of particle transport bias, which is the optimal induced draft fan control sequence.

[0032] Step S003: Based on the dispersion of flue gas velocity at future monitoring times and the diffusion degree over time, obtain the flue gas velocity spatial dispersion diffusion index of each control sequence to be analyzed; based on the deviation of the particulate flow intensity from the uniform flow state at future monitoring times, obtain the particulate flow bias index of each control sequence to be analyzed; based on the flue gas velocity spatial dispersion diffusion index and the particulate flow bias index, select the optimal induced draft fan control sequence from all control sequences to be analyzed; control the induced draft fan of the biomass boiler at the current monitoring time based on the optimal induced draft fan control sequence.

[0033] In actual boiler operation, wear behavior primarily stems from high-speed flue gas flow in localized areas and the long-term concentrated transport of fly ash particles. Specifically, when the spatial difference in flue gas velocity continuously expands during future prediction of a candidate control sequence, it indicates the gradual formation of high-speed flue gas channels in localized areas within the boiler, leading to a decrease in flue gas flow uniformity. Furthermore, the continuous concentrated transport of fly ash particles to these localized areas results in the formation of high-particle-flux scouring channels, subjecting the heating surfaces or flue to long-term concentrated scouring, ultimately increasing the risk of localized boiler wear. Therefore, to achieve stable, low-wear control of the biomass boiler with uniform flue gas velocity, this embodiment requires obtaining a control sequence from the analysis sequence that minimizes the spatial dispersion and diffusion of the boiler flue gas flow field and the degree of particle transport bias. In other words, it requires obtaining a control sequence that makes the boiler flue gas flow field more uniform. The control strategy aims to achieve more uniform particle transport and lower risk of local scour. Specifically, this embodiment will first quantify the spatial dispersion and diffusion index of flue gas velocity and the particle flow bias index of each control sequence under analysis based on the degree of dispersion of flue gas velocity over time at future monitoring times and the deviation of particle flow intensity from a uniform flow state at future monitoring times. The spatial dispersion and diffusion index measures whether the difference in flue gas velocity at each monitoring point in the boiler is widening or narrowing within the future prediction time window. A larger spatial dispersion and diffusion index indicates that a local high-speed flue gas channel is forming, and the flow field uniformity is deteriorating. The particle flow bias index measures whether fly ash particles are more uniformly dispersed or more concentrated in a certain area within the future prediction time window. A larger particle flow bias index indicates that a concentrated particle scour channel is forming.

[0034] Based on the above analysis, this embodiment will next obtain the spatial dispersion diffusion index of the flue gas velocity of each control sequence to be analyzed based on the degree of dispersion of the flue gas velocity at future monitoring times and the degree of diffusion over time. It will also obtain the particle flow bias index of each control sequence to be analyzed based on the degree of deviation of the particle flow intensity from the uniform flow state at future monitoring times. The process of obtaining the spatial dispersion diffusion index of the flue gas velocity and the particle flow bias index of the b-th control sequence to be analyzed in the set of all control sequences to be analyzed will be described as an example. The process of obtaining the spatial dispersion diffusion index of the flue gas velocity and the particle flow bias index of the b-th control sequence to be analyzed is as follows:

[0035] First, the standard deviation of all parameters belonging to the flue gas velocity dimension in the future state vector corresponding to each future monitoring time of the b-th control sequence to be analyzed is denoted as the flue gas velocity dispersion of the b-th control sequence to be analyzed at the corresponding future monitoring time. That is, the standard deviation of all parameters belonging to the flue gas velocity dimension in the future state vector corresponding to the t-th future monitoring time of the b-th control sequence to be analyzed is the flue gas velocity dispersion of the b-th control sequence to be analyzed at the t-th future monitoring time. Then, based on the change in the flue gas velocity dispersion of the b-th control sequence to be analyzed between adjacent future monitoring times and the distance between the future monitoring time and the current monitoring time, the spatial dispersion and diffusion index of the flue gas velocity of the b-th control sequence to be analyzed is obtained. And the expression of the spatial dispersion and diffusion index of the flue gas velocity of the b-th control sequence to be analyzed is:

[0036]

[0037] in, Let be the spatial discrete diffusion index of the smoke velocity for the b-th control sequence to be analyzed, and N be the total number of future monitoring times within the preset future prediction time window. This represents the dispersion of the flue gas velocity of the b-th control sequence to be analyzed at the (t+1)-th future monitoring time within a preset future prediction time window. Let f(t) represent the dispersion of flue gas velocity of the b-th control sequence to be analyzed at the t-th future monitoring time within the preset future prediction time window. max() is the maximum value function, t is the position number of the preset time within the preset future prediction time window, and tanh() is the hyperbolic tangent function, which is used for normalization.

[0038] The maximum value function in this section is only intended to retain positive changes (i.e., the part where the flue gas velocity dispersion increases and diffuses), while negative or zero changes are set to 0; This represents the time decay weighting factor; the smaller t is, The larger the value, the more direct the flow field diffusion has on the current induced draft fan, which means that the earlier or closer to the current monitoring time, the higher the weight is given to positive changes or diffusion. A value greater than 0 indicates that the flue gas velocity in a local area of ​​the boiler shows a positive increasing trend between the t-th and t+1-th future monitoring times. The larger, The larger; and The larger the value, the more likely the spatial difference in flue gas velocity will show a positive growth trend and continue to expand in the future prediction process. This indicates that the local flue gas deviation is gradually enhanced, the uniformity of the boiler flue gas flow field is reduced, and the probability of high-speed flue gas flow forming in local areas is increased. Therefore, the risk of local scouring and wear on the heating surface corresponding to the b-th control sequence to be analyzed is higher. This suggests that if this control strategy is implemented, it may lead to the formation of local high-speed flue gas flow channels and the deterioration of flow field uniformity, resulting in a significant increase in the risk of local wear. Therefore, the probability that the b-th control sequence to be analyzed is the optimal induced draft fan control sequence is smaller; conversely, the smaller the value, the lower the probability. The smaller the value, the more stable or even improved the spatial dispersion of the flue gas flow field of the b-th control sequence to be analyzed will be in the future prediction process. There is no obvious trend of diffusion and deterioration. This indicates that the spatial distribution of the flue gas flow field inside the boiler will be relatively uniform and stable in the future, and the risk of local high-speed flue gas flow will be low. It also indicates that the control strategy helps to maintain the stability of the flow field structure and reduce the risk of local wear. The greater the probability that it will be selected as the optimal induced draft fan control sequence.

[0039] Next, the representative value of uniform particulate flow and the representative particulate flow intensity at each future monitoring time within the preset future prediction time window under the control of the b-th control sequence to be analyzed are obtained. The representative value of uniform particulate flow is... M represents the total number of monitoring points in the biomass boiler. The process for obtaining the representative particulate flow intensity at the t-th future monitoring time within the preset future prediction time window under the b-th control sequence to be analyzed is as follows: For the p-th monitoring point in the set of all monitoring points in the biomass boiler, in the future state vector corresponding to the t-th future monitoring time of the b-th control sequence to be analyzed, the product of the two parameters belonging to the flue gas velocity dimension and the particulate concentration dimension among all parameters belonging to the p-th monitoring point is denoted as the particulate flow intensity of the p-th monitoring point at the t-th future monitoring time under the b-th control sequence to be analyzed. The result of proportionally normalizing the particulate matter flow intensity at all monitoring points at the t-th future monitoring time under the control of the b-th control sequence to be analyzed is denoted as the normalized intensity corresponding to the monitoring point at the t-th future monitoring time under the control of the b-th control sequence to be analyzed. The maximum value among the normalized intensities corresponding to each monitoring point at the t-th future monitoring time under the control of the b-th control sequence to be analyzed is the representative particulate matter flow intensity at the t-th future monitoring time under the control of the b-th control sequence to be analyzed. That is, the representative particulate matter flow intensity at the t-th future monitoring time under the control of the b-th control sequence to be analyzed is... , Let p be the particulate matter flow intensity at the p-th monitoring point in the set of monitoring points at the t-th future monitoring time under the control of the b-th control sequence to be analyzed. Represents the set of monitoring points. The larger the value, the more instantaneous particulate matter flows at the p-th monitoring point at the t-th future monitoring time. This represents the proportion of particulate flow intensity at monitoring point p in the overall boiler at the t-th future monitoring time. A larger ratio indicates that a large number of particles are concentrated and transported to the area of ​​monitoring point p, thus increasing the risk of local scour in that area. The normalized result of the average difference between the representative particulate flow intensity and the representative uniform particulate flow value at each future monitoring time within the preset future prediction time window under the control of the b-th control sequence to be analyzed is denoted as the particulate flow bias index of the b-th control sequence to be analyzed. The expression for the particulate flow bias index of the b-th control sequence to be analyzed is:

[0040]

[0041] in, Let N be the particle flow bias index for the b-th control sequence to be analyzed, and N be the total number of future monitoring times within the preset future prediction time window. Let b be the representative particulate matter flow intensity at the t-th future monitoring time under the control of the b-th control sequence to be analyzed. is a representative value for uniform particulate matter flow, and tanh() is the hyperbolic tangent function, which is used for normalization.

[0042] The purpose is to identify the region with the most concentrated particulate matter flow and the highest intensity at the t-th future monitoring time under the control of the b-th control sequence to be analyzed, which represents the most uneven spatial distribution or the maximum concentration of particulate matter at the t-th future monitoring time under the control of the b-th control sequence to be analyzed. This is also the baseline value under ideal uniform conditions. That is, if there are M monitoring points in the boiler, under absolutely uniform particulate transport conditions, the proportion of particulate flow intensity at each monitoring point should be... ; The difference between the particulate flow intensity in the region with the highest particulate concentration at time t under the control of the b-th control sequence to be analyzed and the particulate flow intensity under the ideal uniform state is represented. This difference quantifies the degree to which particulate transport deviates from the uniform state at time t. The larger the difference, the more serious the phenomenon of particulate concentration in local areas. Local areas are also prone to forming high particulate flux transport channels and obvious flow deviation in flue gas field. This can also lead to continuous concentrated scouring of local heated surfaces or flue areas by fly ash particles, and the risk of local wear will also increase significantly. The larger, The larger; It represents the average intensity of the concentrated transport trend of fly ash particles carried by boiler flue gas in the entire future prediction time window under the control of the b-th control sequence to be analyzed. The larger the value, the more concentrated the particle transport becomes in localized areas, increasing the dominance of these localized areas in the overall particle transport. This indicates a significant particle transport bias in the boiler flue gas field, suggesting a higher probability of forming sustained high-scouring channels in localized areas, resulting in a higher risk of localized wear. It also means that after executing the control sequence to be analyzed, particle transport gradually and continuously concentrates in localized areas, further reducing the probability that this control sequence will be selected as the optimal induced draft fan control sequence. Conversely, a smaller value indicates a lower probability of particle transport being selected as the optimal induced draft fan control sequence. A smaller value means that after executing the control sequence to be analyzed, the spatial transport distribution of particulate matter is more uniform and dispersed, and it also indicates that the control sequence to be analyzed is more likely to be selected as the optimal induced draft fan control sequence.

[0043] After obtaining the spatial discrete diffusion index and particle flow bias index of each control sequence to be analyzed, the optimal induced draft fan control sequence is selected from all the control sequences by combining these indices. The specific process of selecting the optimal induced draft fan control sequence based on the spatial discrete diffusion index and particle flow bias index of each control sequence is as follows: the inverse mapping result between the spatial discrete diffusion index of the smoke velocity of each control sequence and the mean of the particle flow bias index of the corresponding control sequence is recorded as the selection confidence value of the corresponding control sequence. The expression for the selection confidence value of the b-th control sequence is: 1 minus Yes Perform inverse mapping processing. And This reflects the changes in the spatial uniformity of flue gas velocity and the concentration trend of fly ash particle transport under the b-th control sequence to be analyzed. The larger the value, the more the spatial dispersion of the boiler flue gas flow field continues to expand. At the same time, fly ash particles gradually concentrate and are transported to local areas, increasing the probability of high-speed flue gas flow and high particle throughput transport channels in local areas. This reduces the stability of the boiler flue gas flow field and increases the risk of continuous concentrated scouring of local heating surfaces or flue areas, resulting in a significant increase in the risk of local wear. The smaller the value, the better the spatial uniformity of the boiler flue gas field corresponding to the b-th control sequence to be analyzed within the future prediction time window, the more balanced the particle transport distribution, and the lower the risk of local wear.

[0044] Finally, the control sequence to be analyzed corresponding to the largest selection confidence value is recorded as the optimal induced draft fan control sequence. That is, the optimal induced draft fan control sequence is the control sequence to be analyzed with the largest selection confidence value. Furthermore, selecting the control sequence to be analyzed corresponding to the largest selection confidence value for control not only meets the basic operating requirements among the candidate control sequences, but also maximizes the uniformity of the flue gas flow field and the dispersion of particle transport, thereby minimizing the risk of local wear.

[0045] In this embodiment, after obtaining the optimal induced draft fan control sequence, the induced draft fan of the biomass boiler is controlled based on the optimal induced draft fan control sequence at the current monitoring time. That is, the induced draft fan of the biomass boiler is controlled based on the control parameters of the first step in the optimal induced draft fan control sequence at the current monitoring time, which is also the current control cycle. Subsequently, in the next control cycle, boiler operating data is re-acquired and the state vector and prediction model are updated. The above prediction, screening, evaluation and optimization process is repeated, thereby forming a closed-loop system for online monitoring and low-wear regulation of biomass boiler induced draft fan based on a rolling optimization mechanism, realizing continuous optimization control of boiler flue gas flow field uniformity and particle transport balance. Moreover, how the induced draft fan is controlled based on the optimal induced draft fan control sequence in this embodiment is consistent with the traditional method of controlling the induced draft fan after obtaining the optimal induced draft fan control sequence.

[0046] Thus, this embodiment completes the online monitoring and control of low wear at constant flue gas velocity in a biomass boiler. Furthermore, this embodiment directly embeds the flue gas velocity spatial dispersion diffusion index and particle flow bias index into the model predictive control framework's control mechanism, enabling the boiler to operate in a state with the most uniform flue gas flow field and the most dispersed particle scouring. This, in turn, effectively suppresses the occurrence and development of localized wear while ensuring the safe and stable operation of the boiler.

[0047] In summary, this embodiment first monitors the operating parameters of the monitoring points and induced draft fans in the biomass boiler online to obtain the current actual state vector of the biomass boiler at the current monitoring time. Based on the current actual state vector, MPC, and the state-space prediction model of MPC, it obtains the candidate induced draft fan control sequence set at the current monitoring time, the cost function value of each candidate induced draft fan control sequence in the candidate induced draft fan control sequence set, and the future state vector corresponding to each future monitoring time in the preset future prediction time window. Then, based on the overall degree to which the parameters in the future state vector exceed the preset safe operation constraint range, it obtains the constraint violation amount of each candidate induced draft fan control sequence. The system selects control sequences to be analyzed based on constraint violations and cost function values. Then, based on the dispersion of flue gas velocity at future monitoring times and its diffusion over time, it obtains the flue gas velocity spatial dispersion diffusion index for each control sequence. Based on the deviation of particulate flow intensity from a uniform flow state at future monitoring times, it obtains the particulate flow bias index for each control sequence. Based on the flue gas velocity spatial dispersion diffusion index and the particulate flow bias index, the optimal induced draft fan control sequence is selected from all control sequences. Finally, the induced draft fan of the biomass boiler is controlled based on the optimal induced draft fan control sequence at the current monitoring time. Furthermore, this embodiment, by directly embedding the flue gas velocity spatial dispersion diffusion index and the particulate flow bias index into the model predictive control framework's regulation mechanism, enables the boiler to operate in a state with the most uniform flue gas flow field and the most dispersed particulate scouring, thereby suppressing the occurrence and development of localized wear and improving the long-term stability and reliability of the biomass boiler.

[0048] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for online monitoring and control of low wear at constant flue gas velocity in a biomass boiler, characterized in that, The method includes the following steps: By monitoring the operating parameters of the monitoring points and induced draft fans in the biomass boiler online, the current actual state vector of the biomass boiler at the current monitoring time is obtained. Based on the current actual state vector, MPC, and the state space prediction model of MPC, a set of candidate induced draft fan control sequences at the current monitoring time, the cost function value of each candidate induced draft fan control sequence in the set of candidate induced draft fan control sequences, and the future state vector corresponding to each future monitoring time in the preset future prediction time window are obtained. The state vector includes parameters in the dimensions of flue gas velocity and particulate matter concentration. Based on the overall degree to which the parameters in the future state vector exceed the preset safe operation constraint range, the constraint violation amount of each candidate induced draft fan control sequence is obtained, and the control sequence to be analyzed is selected based on the constraint violation amount and the cost function value. Based on the dispersion of flue gas velocity at future monitoring times and the diffusion degree over time, the flue gas velocity spatial dispersion diffusion index of each control sequence to be analyzed is obtained. Based on the deviation of the particulate flow intensity from the uniform flow state at future monitoring times, the particulate flow bias index of each control sequence to be analyzed is obtained. Based on the flue gas velocity spatial dispersion diffusion index and the particulate flow bias index, the optimal induced draft fan control sequence is selected from all control sequences to be analyzed. The induced draft fan of the biomass boiler is controlled based on the optimal induced draft fan control sequence at the current monitoring time.

2. The method for online monitoring and control of low wear at constant flue gas velocity in a biomass boiler as described in claim 1, characterized in that, The method for obtaining the constraint violation amount of the i-th candidate induced draft fan control sequence includes: The sum of the constraint violation degrees of all parameters in all future state vectors corresponding to the i-th candidate induced draft fan control sequence is denoted as the constraint violation amount of the i-th candidate induced draft fan control sequence; the constraint violation degree of the j-th parameter in the future state vector corresponding to the i-th candidate induced draft fan control sequence at t future monitoring times is... max() is the function to find the maximum value. Let j be the value of the j-th parameter in the future state vector corresponding to the i-th candidate induced draft fan control sequence at t future monitoring times. This represents the maximum value within the preset safe operating constraint range for the parameter type to which the j-th parameter belongs. This is the minimum value of the preset safe operation constraint range for the parameter type to which the j-th parameter belongs.

3. The method for online monitoring and control of low wear at constant flue gas velocity in a biomass boiler as described in claim 1, characterized in that, Methods for screening control sequences to be analyzed include: Among the cost function values ​​of all candidate induced draft fan control sequences with a constraint violation of 0, the largest cost function value is selected and recorded as the basic operating cost value; candidate induced draft fan control sequences with cost function values ​​less than the basic operating cost value are all recorded as control sequences to be analyzed.

4. The method for online monitoring and control of low wear at constant flue gas velocity in a biomass boiler as described in claim 1, characterized in that, Methods for obtaining the spatial discrete diffusion index of smoke velocity include: The standard deviation of all parameters belonging to the flue gas velocity dimension in the future state vector corresponding to each future monitoring time of each control sequence to be analyzed is denoted as the flue gas velocity dispersion of the corresponding control sequence to be analyzed at the corresponding future monitoring time. Based on the change in the dispersion of flue gas velocity between adjacent future monitoring times for each control sequence to be analyzed, and the distance between the future monitoring time and the current monitoring time, the spatial dispersion diffusion index of flue gas velocity for each control sequence to be analyzed is obtained.

5. The method for online monitoring and control of low wear at constant flue gas velocity in a biomass boiler as described in claim 4, characterized in that, The expression for the spatial discrete diffusion exponent of the smoke velocity of the b-th control sequence to be analyzed is: ； in, Let be the spatial discrete diffusion index of the smoke velocity for the b-th control sequence to be analyzed, and N be the total number of future monitoring moments in the preset future prediction time window. This represents the dispersion of the flue gas velocity of the b-th control sequence to be analyzed at the (t+1)-th future monitoring time within a preset future prediction time window. Let f(t) represent the dispersion of flue gas velocity of the b-th control sequence to be analyzed at the t-th future monitoring time. max() is the maximum value function, t is the position number of the preset time in the preset future prediction time window, and tanh() is the hyperbolic tangent function.

6. The method for online monitoring and control of low wear at constant flue gas velocity in a biomass boiler as described in claim 1, characterized in that, The method for obtaining the particle flow bias index of the b-th control sequence to be analyzed includes: The normalized result of the mean difference between the representative particulate flow intensity and the uniform representative value of particulate flow at each future monitoring time within the preset future prediction time window under the control of the b-th control sequence to be analyzed is denoted as the particulate flow bias index of the b-th control sequence to be analyzed; the uniform representative value of particulate flow is... M represents the total number of monitoring points in the biomass boiler; The representative particulate matter flow intensity at the t-th future monitoring time within the preset future prediction time window under the control of the b-th control sequence to be analyzed is the maximum value among the normalized intensities corresponding to each monitoring point at the t-th future monitoring time. The normalized intensities corresponding to each monitoring point at the t-th future monitoring time are the result of proportional normalization of the particulate matter flow intensity at each monitoring point at the t-th future monitoring time. The particulate matter flow intensity at the p-th monitoring point at the t-th future monitoring time is the product of the two parameters of the b-th control sequence to be analyzed at the t-th future state vector, which belong to the p-th monitoring point and belong to the flue gas velocity dimension and the particulate matter concentration dimension.

7. The method for online monitoring and control of low wear at constant flue gas velocity in a biomass boiler as described in claim 1, characterized in that, Methods for obtaining the optimal induced draft fan control sequence include: The inverse mapping result between the spatial discrete diffusion index of smoke velocity of each control sequence to be analyzed and the mean of the particle flow bias index of the corresponding control sequence to be analyzed is recorded as the selection confidence value of the corresponding control sequence to be analyzed, and the control sequence to be analyzed corresponding to the largest selection confidence value is recorded as the optimal induced draft fan control sequence.

Images