Particle filter based gradient estimation of methane production rate and extreme value search method and system for anaerobic digestion process

Abstract

The application discloses a kind of gradient estimation of anaerobic digestion process methane yield based on particle filtering and extreme value search method and system, and it is related to anaerobic digestion process intelligent control technical field.The application is aimed at the optimal working point of anaerobic digestion process methane yield easy to deviate, and there are problems of insufficient precision, slow optimization speed and steady-state oscillation in traditional extreme value search algorithm relying on high-low pass filter set gradient estimation, gradient estimation is converted into state estimation problem by tangent approximation method, particle filtering algorithm is introduced, and gradient accurate estimation is realized after weighted particle approximation system state posterior probability distribution, replace traditional filter set;While designing amplitude self-adaptive attenuation module to update disturbance signal amplitude.The application significantly improves the gradient estimation precision, speeds up the optimization convergence speed, eliminates steady-state oscillation, and is suitable for anaerobic digestion process feeding optimization control.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for anaerobic digestion processes, and in particular to a method and system for estimating the methane yield gradient and searching for extreme values ​​in anaerobic digestion processes based on particle filtering. Background Technology

[0002] Anaerobic digestion technology is a core green technology for the harmless disposal and resource utilization of organic waste. It can convert various complex organic wastes such as livestock and poultry manure, kitchen waste, and industrial organic wastewater into biogas with methane as the main component. It has both ecological and environmental benefits and clean energy production value, and has broad engineering application prospects.

[0003] During the engineering operation of anaerobic digestion, the system operating conditions are affected by multiple factors, including substrate concentration fluctuations, changes in feeding rate, accumulation of inhibitors, and dynamic evolution of microbial community activity. The optimal operating point for methane yield continues to deviate, and its optimal setpoint is difficult to determine accurately. It is easy to encounter problems such as excessive substrate addition leading to system acidification or insufficient addition leading to low yield, which seriously restricts the operating efficiency and gas production stability of the anaerobic digestion system.

[0004] Currently, extreme value search algorithms are commonly used to optimize the feed control in anaerobic digestion processes. This method extracts the gradient information of methane yield relative to dilution rate by injecting a small sinusoidal perturbation signal to achieve extreme value optimization, and has the advantage of not relying on a precise system mechanism model. However, existing extreme value search algorithms rely on high-pass and low-pass filter banks to achieve gradient estimation. During modulation and demodulation, some high-frequency components are inevitably ignored, resulting in insufficient gradient estimation accuracy, which directly limits the optimization convergence speed. At the same time, there is also the problem of steady-state oscillation, making it difficult to meet the dual requirements of fast optimization and stable steady-state operation.

[0005] As a sequential Monte Carlo state estimation method, particle filtering does not require the assumption that noise follows a specific distribution. It can accurately handle the state estimation problem of nonlinear systems by approximating the posterior probability distribution of the system state through weighted particles, providing a brand-new technical path to solve the industry pain point of insufficient gradient estimation accuracy. Summary of the Invention

[0006] To address these issues, this invention provides a method and system for methane yield gradient estimation and extreme value search in anaerobic digestion based on particle filtering. This method solves the problems in existing technologies, such as the difficulty in determining the optimal setpoint for methane yield in anaerobic digestion, the tendency for substrate fluctuations and microbial community dynamics to cause deviations in the optimal yield operating point, and the fact that existing extreme value search algorithms rely on high-pass and low-pass filter banks to achieve gradient estimation while ignoring some high-frequency components generated by modulation and demodulation, resulting in insufficient gradient estimation accuracy, limited optimization convergence speed, and steady-state oscillation defects.

[0007] To address the aforementioned technical problems, embodiments of the present invention provide a method for estimating the methane yield gradient and searching for extreme values ​​in an anaerobic digestion process based on particle filtering. This method includes the following steps:

[0008] Step S1: Determine the control input variables and controlled output variables of the anaerobic digestion process, and determine the sinusoidal perturbation signal form of the extreme value search algorithm;

[0009] Step S2: Build a basic framework for feeding control in the anaerobic digestion process based on the extreme value search algorithm, and establish a dynamic mapping relationship between input variables and output variables;

[0010] Step S3: Transform the gradient estimation problem of methane yield relative to input dilution rate into a state estimation problem of nonlinear system, and construct the corresponding discrete state-space equation;

[0011] Step S4: Based on the discrete state space equation, the posterior probability distribution of the system state is approximated by a weighted set of particles to achieve accurate estimation of the gradient information of methane yield relative to dilution rate.

[0012] Step S5: Input the gradient information output by the particle filter into the pre-established amplitude adaptive attenuation module to complete the dynamic update of the amplitude of the sinusoidal disturbance signal. Simultaneously, update the estimated value of the optimal dilution rate by the gradient information through the integrator, thereby realizing the rapid extreme value optimization and steady-state oscillation stability control of methane yield in the anaerobic digestion process.

[0013] Preferably, in step S1, the control input variable for the anaerobic digestion process is the dilution rate. The acceleration rate of the substrate is controlled by the dilution rate to regulate the matrix concentration; the controlled output variable is the methane yield. Input signal From the initial input Sinusoidal perturbation signal, estimated value of optimal input dilution rate It consists of three superimposed parts, and the expression is:

[0014] ;

[0015] in, The amplitude of the sinusoidal disturbance signal. The angular frequency of the sinusoidal disturbance signal. It is a continuous-time variable.

[0016] Preferably, in step S2, the execution flow of the basic framework for feed control in the anaerobic digestion process based on the extreme value search algorithm is as follows:

[0017] Step S21: Initialize the input dilution rate for the anaerobic digestion process;

[0018] Step S22: Superimpose a sinusoidal perturbation signal with a set amplitude and frequency onto the input dilution rate;

[0019] Step S23: Measure the response of the output methane yield to the disturbance signal, and extract the gradient information of the methane yield relative to the dilution rate by filtering through a high-pass and low-pass filter bank.

[0020] Step S24: Multiply the gradient information by the integrator gain and input it into the integrator to iteratively update the dilution rate estimate;

[0021] Step S25: Determine whether the gradient information has converged to 0. If it has not converged, proceed to step S22. If it has converged, output the current optimal dilution rate.

[0022] Preferably, in step S3, the method of transforming the gradient estimation problem of methane yield relative to the input dilution rate into a state estimation problem of a nonlinear system and constructing the corresponding discrete state-space equations includes:

[0023] Provided that the amplitude A of the sinusoidal perturbation signal is sufficiently small, the input perturbation causes the output methane yield to exhibit an approximately sinusoidal oscillation, as expressed by:

[0024] ;

[0025] in, for methane yield at time t, The initial methane yield, This represents the gradient of methane yield relative to dilution rate. The angular frequency of the sinusoidal disturbance signal. It is a continuous-time variable;

[0026] Transform the methane yield output expression into a linear regression form:

[0027] ;

[0028] in, This is the baseline term for the methane yield output;

[0029] Let the state variable , This transforms the gradient estimation problem into a problem involving state variables. , The problem of state estimation for discrete systems.

[0030] Preferably, the construction of the corresponding discrete state-space equations includes state transition equations and observation equations, specifically in the following form:

[0031] State transition equation: ;

[0032] Observation equation: ;

[0033] in, for The system state vector at time t. For system process noise, To observe noise in the system, The discrete state-space equation satisfies the system observability condition, where the time delay step is the observed data.

[0034] Preferably, the time delay step in the observation equation The corresponding duration is or ,in The period of the sinusoidal perturbation signal is chosen to ensure the observability of the discrete-time system, enabling the observed data to effectively reflect the correlation between the input perturbation and the output methane yield response.

[0035] Preferably, in step S4, the method for accurately estimating the gradient information of methane yield relative to dilution rate by approximating the posterior probability distribution of the system state through a weighted particle ensemble based on the discrete state-space equation includes:

[0036] Step S41: Initialize the particle set and set the number of particles. Random sampling from the initial state prior distribution 100 particles form the initial particle set, and each particle is assigned an equal initial weight.

[0037] Step S42: Using the system state transition distribution as the importance sampling distribution, from... Sampled from the particle set at time t. Predicted particles at specific times;

[0038] Step S43: Combining The observed methane yield at time t is used to calculate the importance weight of each predicted particle through the observation likelihood function, and the weights of all particles are normalized so that the sum of the weights of all particles is 1.

[0039] Step S44: Approximate the posterior expectation of the system state by weighted particle summation, and obtain... The state estimate at time t is the gradient information of the output methane yield relative to the dilution rate during the anaerobic digestion process;

[0040] Step S45: When the number of effective particles is lower than the set threshold, perform a resampling operation on the particle set, remove particles with too small weights, and duplicate high-weight particles to maintain the diversity of the particle set;

[0041] Repeat steps S42 to S45 to achieve real-time sequential estimation of gradient information.

[0042] Preferably, in step S5, the pre-established calculation expression for the amplitude adaptive attenuation module is:

[0043] ;

[0044] in, The amplitude of the sinusoidal disturbance signal. The initial amplitude of the sinusoidal disturbance signal is given. For custom attenuation rate adjustment coefficient, The gradient information is output by the particle filter. When the gradient information amplitude is large during the optimization stage, a high disturbance signal amplitude is maintained to improve the gradient estimation sensitivity and optimization convergence speed. When the gradient information approaches 0 in the steady state stage, the disturbance signal amplitude is adaptively reduced until the impact of the disturbance on the system operation can be ignored, thus eliminating steady-state oscillations.

[0045] Preferably, in step S5, the method of inputting the gradient information output by the particle filter into a pre-established amplitude adaptive attenuation module to complete the dynamic update of the amplitude of the sinusoidal disturbance signal, and simultaneously updating the estimated value of the optimal dilution rate by iteratively updating the gradient information through an integrator, includes:

[0046] Step S51: Initialize the input dilution rate for the anaerobic digestion process and simultaneously initialize the particle set for particle filtering;

[0047] Step S52: Superimpose the sinusoidal perturbation signal updated by the amplitude adaptive attenuation module onto the input dilution rate;

[0048] Step S53: Real-time measurement Time and The output methane yield response data to the perturbation signal at any given time is used to input the dilution rate data and the methane yield response data into a particle filter, which outputs the gradient information of the methane yield relative to the dilution rate. ;

[0049] Step S54: Transfer gradient information The value is multiplied by the integrator gain and then input into the integrator to iteratively update the estimate of the optimal dilution rate.

[0050] Step S55: Determine gradient information If the convergence reaches 0, proceed to step S52. If the convergence occurs, continue to output the current optimal dilution rate to maintain the optimal yield operation of the anaerobic digestion system.

[0051] This invention also provides a particle filtering-based system for estimating the methane yield gradient and searching for extreme values ​​in an anaerobic digestion process. This system is used to implement the aforementioned particle filtering-based method for estimating the methane yield gradient and searching for extreme values ​​in an anaerobic digestion process, specifically including:

[0052] The anaerobic digestion process variable and disturbance signal definition module is used to determine the control input variables and controlled output variables of the anaerobic digestion process, and to determine the sinusoidal disturbance signal form of the extreme value search algorithm;

[0053] The module for building the basic framework of extreme value search feeding control is used to build the basic framework of feeding control for anaerobic digestion based on the extreme value search algorithm and to establish a dynamic mapping relationship between input variables and output variables.

[0054] The gradient estimation-state estimation transformation and system modeling module is used to transform the gradient estimation problem of methane yield relative to the input dilution rate into the state estimation problem of a nonlinear system and construct the corresponding discrete state-space equations.

[0055] The gradient information estimation module based on particle filtering is used to accurately estimate the gradient information of methane yield relative to dilution rate by approximating the posterior probability distribution of the system state through a weighted set of particles based on the discrete state space equation.

[0056] The adaptive extremum optimization and steady-state control module is used to input the gradient information output by the particle filter into the pre-established amplitude adaptive decay module to complete the dynamic update of the amplitude of the sinusoidal disturbance signal. Simultaneously, the gradient information is iteratively updated through the integrator to update the estimated value of the optimal dilution rate, thereby realizing rapid extremum optimization and steady-state oscillation stability control of methane yield in the anaerobic digestion process.

[0057] As can be seen from the above technical solutions, this invention application has the following beneficial effects:

[0058] (1) This invention uses a particle filter algorithm to replace the gradient estimation method of high-pass and low-pass filter banks in the traditional extreme value search algorithm. The gradient estimation problem is transformed into a nonlinear system state estimation problem by using the tangent approximation method. The weighted particle set is used to approximate the posterior probability distribution of the system state. There is no need to assume that the noise follows a specific distribution. The correlation information between the input disturbance and the output response can be completely preserved. The gradient estimation distortion problem caused by the loss of high-frequency components during the modulation and demodulation process of the traditional filter bank is avoided. The gradient estimation result is highly matched with the real dynamic of methane yield, providing accurate directional guidance for extreme value optimization.

[0059] (2) Based on the high-precision gradient estimation results, this invention can accurately lock the extreme value search direction, avoiding optimization iteration oscillation and direction errors caused by gradient information deviation. Simulation verification shows that compared with the traditional extreme value search algorithm, this invention can shorten the time for the dilution rate to converge to the optimal value from 490h to 388h, significantly improving the optimization efficiency; at the same time, while accelerating the optimization speed, it can ensure that the methane yield eventually converges to the optimal steady-state value consistent with the traditional method, improving the dynamic response speed without sacrificing the long-term steady-state gas production performance of the anaerobic digestion system.

[0060] (3) This invention designs an amplitude adaptive attenuation module that is linked to gradient information. During the optimization phase, it maintains a high perturbation amplitude to improve the gradient estimation sensitivity, and during the steady-state phase, it adaptively reduces the perturbation amplitude until its impact on the system is negligible. This fundamentally eliminates the steady-state oscillation problem inherent in traditional extreme value search algorithms, achieving a balance between rapid optimization and stable operation. At the same time, this invention does not rely on a precise mechanism model of the anaerobic digestion process. It can adaptively adapt to complex working conditions such as substrate concentration fluctuations, changes in bacterial community activity, and inhibitor accumulation. It can also be used as a soft measurement method to cope with actual engineering scenarios involving sensor distortion and measurement lag, possessing strong robustness and engineering application value. Attached Figure Description

[0061] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly described below. Referring to the drawings will make the features and advantages of the present invention clearer. The drawings are illustrative and should not be construed as limiting the present invention in any way. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0062] Figure 1 This is a flowchart of a method for estimating the gradient of methane yield and searching for extreme values ​​in an anaerobic digestion process based on particle filtering, provided by the present invention.

[0063] Figure 2 This is a block diagram of the methane yield extremum search control for the anaerobic digestion process based on particle filtering in this invention;

[0064] Figure 3 This is a graph showing the dilution rate variation during extreme value search in the anaerobic digestion process based on particle filtering in this invention.

[0065] Figure 4 This is a graph showing the methane yield variation during extreme value search in the anaerobic digestion process based on particle filtering in this invention.

[0066] Figure 5 This is a methane yield prediction graph based on particle filtering for extreme value search in the anaerobic digestion process in this invention.

[0067] Figure 6 This is a block diagram of a methane yield gradient estimation and extreme value search system based on particle filtering for anaerobic digestion provided by the present invention. Detailed Implementation

[0068] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0069] This invention provides a method and system for methane yield gradient estimation and extreme value search in anaerobic digestion based on particle filtering. It is applicable to anaerobic digestion processes using organic waste such as livestock and poultry manure, kitchen waste, and industrial organic wastewater as substrates. The system is used for online optimization control of the feeding rate and real-time extreme value search of methane yield in anaerobic digestion. It addresses the technical problems in existing technologies, such as difficulty in determining the optimal setpoint for methane yield in anaerobic digestion, fluctuations in substrate concentration, accumulation of inhibitors, and changes in microbial activity, which can easily lead to deviations in the optimal yield operating point. Furthermore, existing extreme value search algorithms rely on high-pass and low-pass filter banks to achieve gradient estimation, neglecting some high-frequency components generated by modulation and demodulation, resulting in insufficient gradient estimation accuracy, limited convergence speed, and steady-state oscillation defects.

[0070] In this invention, the proposed particle filtering-based method for estimating the methane yield gradient and searching for extreme values ​​in anaerobic digestion is referred to as the PFUESC algorithm, and the traditional algorithm for searching extreme values ​​in anaerobic digestion without steady-state oscillation is referred to as the UESC algorithm.

[0071] Example 1:

[0072] This embodiment provides a method for estimating the methane yield gradient and searching for extreme values ​​in an anaerobic digestion process based on particle filtering. Figure 1 and Figure 2 As shown, the method includes the following steps:

[0073] In step S1, the control input variables and controlled output variables of the anaerobic digestion process are determined, and the sinusoidal perturbation signal form of the extreme value search algorithm is determined.

[0074] This step first clarifies the core control variables and controlled variables of the anaerobic digestion process, laying the foundation for subsequent extreme value optimization and gradient estimation. Specifically:

[0075] Control input variable: dilution rate The bottom flow acceleration rate affects the substrate concentration in the anaerobic reactor through the dilution rate, and is the core control parameter for anaerobic digestion feeding control.

[0076] Controlled output variable: methane yield It is the core evaluation index of gas production efficiency in the anaerobic digestion process, and also the target for extreme value optimization in this invention.

[0077] A small sinusoidal perturbation signal is superimposed on the input dilution rate to excite the system's output response, thereby extracting gradient information. The final input signal... From the initial input Estimates of sinusoidal perturbation signal and optimal input dilution rate It consists of three parts superimposed, and its expression is:

[0078] ;

[0079] in, The amplitude of the sinusoidal disturbance signal. The angular frequency of the sinusoidal disturbance signal. It is a continuous-time variable.

[0080] In step S2, a basic framework for feeding control of the anaerobic digestion process based on the extreme value search algorithm is established, and a dynamic mapping relationship between input variables and output variables is established.

[0081] This step establishes the basic framework for feed control using the traditional extreme value search algorithm, clarifying the fundamental logic of extreme value optimization and providing a framework foundation for subsequent improvements to the algorithm based on particle filtering. The execution flow of the basic framework for feed control in the anaerobic digestion process based on the extreme value search algorithm is as follows:

[0082] Step S21: Initialize the input dilution rate of the anaerobic digestion process and set the initial operating parameters;

[0083] Step S22: Add an amplitude value to the input dilution rate. , frequency is A sinusoidal disturbance signal is injected into the system to provide a continuous excitation signal;

[0084] Step S23: Measure the response of the output methane yield to the disturbance signal, using a cutoff frequency of [frequency value missing]. The high-pass filter and cutoff frequency are The low-pass filter bank is used for filtering, retaining the AC response component and removing the DC component to obtain the gradient information ξ of methane yield relative to dilution rate.

[0085] Step S24: Transfer gradient information With integrator gain Multiply the gradient signal to accelerate the convergence speed of the optimization, and then input it into the integrator to iteratively update the dilution rate estimate.

[0086] Step S25: Determine gradient information If the system converges to 0, proceed to step S22 to continue iterative optimization. If the system converges to 0, it indicates that the system has reached the optimal operating point for methane yield and continues to output the current optimal dilution rate.

[0087] In step S3, the gradient estimation problem of methane yield relative to the input dilution rate is transformed into a state estimation problem of a nonlinear system, and the corresponding discrete state-space equation is constructed.

[0088] This step transforms the gradient estimation problem into a state estimation problem using the tangent approximation method, constructing a matching mathematical model for the application of the particle filter algorithm. The specific implementation method is as follows:

[0089] The amplitude of the sinusoidal disturbance signal Provided the input is small enough, a sinusoidal perturbation will cause the output methane yield to also exhibit an approximately sinusoidal shape, with the methane yield initially revolving around its initial value. Oscillation, its expression is:

[0090] ;

[0091] in, for methane yield at time t, The initial methane yield, This represents the gradient of methane yield relative to the dilution rate, which is the core target quantity that this invention needs to accurately estimate. The angular frequency of the sinusoidal disturbance signal. It is a continuous-time variable.

[0092] Based on the above expression, the methane yield output expression for the anaerobic digestion process is transformed into a linear regression form:

[0093] ;

[0094] in, is the baseline term for the methane yield output, and is a slowly varying parameter that changes over time.

[0095] In the discrete-time domain, particle filtering is used to estimate the aforementioned unknown parameters. For and At that time, let the state variable , ,in By setting the time step to discrete, the gradient estimation problem can be further transformed into a problem involving parameters. , This problem involves the state estimation of discrete systems, and further, the gradient information of the output methane yield relative to the input dilution rate in the anaerobic digestion process can be estimated using a particle filter.

[0096] Based on the above definition of state variables, the corresponding discrete state-space equations are constructed, including state transition equations and observation equations, in the following specific form:

[0097] State transition equation: ;

[0098] Observation equation: ;

[0099] in, for The system state vector at time t. For system process noise, To observe noise in the system, The discrete state-space equation satisfies the system observability condition, where the time delay step is the observed data.

[0100] Regarding the time delay step of the observation data The choice of duration must ensure the observability of the discrete-time system, and its corresponding duration must be determined by the period of the perturbation signal. Determined, usually selected as or ,in The period of the sinusoidal perturbation signal is used. This choice ensures that the observation data effectively reflects the correlation between the input perturbation and the output response, providing reliable observational data support for accurate estimation of gradient information, and ultimately achieving gradient estimation of the output methane yield relative to the input dilution rate. In this embodiment, the time delay step is... The corresponding duration selection is .

[0101] In step S4, based on the discrete state space equation, the posterior probability distribution of the system state is approximated by a weighted set of particles, thereby achieving an accurate estimation of the gradient information of methane yield relative to dilution rate.

[0102] This step is the core innovation of this invention. It replaces the high-pass and low-pass filter bank gradient estimation method in the traditional extremum search algorithm with a particle filter algorithm. Accurate gradient estimation is achieved through a sequential Monte Carlo method, addressing the core pain point of insufficient gradient estimation accuracy in traditional methods. The specific execution steps of the particle filter algorithm for gradient estimation are as follows:

[0103] Step S41: Initialize the particle set and define the number of particles. Random sampling from the initial state prior distribution 100 particles, forming the initial particle set Each particle is assigned an equal initial weight to obtain an initial weighted particle set. ,in , .

[0104] Step S42: Based on the importance sampling distribution, from Sampled from the particle set at time t. Predicted particles at time The choice of importance sampling distribution directly affects the estimation accuracy of particle filtering. In this embodiment, the system state transition distribution is used as the importance sampling distribution to ensure that the predicted particles can follow the dynamic changes of the system state well and adapt to the nonlinear dynamic characteristics of the anaerobic digestion process.

[0105] Step S43: Combining Observed methane yield at time The importance weight of each predicted particle is calculated based on the observed likelihood function. The larger the weight value, the higher the degree of matching between the particle and the observed methane yield, and the better it reflects the true state of the system.

[0106] After calculating the weights of each single particle, the weights of all particles are normalized using the following formula:

[0107] ;

[0108] Make the sum of the weights of all particles equal to 1 to ensure the reasonable probability distribution of the weights.

[0109] Step S44: Approximate the posterior expectation of the system state using weighted particle summation to obtain... State estimate at time 1 This estimate represents the gradient information of methane yield relative to dilution rate during anaerobic digestion. ;

[0110] Step S45: As particle weights gradually differentiate during iteration, some particle weights approach 0, while others approach 1, leading to reduced particle diversity and particle degradation, which severely affects the accuracy of state estimation. Therefore, this step sets an effective particle count threshold. When the particle count falls below a set threshold, a resampling operation is performed on the particle set to remove particles with excessively low weights and duplicate high-weight particles, ensuring both the diversity of the particle set and the accuracy of the estimation. The formula for calculating the effective particle count is as follows:

[0111] ;

[0112] By repeating steps S42 to S45 above, real-time sequential estimation of the gradient information of the output methane yield relative to the input dilution rate in the anaerobic digestion process can be achieved.

[0113] In step S5, the gradient information output by the particle filter is input into the pre-established amplitude adaptive attenuation module to complete the dynamic update of the amplitude of the sinusoidal disturbance signal. Simultaneously, the gradient information is iteratively updated through the integrator to update the estimated value of the optimal dilution rate, thereby realizing rapid extreme value optimization and steady-state oscillation stability control of methane yield in the anaerobic digestion process.

[0114] This step constructs a dual-loop closed-loop control link. On the one hand, it solves the steady-state oscillation problem of the traditional extreme value search algorithm through an amplitude adaptive decay module. On the other hand, it iteratively updates the optimal dilution rate through gradient information to achieve rapid extreme value optimization, and finally completes the closed-loop optimization control of feeding in the anaerobic digestion process.

[0115] First, an amplitude adaptive attenuation module is pre-established, and its calculation expression is as follows:

[0116] ;

[0117] in, The amplitude of the sinusoidal disturbance signal. The initial amplitude of the sinusoidal disturbance signal is given. For custom attenuation rate adjustment coefficient, This refers to the gradient information output by the particle filter.

[0118] The working mechanism of this amplitude adaptive decay module is as follows: during the algorithm optimization phase, gradient information... When the amplitude of the gradient is large, maintaining a high-amplitude perturbation signal enhances the sensitivity and accuracy of gradient estimation, accelerating the convergence speed of the anaerobic digestion process towards the optimal dilution rate operating point. However, once the anaerobic digestion process enters a steady-state operation phase, the gradient information... As the value approaches zero, the amplitude of the disturbance signal is adaptively reduced until its impact on the system becomes negligible. This design ensures rapid optimization of the dilution rate and eliminates steady-state oscillations that occur when extreme value search algorithms are applied to feed control in anaerobic digestion processes, achieving both rapid process optimization and steady-state oscillation-free extreme value search feed control.

[0119] Based on the aforementioned particle filter gradient estimation and amplitude adaptive attenuation module, the improved methane yield extremum search and optimization process for the anaerobic digestion process of this invention specifically includes the following steps:

[0120] Step S51: Initialize the input dilution rate for the anaerobic digestion process and simultaneously initialize the particle set for particle filtering;

[0121] Step S52: Superimpose the sinusoidal perturbation signal updated by the amplitude adaptive attenuation module onto the input dilution rate;

[0122] Step S53: Real-time measurement Time and The output methane yield response data to the perturbation signal at each time step is used. The input dilution rate data and methane yield response data are then input into a particle filter. The particle filter algorithm is used to estimate the gradient information of methane yield relative to the dilution rate. ;

[0123] Step S54: Transfer gradient information The value is multiplied by the integrator gain to accelerate convergence, and then fed into the integrator to iteratively update the estimate of the optimal dilution rate.

[0124] Step S55: Determine gradient information If the convergence reaches 0, and if it does not converge, proceed to step S52 to continue iterative optimization. If it converges, continue to output the current optimal dilution rate to maintain the optimal yield operation of the anaerobic digestion system.

[0125] To verify the technical effectiveness of the method proposed in this invention, this embodiment compares and verifies the PFUESC algorithm with the traditional UESC algorithm through simulation experiments. The simulation results are as follows: Figure 3 , Figure 4 , Figure 5 As shown, the details are as follows:

[0126] Figure 3 The dilution rate comparison curves show that the PFUESC algorithm optimizes the dilution rate significantly faster than the UESC algorithm, reaching the optimal value at 388 hours. In contrast, the UESC algorithm only converged to the optimal value at 490h, indicating that the PFUESC algorithm successfully accelerated the convergence speed by using particle filtering to estimate the gradient.

[0127] Figure 4 The methane yield comparison curve shows the maximum methane yield under the PFUESC algorithm during the optimization process. The maximum methane yield under the UESC algorithm is The steady-state methane yield values ​​for both algorithms are... The methane yields eventually converged to the same steady-state value, indicating that in terms of long-term operating performance, both methods can drive the methane yield of the anaerobic digestion process to a maximum production state. The method proposed in this invention improves the convergence speed without sacrificing the steady-state gas production performance of the system.

[0128] Figure 5The image shows the effect curve of particle filtering on methane yield estimation in the PFUESC algorithm. The estimated methane yield curve highly coincides with the actual yield curve, especially in the steady-state stage where the deviation between the estimated and actual values ​​is extremely small. This proves that particle filtering can accurately capture the dynamic changes in methane yield and effectively achieve gradient estimation of the output methane yield relative to the input dilution rate. This simulation result also provides a new method for online estimation of methane yield in anaerobic digestion processes, particularly suitable for practical engineering scenarios with sensor distortion or measurement lag. It can be used as a soft sensing tool to assist in the monitoring and control of anaerobic digestion processes.

[0129] Example 2:

[0130] like Figure 6 As shown, this embodiment provides a particle filtering-based system for estimating the methane yield gradient and searching for extreme values ​​in an anaerobic digestion process. This system is used to implement the particle filtering-based method for estimating the methane yield gradient and searching for extreme values ​​in an anaerobic digestion process described in Embodiment 1. Specifically, it includes a module for defining anaerobic digestion process variables and disturbance signals, a module for building a basic framework for extreme value search and feed control, a module for gradient estimation-state estimation transformation and system modeling, a module for accurate gradient information estimation based on particle filtering, and a module for adaptive extreme value optimization and steady-state control. The specific implementation methods of each module are as follows:

[0131] The anaerobic digestion process variable and disturbance signal definition module 100 is used to determine the control input variables and controlled output variables of the anaerobic digestion process, and to determine the sinusoidal disturbance signal form of the extreme value search algorithm. The specific implementation of this module is consistent with the content of step S1 in Embodiment 1, and will not be repeated here.

[0132] The extreme value search feeding control framework construction module 200 is used to build the basic framework for feeding control in the anaerobic digestion process based on the extreme value search algorithm, and to establish a dynamic mapping relationship between input variables and output variables. The specific implementation of this module is the same as the content of step S2 in Example 1, and will not be repeated here.

[0133] The gradient estimation-state estimation transformation and system modeling module 300 is used to transform the gradient estimation problem of methane yield relative to the input dilution rate into a state estimation problem of a nonlinear system, and construct the corresponding discrete state-space equations. The specific implementation of this module is consistent with the content of step S3 in Example 1, and will not be repeated here.

[0134] The gradient information estimation module 400 based on particle filtering is used to accurately estimate the gradient information of methane yield relative to dilution rate by approximating the posterior probability distribution of the system state through a weighted particle set based on the discrete state space equation. The specific implementation of this module is consistent with step S4 in Embodiment 1, and will not be repeated here.

[0135] The adaptive extremum optimization and steady-state control module 500 is used to input the gradient information output by the particle filter into a pre-established amplitude adaptive attenuation module to complete the dynamic update of the amplitude of the sinusoidal disturbance signal. Simultaneously, the gradient information is iteratively updated via an integrator to the estimated value of the optimal dilution rate, thereby achieving rapid extremum optimization and steady-state oscillation-free stable control of the methane yield in the anaerobic digestion process. The specific implementation of this module is consistent with step S5 in Example 1, and will not be repeated here.

[0136] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0137] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0138] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0139] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for estimating the gradient of methane yield and searching for extreme values ​​in an anaerobic digestion process based on particle filtering, characterized in that, Includes the following steps: Step S1: Determine the control input variables and controlled output variables of the anaerobic digestion process, and determine the sinusoidal perturbation signal form of the extreme value search algorithm; Step S2: Build a basic framework for feeding control in the anaerobic digestion process based on the extreme value search algorithm, and establish a dynamic mapping relationship between input variables and output variables; Step S3: Transform the gradient estimation problem of methane yield relative to input dilution rate into a state estimation problem of nonlinear system, and construct the corresponding discrete state-space equation; Step S4: Based on the discrete state space equation, the posterior probability distribution of the system state is approximated by a weighted set of particles to achieve accurate estimation of the gradient information of methane yield relative to dilution rate. Step S5: Input the gradient information output by the particle filter into the pre-established amplitude adaptive attenuation module to complete the dynamic update of the amplitude of the sinusoidal disturbance signal. Simultaneously, update the estimated value of the optimal dilution rate by the gradient information through the integrator, thereby realizing the rapid extreme value optimization and steady-state oscillation stability control of methane yield in the anaerobic digestion process.

2. The method for estimating methane yield gradient and searching for extreme values ​​in anaerobic digestion based on particle filtering according to claim 1, characterized in that, In step S1, the control input variable for the anaerobic digestion process is the dilution rate. The acceleration rate of the substrate is controlled by the dilution rate to regulate the matrix concentration; the controlled output variable is the methane yield. Input signal From the initial input Sinusoidal perturbation signal, estimated value of optimal input dilution rate It consists of three superimposed parts, and the expression is: ； in, The amplitude of the sinusoidal disturbance signal. The angular frequency of the sinusoidal disturbance signal. It is a continuous-time variable.

3. The method for estimating methane yield gradient and searching for extreme values ​​in anaerobic digestion based on particle filtering according to claim 1, characterized in that, In step S2, the execution flow of the basic framework for feed control in the anaerobic digestion process based on the extreme value search algorithm is as follows: Step S21: Initialize the input dilution rate for the anaerobic digestion process; Step S22: Superimpose a sinusoidal perturbation signal with a set amplitude and frequency onto the input dilution rate; Step S23: Measure the response of the output methane yield to the disturbance signal, and extract the gradient information of the methane yield relative to the dilution rate by filtering through a high-pass and low-pass filter bank. Step S24: Multiply the gradient information by the integrator gain and input it into the integrator to iteratively update the dilution rate estimate; Step S25: Determine whether the gradient information has converged to 0. If it has not converged, proceed to step S22. If it has converged, output the current optimal dilution rate.

4. The method for estimating the methane yield gradient and searching for extreme values ​​in an anaerobic digestion process based on particle filtering according to claim 1, characterized in that, In step S3, the method of transforming the gradient estimation problem of methane yield relative to the input dilution rate into a state estimation problem of a nonlinear system and constructing the corresponding discrete state-space equations includes: Provided that the amplitude A of the sinusoidal perturbation signal is sufficiently small, the input perturbation causes the output methane yield to exhibit an approximately sinusoidal oscillation, as expressed by: ； in, for methane yield at time t, The initial methane yield, This represents the gradient of methane yield relative to dilution rate. The angular frequency of the sinusoidal disturbance signal. For continuous time variables; Transform the methane yield output expression into a linear regression form: ； in, This is the baseline term for the methane yield output; Let the state variable , This transforms the gradient estimation problem into a problem involving state variables. , The problem of state estimation for discrete systems.

5. The method for estimating methane yield gradient and searching for extreme values ​​in anaerobic digestion based on particle filtering according to claim 4, characterized in that, The construction of the corresponding discrete state-space equations includes state transition equations and observation equations, specifically in the following form: State transition equation: ; Observation equation: ; in, for The system state vector at time t. For system process noise, To observe noise in the system, The discrete state-space equation satisfies the system observability condition, where the time delay step is the observed data.

6. The method for estimating methane yield gradient and searching for extreme values ​​in anaerobic digestion based on particle filtering according to claim 5, characterized in that, The time delay step in the observation equation The corresponding duration is or ,in The period of the sinusoidal perturbation signal is chosen to ensure the observability of the discrete-time system, enabling the observed data to effectively reflect the correlation between the input perturbation and the output methane yield response.

7. The method for estimating methane yield gradient and searching for extreme values ​​in anaerobic digestion based on particle filtering according to claim 1, characterized in that, In step S4, the method for accurately estimating the gradient information of methane yield relative to dilution rate by approximating the posterior probability distribution of the system state through a weighted set of particles based on the discrete state-space equation includes: Step S41: Initialize the particle set and set the number of particles. Random sampling from the initial state prior distribution 100 particles form the initial particle set, and each particle is assigned an equal initial weight. Step S42: Using the system state transition distribution as the importance sampling distribution, from... Sampled from the particle set at time t. Predicting particles at specific times; Step S43: Combining The observed methane yield at time t is used to calculate the importance weight of each predicted particle through the observation likelihood function, and the weights of all particles are normalized so that the sum of the weights of all particles is 1. Step S44: Approximate the posterior expectation of the system state by weighted particle summation, and obtain... The state estimate at time t is the gradient information of the output methane yield relative to the dilution rate during the anaerobic digestion process; Step S45: When the number of effective particles is lower than the set threshold, perform a resampling operation on the particle set, remove particles with too small weights, and duplicate high-weight particles to maintain the diversity of the particle set; Repeat steps S42 to S45 to achieve real-time sequential estimation of gradient information.

8. The method for estimating the methane yield gradient and searching for extreme values ​​in an anaerobic digestion process based on particle filtering according to claim 1, characterized in that, In step S5, the pre-established calculation expression for the amplitude adaptive attenuation module is: ； in, The amplitude of the sinusoidal disturbance signal. The initial amplitude of the sinusoidal disturbance signal is given. For the custom attenuation rate adjustment coefficient, The gradient information is output by the particle filter. When the gradient information amplitude is large during the optimization stage, a high disturbance signal amplitude is maintained to improve the gradient estimation sensitivity and optimization convergence speed. When the gradient information approaches 0 in the steady state stage, the disturbance signal amplitude is adaptively reduced until the impact of the disturbance on the system operation can be ignored, thus eliminating steady-state oscillations.

9. The method for estimating the methane yield gradient and searching for extreme values ​​in an anaerobic digestion process based on particle filtering according to claim 1, characterized in that, In step S5, the method of inputting the gradient information output by the particle filter into a pre-established amplitude adaptive attenuation module to complete the dynamic update of the amplitude of the sinusoidal disturbance signal, and simultaneously updating the estimated value of the optimal dilution rate by iteratively updating the gradient information through an integrator, includes: Step S51: Initialize the input dilution rate for the anaerobic digestion process and simultaneously initialize the particle set for particle filtering; Step S52: Superimpose the sinusoidal perturbation signal updated by the amplitude adaptive attenuation module onto the input dilution rate; Step S53: Real-time measurement Time and The output methane yield response data to the perturbation signal at any given time is used to input the dilution rate data and the methane yield response data into a particle filter, which outputs the gradient information of the methane yield relative to the dilution rate. ; Step S54: Transfer gradient information The value is multiplied by the integrator gain and then input into the integrator to iteratively update the estimate of the optimal dilution rate. Step S55: Determine gradient information If the convergence reaches 0, proceed to step S52. If the convergence occurs, continue to output the current optimal dilution rate to maintain the optimal yield operation of the anaerobic digestion system.

10. A particle filtering-based system for estimating methane yield gradient and searching for extreme values ​​in an anaerobic digestion process, characterized in that, The system is used to implement the particle filtering-based method for estimating the methane yield gradient and searching for extreme values ​​in an anaerobic digestion process as described in any one of claims 1 to 9, specifically including: The anaerobic digestion process variable and disturbance signal definition module is used to determine the control input variables and controlled output variables of the anaerobic digestion process, and to determine the sinusoidal disturbance signal form of the extreme value search algorithm; The module for building the basic framework of extreme value search feeding control is used to build the basic framework of feeding control for anaerobic digestion based on the extreme value search algorithm and to establish a dynamic mapping relationship between input variables and output variables. The gradient estimation-state estimation transformation and system modeling module is used to transform the gradient estimation problem of methane yield relative to the input dilution rate into the state estimation problem of a nonlinear system and construct the corresponding discrete state-space equations. The gradient information estimation module based on particle filtering is used to accurately estimate the gradient information of methane yield relative to dilution rate by approximating the posterior probability distribution of the system state through a weighted set of particles based on the discrete state space equation. The adaptive extremum optimization and steady-state control module is used to input the gradient information output by the particle filter into the pre-established amplitude adaptive decay module to complete the dynamic update of the amplitude of the sinusoidal disturbance signal. Simultaneously, the gradient information is iteratively updated through the integrator to update the estimated value of the optimal dilution rate, thereby realizing rapid extremum optimization and steady-state oscillation stability control of methane yield in the anaerobic digestion process.

Images