A method for predicting the combustion efficiency of a biomass combined heat and power system

Abstract

The present application provides a kind of biomass cogeneration system combustion efficiency prediction method, the method obtains the historical data of biomass cogeneration system, clusters historical data, extracts the combustion rule under different working conditions;Adaptive adjustment of fuzzy area of prediction model, the combustion rule of system is optimized by particle swarm algorithm;Based on fuzzy system and the parameter training prediction model after optimization;Acquire the measured parameter data of biomass cogeneration system, to predict the combustion efficiency with the trained prediction model.The present application quantitatively and qualitatively analyzes the multi-condition of biomass cogeneration system;Under multi-condition, it can adaptively adjust the static parameters of model;Compared with traditional algorithm, it can improve the prediction accuracy under the premise of ensuring the running speed, and has strong interpretability.

Description

Technical Field

[0001] This invention pertains to the technical field of general control or regulation systems; functional units of such systems; and monitoring or testing devices for such systems or units, and particularly relates to a method for predicting the combustion efficiency of a biomass cogeneration system using a data-driven soft computing approach. Background Technology

[0002] The rapid development of the thermal power industry in recent years has given biomass a place as a fuel. The large amounts of straw and forest biomass produced annually by major agricultural producing countries urgently need further development and utilization.

[0003] Biomass cogeneration systems are systems that use biomass as fuel to produce electricity. They not only effectively alleviate over-reliance on traditional fossil fuels but also improve the utilization rate of waste resources and reduce environmental pollution, leading to their widespread application in the energy sector. Circulating fluidized bed (CFBB) boilers are key equipment in biomass cogeneration systems, and their operation directly determines the stability of various system conditions. Therefore, combustion efficiency is currently used to measure the performance of CFBB boilers. Combustion efficiency mainly refers to the effective heat absorbed by the working fluid from the total heat of the fuel. Real-time evaluation is necessary to optimize the combustion process in biomass cogeneration systems.

[0004] Currently, direct methods, indirect methods, and soft calculation methods are mainly used to predict the combustion efficiency of biomass cogeneration systems. Direct methods require specialized equipment and technical support, are costly, and are significantly affected by the combustion process, easily influenced by variations in the combustion process and sensor accuracy, thus affecting the accuracy of the measurement results. Indirect methods measure parameters in the fuel and flue gas, such as calorific value, oxygen content, flue gas temperature, and fuel carbon content, and then calculate the boiler's combustion efficiency using empirical formulas. Indirect methods are relatively simple and convenient to calculate combustion efficiency, have lower costs, and are relatively easy to operate, allowing for rapid on-site detection and calculation; however, they cannot account for uncertainties in the combustion process.

[0005] Considering that soft computing methods can handle nonlinear relationships, are suitable for modeling complex systems, do not rely on theoretical models, can make predictions using existing data, and do not require expensive experimental equipment, they have a promising future for predicting combustion efficiency. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention proposes a method for predicting the combustion efficiency of a biomass cogeneration system. This method utilizes fuzzy association rules to establish close relationships between performance variables of the boiler combustion system, thereby predicting boiler efficiency.

[0007] The technical solution adopted by this invention to solve its technical problem is a method for predicting the combustion efficiency of a biomass cogeneration system. The method acquires historical data of the biomass cogeneration system, clusters the historical data, extracts combustion rules under different operating conditions, adaptively adjusts the fuzzy region of the prediction model, and optimizes the combustion rules of the system using a particle swarm optimization algorithm.

[0008] A prediction model is generated based on a fuzzy system; the test parameter data of the biomass cogeneration system is obtained to train the prediction model; and the combustion efficiency is predicted using the trained prediction model.

[0009] Preferably, extracting combustion rules under different operating conditions includes the following steps:

[0010] Step 1.1: Given historical data D = {X, y}, where X is the input variable, y is the target variable, and X ∈ R. M*n y∈R M *1 M represents the total number of historical data points, and n represents the total number of input variable attributes.

[0011] Step 1.2: Divide the system combustion state into different operating conditions based on performance variables using the fuzzy mean algorithm;

[0012] Step 1.3: For each operating condition k, a soft calculation method is used, through the dataset d k Use fuzzy association rules to establish the relationship between system performance variables and variables, and determine the mapping relationship f. k :(X)→y, generate the prediction model and the corresponding combustion rules, D=∑d k .

[0013] Preferably, in step 1.3, the dataset d for different working conditions... k Evenly distribute each variable within the operating condition. The domain interval is divided into c i A blurry area Mark the fuzzy region of each variable as i is the index of any operating condition variable; the target variable y k Divided into c0 fuzzy regions, marked as

[0014] The fuzzification of data is achieved by assigning membership functions to the fuzzy regions of each input and output variable using Gaussian membership functions. Based on this, the region where the real input data is located is identified. A fuzzy rule base is generated based on the training set, with one fuzzy rule generated for each data point.

[0015] Preferably, the Gaussian membership function assignment of the fuzzy region of the input variable satisfies

[0016]

[0017] The membership assignment of the fuzzy region of the output variable satisfies

[0018]

[0019] In l i When the membership degree reaches its maximum value, then This section describes the fuzzification of data, transforming the specific numerical values ​​of variables in the record into fuzzy regions corresponding to the variables in the prediction model. The membership function represents the attribute of the fuzzy record, and the fuzzy rule is a combination of fuzzy partitions (operating conditions) of different variables.

[0020] Preferably, with and Let represent the membership functions of the input and output variables corresponding to the p-th rule under working condition k, respectively. The working condition fuzzy system (prediction model) is as follows:

[0021]

[0022] in, The predicted value of the fuzzy system; the membership degree of the p-th rule for the k-th working condition. When the maximum value is reached, the output value is taken as... N represents the total number of rules, and i represents the sequence number of any working condition variable.

[0023] Preferably, the combustion rules of the fuzzy system are adaptively adjusted based on the values ​​of the maximum permissible error and the root mean square error.

[0024] Preferably, the number of training sets for condition k is N. k Target error function Where y is the true value, and α is the maximum allowable error. When E>α, c is increased. i The value of is determined until E ≤ α is satisfied.

[0025] Preferably, a swarm of random particles is initialized using a particle swarm optimization algorithm, and the optimal solution is found iteratively; the predicted value output by the system is then rewritten as... in,

[0026] In each iteration, the particle updates the parameter θ using the following formula.

[0027]

[0028]

[0029] w(t)=(w ini -w end (tt) max ) / t max +w end

[0030] Where θ is the particle's position, V is the particle's velocity, and t is the particle's velocity. max Let w(t) be the maximum number of iterations, w(t) be the inertia factor, d1 be the individual learning factor, d2 be the global learning factor, and r1 and r2 be random numbers between zero and one. This represents the historical best value of an individual particle at the t-th iteration. The optimal value in the global history of the particle swarm is obtained at the t-th iteration.

[0031] The particle swarm optimization algorithm adjusts the fuzzy rules, namely... The value of makes the predicted output more accurate.

[0032] This invention proposes a method for predicting the combustion efficiency of a biomass cogeneration system. The method acquires historical data of the biomass cogeneration system, clusters the historical data, and extracts combustion rules under different operating conditions; adaptively adjusts the fuzzy region of the prediction model, and optimizes the system's combustion rules using a particle swarm optimization algorithm; trains the prediction model based on the fuzzy system and the optimized parameters; acquires the test parameter data of the biomass cogeneration system, and uses the trained prediction model to predict the combustion efficiency.

[0033] The beneficial effects of this invention are mainly reflected in:

[0034] (1) Quantitative and qualitative analysis of the biomass cogeneration system under multiple operating conditions was conducted;

[0035] (2) Under multiple working conditions, the static parameters of the model can be adaptively adjusted;

[0036] (3) Compared with traditional algorithms, it can improve prediction accuracy while ensuring running speed and has strong interpretability. Attached Figure Description

[0037] Figure 1 This is a flowchart of the present invention;

[0038] Figure 2 This is a flowchart of the model optimization process of the present invention;

[0039] Figure 3 The figure shows the simulation results of the algorithm of this invention. Detailed Implementation

[0040] The present invention will be further described in detail below with reference to embodiments, but the scope of protection of the present invention is not limited thereto.

[0041] This invention relates to a method for predicting the combustion efficiency of a biomass cogeneration system. The method involves establishing a prediction model for the combustion efficiency of the biomass cogeneration system; initializing the parameters of the prediction model and parameter optimization algorithm; training the prediction model offline; acquiring parameter data of the biomass cogeneration system; and obtaining the predicted value of the combustion efficiency of the biomass cogeneration system from the trained prediction model. The method is applied to a biomass cogeneration system, where the prediction model outputs the predicted system combustion efficiency value by inputting variable data.

[0042] Specifically, to address the time-varying nature of biomass cogeneration systems, historical data is clustered into different operating conditions, and combustion rules are extracted under these conditions. Based on the values ​​of the maximum permissible error and root mean square error, the static parameters of the fuzzy system are adaptively adjusted. Finally, the fuzzy rules of the fuzzy system are deeply optimized using a particle swarm optimization algorithm to improve the approximation accuracy of the fuzzy model, thereby completing the prediction of combustion efficiency. After initializing the parameters of the prediction model and parameter optimization algorithm, the prediction model is trained offline to obtain parameter data of the biomass cogeneration system. The trained prediction model then obtains the predicted value of the combustion efficiency of the biomass cogeneration system.

[0043] The main execution part of this invention is implemented on a process control computer that predicts the concentration of pollutants in a biomass cogeneration system.

[0044] The application process of this invention is divided into three stages:

[0045] (1) Parameter initialization: Set appropriate initial values ​​for the number of fuzzy regions divided by the fuzzy system in the proposed method, and import the relevant variable data of the biomass cogeneration system and the corresponding boiler efficiency into the model.

[0046] (2) Offline training: such as Figure 1 As shown.

[0047] First, given historical data D = {X, y}, X ∈ R M*n y∈R M*1 , where X is the input variable, y is the target variable, M is the total number of historical data, and n is the total number of input variable attributes.

[0048] The main components of the biomass circulating fluidized bed boiler include: furnace, separator, superheater, economizer, return feeder, and dust collector. The combustion process can be summarized as follows: biomass fuel, such as straw, is crushed, mixed with a desulfurizing agent, and then fed into the furnace. It burns under the heating action of the fluidized material, while primary and secondary air fans respectively supply primary and secondary air to the furnace for fuel combustion. Under the action of the high-speed upward airflow, the fuel moves towards the upper part of the furnace. Coarse fuel particles are carried into the dense phase zone of the furnace for combustion, while fine particles are suspended and burned with the flue gas in the dilute phase zone. Some of the entrained fine particles enter the return feeder under the action of the separator and are then sent back to the furnace for secondary combustion. The flue gas exchanges heat with the heating surface at the tail end through the superheater and is then purified by the dust collector before being discharged. The heat generated during the entire combustion process is captured by the heat exchanger for power generation and heating. The collected data includes 23 input variable attributes: generator active power, instantaneous value of main steam flow, boiler main steam header outlet temperature, primary air main pipe flow, primary hot air outlet temperature, secondary air main pipe flow, secondary hot air outlet temperature, flue gas temperature (left), flue gas temperature (right), combustion chamber sub-boiling temperature (front), combustion chamber sub-boiling temperature (back), combustion chamber mid-boiling temperature (left), combustion chamber mid-temperature (right), combustion chamber mid-temperature (left), combustion chamber mid-temperature (right), furnace outlet temperature (left), furnace outlet temperature (right), low-temperature superheater outlet oxygen content (left), low-temperature superheater outlet oxygen content (right), cyclone outlet temperature (left), cyclone outlet temperature (right), air preheater outlet flue gas oxygen content (left), air preheater outlet flue gas oxygen content (right).

[0049] Different boiler operating conditions directly lead to different operating characteristics of the boiler combustion process, resulting in different optimization strategies under the same modeling and optimization methods. Therefore, it is necessary to divide the overall system into different partitions. The Fuzzy Means (FCM) algorithm is used to divide the boiler operating conditions into different partitions. This approach directly reduces the size of the modeling dataset, which has a beneficial impact on reducing computation time during the modeling process.

[0050] For example, using unit load as a constraint variable, the dataset is divided into groups consisting of... The three partitions, Represents the constraint variable, the k-th fuzzy partition of the unit load uc; the dataset d after partitioning different performance variables. k = {X, y}, X∈R N*n y∈R N*1 D = ∑d k N represents the total number of data points within a partition. A data-driven approach is used for each partition, through the dataset d. k Determine the mapping relationship f k :(X)→y, thereby modeling and generating a fuzzy logic system and the corresponding combustion rules.

[0051] The generator active power performance variable is used as an external constraint variable. This performance variable is determined by expert prior knowledge and generally consists of external variables such as the properties of the combustible material and boiler performance. Different values ​​of the performance variable represent different operating conditions of the boiler. The original dataset D is clustered using the Fuzzy Means (FCM) algorithm. In this embodiment, the original data is divided into three fuzzy partitions: low, medium, and high, where 1 ≤ k ≤ 3. The partitioned dataset d... k = {X, y}, X∈R N*n y∈R N*1 , where D=∑d k For each partition We can use dataset d k Determine the mapping relationship f k :(X)→y, thereby generating combustion rules for the fuzzy system.

[0052] After dividing the historical data into different partitions, the datasets d in each partition are... k Distribute each variable within the partition evenly. The domain interval is divided into c i A blurry area Different variables It can have different numbers of fuzzy regions c. i Mark the fuzzy region of each variable as The target variable y k Divided into c0 fuzzy regions, marked as

[0053] The Gaussian membership function is used to assign membership functions to the fuzzy regions of each input and output variable. Equation (1) is the Gaussian membership function assignment for the fuzzy regions of the input variables, and Equation (2) is the membership function assignment for the fuzzy regions of the output variables. If in l i When the membership degree reaches its maximum value, then

[0054]

[0055]

[0056] A fuzzy rule base is generated based on the training set, with one fuzzy rule generated for each data point. The rule for the k-th partition is the p-th fuzzy rule. Defined as: if for and for …,and for Then y k for p = 1, 2, ..., N.

[0057] From the perspective of data mining technology, support reflects the degree to which fuzzy rules support real-world data. Only when a fuzzy rule has high support does it have practical significance. The support value can be used as a basis for deleting redundant rules, and it can also be used as the weight of the fuzzy rule during programming, affecting the accuracy of the prediction model. This can be optimized using algorithms such as particle swarm optimization. Therefore, in the process of extracting fuzzy rules from numerical data, support can be improved by using it as a standard for a specific fuzzy sampling space. Support can be defined as Equation (3), where... and Let represent the membership functions of the input and output variables corresponding to the p-th rule in the k-partition.

[0058]

[0059] The above formula can be simplified as:

[0060]

[0061] The partitioned fuzzy system takes the form of equation (5), where, The predicted value of the fuzzy system; the membership degree of the p-th rule for the k-th partition. When the maximum value is reached, the output value is taken as...

[0062]

[0063] Equation (5) can be rewritten as:

[0064]

[0065]

[0066]

[0067]

[0068] like Figure 2 As shown, the extracted combustion rules are first initially optimized: it is determined whether the error loss function E is less than the maximum allowable error α. If not, the number of fuzzy partitions c is increased. i Re-extract the fuzzy rules; if successful, use the particle swarm optimization algorithm to perform deep optimization on the static parameters and initialize the maximum number of iterations t. max Update the function fitness value, historical best position, and global best position of each particle in the particle swarm.

[0069] Assuming the number of training sets in the k-partition is R, the target error function E is defined by formula (10), where y is the true value:

[0070]

[0071] The parameter optimization process of the model is as follows: Figure 2 As shown, since the fuzzy system is a generalized approximator, by adding a new fuzzy region C in the input-output space... i Improving performance is always possible. Therefore, a maximum allowable error α is introduced; when E > α, C is increased. i The value of c is calculated until the condition is met. In this embodiment, c is ultimately obtained. i The value is 8, and the value of c0 is 50.

[0072] The static parameters of the fuzzy system are further optimized using the particle swarm optimization (PSO) algorithm. The initial number of particles is set to 100, and the maximum number of iterations is 100. The PSO algorithm initializes a swarm of random particles (random solutions). Then, the optimal solution is found through iteration. In each iteration, the particles update the parameter θ using the following formula.

[0073]

[0074]

[0075] w(t)=(w ini -w end (tt) max ) / t max +w end (12)

[0076] θ is the particle's position, V is the particle's velocity, and t is the particle's velocity. max t is the maximum number of iterations; w(t) is the inertia factor. Dynamic values ​​often yield better optimization results than fixed values. The larger the value, the stronger the global optimization ability and the weaker the local optimization ability. A linear decreasing weight strategy is commonly used, as shown in formula (13); c1 is the individual learning factor, c2 is the global learning factor, and d1 and d2 are random numbers between zero and one. This represents the historical best value of an individual particle at the t-th iteration. The value is the global historical optimum of the particle swarm at the t-th iteration.

[0077] (3) Online Prediction: By predicting the variable data of the biomass cogeneration system online, the corresponding boiler efficiency value can be predicted in real time. The simulation prediction diagram is shown below. Figure 3 As shown.

[0078] Different operating conditions directly lead to different characteristics of biomass cogeneration systems. Therefore, it is necessary to divide the whole system into sections.

[0079] The present invention relates to a computer-readable storage medium storing a program for predicting the combustion efficiency of a biomass cogeneration system, wherein the program, when executed by a processor, implements the aforementioned method for predicting the combustion efficiency of a biomass cogeneration system; and also relates to a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the aforementioned method for predicting the combustion efficiency of a biomass cogeneration system.

[0080] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention 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.

[0081] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. 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 illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0082] 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 function specified in one or more boxes.

[0083] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment 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.

[0084] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0085] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for predicting the combustion efficiency of a biomass cogeneration system, the method comprising: The method acquires historical data from a biomass cogeneration system, clusters the historical data, and extracts combustion rules under different operating conditions, including the following steps: ​ Step 1.1: Given historical data on biomass fuel combustion Where X is the input variable and y is the target variable. , M represents the total number of historical data points, and n represents the total number of input variable attributes. The input variables include generator active power, instantaneous main steam flow rate, boiler main steam header outlet temperature, primary air main flow rate, primary hot air outlet temperature, secondary air main flow rate, secondary hot air outlet temperature, left exhaust gas temperature, right exhaust gas temperature, front sub-boiling temperature of the combustion chamber, rear sub-boiling temperature of the combustion chamber, left-side mid-boiling temperature of the combustion chamber, right-side mid-temperature of the combustion chamber, left-side mid-temperature of the combustion chamber, left-side outlet temperature of the furnace, right-side outlet temperature of the furnace, left-side outlet oxygen content of the low-temperature superheater, right-side outlet oxygen content of the low-temperature superheater, left-side outlet temperature of the cyclone separator, right-side outlet temperature of the cyclone separator, left-side outlet oxygen content of the air preheater, and right-side outlet oxygen content of the air preheater. The generator active power performance variable is used as an external constraint variable. Step 1.2: Divide the system combustion state into different operating conditions based on performance variables using the fuzzy mean algorithm; Step 1.3: For each operating condition k, a soft calculation method is used, through the dataset... Use fuzzy association rules to establish the relationships between system performance variables and variables, and determine the mapping relationships. Generate prediction models and corresponding combustion rules. Data sets for different working conditions Evenly distribute each variable within the operating condition. The domain interval is divided into c i A blurry area Mark the fuzzy region of each variable as , where i is the index of any working condition variable; the target variable Divided into A blurred region, marked as ; The membership function of each input and output variable fuzzy region is assigned using the Gaussian membership function. A fuzzy rule base is generated based on the training set, with one fuzzy rule generated for each data point. In the process of extracting fuzzy rules from numerical data, the support is used as the standard for a specific fuzzy sampling space to improve the process. The fuzzy region of the prediction model is adaptively adjusted, and the extracted combustion rules are initially optimized: it is determined whether the error loss function E is less than the maximum allowable error. If it fails, increase the number of fuzzy partitions. Re-extract the fuzzy rules; if successful, rewrite the fuzzy system of operating conditions as follows: ,in, , , , For the predicted value of the fuzzy system, For the optimization parameters of the particle swarm, The output value for the p-th rule in condition k is... Let be the membership degree of all fuzzy rules under condition k. The product of the normalized membership degrees of the p-th rule. Let N be the membership function of the input variable corresponding to the p-th rule under condition k, and N be the total number of rules; the particle swarm optimization algorithm is used to evaluate the system parameters. To optimize the system, the inertia factor is dynamically adjusted based on a linearly decreasing weight strategy. A prediction model is generated based on a fuzzy system; the test parameter data of the biomass cogeneration system is obtained to train the prediction model; and the combustion efficiency is predicted using the trained prediction model.

2. The method for predicting the combustion efficiency of a biomass cogeneration system according to claim 1, characterized in that: The Gaussian membership function assignment of the fuzzy region of the input variable satisfies ； The membership assignment of the fuzzy region of the output variable satisfies ； exist When the membership degree reaches its maximum value, then .

3. The method for predicting the combustion efficiency of a biomass cogeneration system according to claim 2, characterized in that: by and Let the membership functions of the input and output variables corresponding to the p-th rule in working condition k be denoted as , and the working condition fuzzy system be . ， in, The predicted value of the fuzzy system; the membership degree of the p-th rule for the k-th working condition. When the maximum value is reached, the output value is taken as... N is the total number of rules, and i is the index of any working condition variable.

4. The method for predicting the combustion efficiency of a biomass cogeneration system according to claim 1, characterized in that: Let the number of training sets for condition k be... Target error function Where y is the true value, and For the maximum permissible error, when At that time, increase The value, until satisfied. .

5. The method for predicting the combustion efficiency of a biomass cogeneration system according to claim 1, characterized in that: A swarm of random particles is initialized using the particle swarm optimization algorithm, and the optimal solution is found through iteration. In each iteration, the particle updates its parameters using the following formula. , ， ， ， in, V is the particle's position, and V is the particle's velocity. The maximum number of iterations, Inertia factor For individual learning factors, As a global learning factor, , A random number between zero and one. Let be the historical best value of the particle at the t-th iteration. The value is the global historical optimum of the particle swarm at the t-th iteration.

Images