Desulfurization system operation optimization method and device based on random forest regression and enumeration hybrid algorithm, storage medium and equipment
By employing a hybrid algorithm of random forest regression and enumeration in the desulfurization system of thermal power plants, a multi-objective optimization model is constructed and the optimal solution is selected. This solves the problems of long training cycles and poor real-time performance in traditional methods, and achieves efficient energy consumption optimization and emission compliance of the desulfurization system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-06
- Publication Date
- 2026-03-10
AI Technical Summary
Existing control methods for desulfurization systems in thermal power plants suffer from inaccurate state prediction and lengthy training cycles due to massive amounts of data, making it difficult to achieve satisfactory results in terms of real-time performance and energy efficiency.
A hybrid algorithm based on random forest regression and enumeration was adopted. By constructing a desulfurization cost model and a multi-objective optimization model, the optimal solution was selected by using the random forest regression prediction model and the enumeration method. The combination of slurry circulation pumps was optimized to reduce energy consumption and ensure that SO2 emissions meet the standards.
It significantly shortened the training time, improved the real-time performance and energy efficiency of the desulfurization system, optimized the combination of circulating pumps, reduced energy consumption, and ensured that SO2 emissions met the standards.
Smart Images
Figure CN121637979A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application belongs to the technical field of operation optimization of a desulfurization system of a thermal power plant, and in particular relates to a desulfurization system operation optimization method based on a random forest regression and an enumeration hybrid algorithm. BACKGROUND
[0002] At present, the energy structure in China is constantly transforming towards new energy. Coal-fired power generation, as a stable energy supply form, will play an important base load role for a period of time, but faces the dual challenges of low carbonization and environmental protection, and controlling its emissions is a key part of environmental sustainable development. The desulfurization system has characteristics such as multivariable coupling and strong nonlinearity, and the traditional control method cannot achieve satisfactory control effect due to inaccurate state prediction.
[0003] Some existing methods for operation optimization of a desulfurization system of a thermal power plant mainly focus on regulating system parameters, data-driven, integrated learning algorithm, etc. The real-time performance of the desulfurization system control is less considered, which has a large lag, and the massive data leads to difficulties in adjusting algorithm parameters and a long training period. Therefore, how to further reduce the training time of the desulfurization system operation optimization, improve the real-time performance of the optimization and energy efficiency under the premise of considering that the control of the desulfurization system meets the national minimum emission standard has become an important topic in current research and practice. SUMMARY
[0004] (I) Technical problem to be solved by the present application
[0005] How to provide a desulfurization system operation optimization method based on a random forest regression and an enumeration hybrid algorithm, which can reduce the training time of operation optimization and improve the real-time performance of optimization.
[0006] (II) Technical solution of the present application
[0007] A desulfurization system operation optimization method based on a random forest regression and an enumeration hybrid algorithm, the method comprising:
[0008] Preprocessing historical data of the desulfurization system to obtain a resource load time series data set;
[0009] Constructing a desulfurization cost model based on desulfurization material consumption cost and energy consumption cost;
[0010] Determining an optimization target according to the desulfurization cost model analysis;
[0011] Constructing a multi-objective optimization model based on desulfurization cost and desulfurization efficiency;
[0012] Based on the resource load time series data set, a hybrid optimization algorithm based on a random forest and an enumeration method is used to screen to obtain an optimal solution that meets the optimization target of the multi-objective optimization model.
[0013] Optionally, the historical data includes boiler load, desulfurization efficiency, raw flue gas flow, net flue gas flow, and inlet and outlet SO2 concentrations.
[0014] Optionally, a desulfurization cost model based on desulfurization material consumption cost and energy consumption cost is constructed, including:
[0015] A material consumption cost model is constructed, which is used to analyze the cost of desulfurization agent limestone, and determine the mapping relationship between unit SO2 removal and material consumption;
[0016] An energy consumption cost model is constructed, which is used to analyze the power consumption of main desulfurization equipment, and determine the mapping relationship between unit SO2 removal and energy consumption.
[0017] Optionally, the material consumption cost model is specifically:
[0018]
[0019] wherein, is the limestone cost, η is the desulfurization efficiency, Q in is the raw flue gas flow, is the inlet SO2 concentration, Ca / S is the calcium-sulfur ratio, λ is the absorbent purity, and β is the absorbent utilization rate, is the limestone unit price, and B is the boiler load;
[0020] The energy consumption cost model is specifically:
[0021]
[0022] wherein, F dh is the power consumption cost, P d is the plant power price, U i is the equipment operating voltage, I i is the equipment operating current, cosθ i is the power factor.
[0023] Optionally, the optimization target is the slurry circulating pump combination mode.
[0024] Optionally, the objective function of the multi-objective optimization model is:
[0025]
[0026] wherein, z i is the start-stop state of the i-th pump, z i = 0 indicates stop, z i = 1 indicates operation, α and β are weights, P i is the power of the i-th pump, SO2 concentration of the clean flue gas, λ s SO2 emission standard.
[0027] Optionally, based on the resource load time series dataset, a hybrid optimization algorithm based on random forest and enumeration method is used to screen the optimal solution that meets the optimization target of the multi-objective optimization model, including:
[0028] A random forest regression prediction model is trained using the resource load time series dataset;
[0029] The trained random forest regression prediction model is used to predict the corresponding key indicators of various pump combinations;
[0030] Various pump combinations and corresponding key indicators are substituted into the multi-objective optimization model to obtain target optimization values;
[0031] The pump combination corresponding to the smallest target optimization value is used as the optimal solution that meets the optimization target of the multi-objective optimization model.
[0032] The application also discloses a desulfurization system operation optimization device based on a random forest regression and enumeration hybrid algorithm, which comprises:
[0033] A preprocessing module configured to preprocess historical data of the desulfurization system to obtain a resource load time series dataset;
[0034] A cost construction module configured to construct a desulfurization cost model based on desulfurization material consumption cost and energy consumption cost;
[0035] A cost analysis module configured to determine an optimization target according to the desulfurization cost model;
[0036] An optimization model construction module configured to construct a multi-objective optimization model based on desulfurization cost and desulfurization efficiency;
[0037] An optimization solving module configured to, based on the resource load time series dataset, use a hybrid optimization algorithm based on random forest and enumeration method to screen an optimal solution that meets the optimization target of the multi-objective optimization model.
[0038] The application also discloses a computer readable storage medium storing a desulfurization system operation optimization program based on a random forest regression and enumeration hybrid algorithm, and the desulfurization system operation optimization program based on the random forest regression and enumeration hybrid algorithm is executed by a processor to implement the above-mentioned desulfurization system operation optimization method based on the random forest regression and enumeration hybrid algorithm.
[0039] The application also discloses a computer device, which comprises a computer readable storage medium, a processor and a desulfurization system operation optimization program based on a random forest regression and enumeration hybrid algorithm stored in the computer readable storage medium.
[0040] (III) beneficial effects
[0041] The desulfurization system operation optimization method, device, storage medium and equipment based on the random forest regression and enumeration hybrid algorithm have the following technical effects relative to existing methods.
[0042] (1) The traditional control method has inaccuracy in state prediction and complexity caused by massive historical data, and the scheme constructs a random forest-enumeration hybrid algorithm, which can directly filter out the optimal configuration and avoid local optimization. The method is based on the feature importance ranking method of the random forest, reduces the input feature dimension, greatly shortens the training time, and significantly improves the real-time performance and energy efficiency of the desulfurization system pump group optimization.
[0043] (2) The scheme combines the prediction ability of the random forest regression and the global search characteristics of the enumeration method. The advantages of the random forest in processing high-dimensional nonlinear problems are maintained, and the massive calculation of the traditional multi-objective optimization algorithm is avoided through the enumeration strategy, which is suitable for rapid optimization and decision-making in actual fields. The advantage of the method is low calculation cost, no need to calculate the Pareto front, faster prediction speed, no need for additional algorithm design, and only the number and depth of trees need to be adjusted.
[0044] (3) The scheme adopts a multi-objective optimization weighted summation method. The weighted summation method simultaneously incorporates two objectives into the optimization search process by constructing a single objective function. This ensures that the optimization algorithm can find a globally optimal compromise solution in the entire solution space, rather than a local optimal or suboptimal solution. The final solution is a systematic, globally optimal balance point with higher engineering practical value. The weighted summation method has strong flexibility. The weight coefficient can be dynamically adjusted according to policy requirements, seasonal changes or market electricity prices, etc., to quickly obtain a new optimal operation strategy.
[0045] In summary, the scheme focuses on the circulating pump with the highest energy consumption in the desulfurization system operation, predicts the optimal operation combination of the slurry circulating pump under each working condition through a random forest regression and enumeration hybrid algorithm, and realizes the minimization of circulating pump energy consumption while ensuring that the SO2 emission meets the standard, thereby achieving significant energy saving and consumption reduction effects. BRIEF DESCRIPTION OF DRAWINGS
[0046] Figure 1 A flow chart of a desulfurization system operation optimization method based on a random forest regression and enumeration hybrid algorithm according to one or more embodiments.
[0047] Figure 2 A principle block diagram of a desulfurization system operation optimization device based on a random forest regression and enumeration hybrid algorithm according to one or more embodiments.
[0048] Figure 3 A schematic diagram of a computer device according to one or more embodiments.
[0049] Figure 4 A schematic diagram of energy consumption comparison and cost difference before and after optimization according to one or more embodiments. DETAILED DESCRIPTION
[0050] In order to make the objects, technical solutions and advantages of the present application clearer, further detailed description will be made to the present application with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not used to limit the present application.
[0051] Before describing various embodiments of the present application in detail, first, the technical concept of the present application is briefly described: the desulfurization system operation optimization method of the thermal power plant has the problems of poor optimization real-time performance and long training period. Therefore, the present application provides a desulfurization system operation optimization method based on a random forest regression and enumeration hybrid algorithm, and the key improvement point is to analyze the desulfurization cost to determine the optimization target and to build a multi-objective optimization model, and to use a hybrid optimization algorithm based on random forest and enumeration method to obtain the optimal solution of the optimization target conforming to the multi-objective optimization model. This method greatly shortens the training time and significantly improves the real-time performance and energy efficiency of the desulfurization system pump group optimization. The specific principles of the desulfurization system operation optimization method based on a random forest regression and enumeration hybrid algorithm of the present application will be described in combination with more embodiments.
[0052] Specifically, as shown in Figure 1 the desulfurization system operation optimization method based on a random forest regression and enumeration hybrid algorithm of the present embodiment includes the following steps:
[0053] Step S10, pre-processing the historical data of the desulfurization system to obtain a resource load time series data set;
[0054] Step S20, building a desulfurization cost model based on the desulfurization material consumption cost and the energy consumption cost;
[0055] Step S30, analyzing and determining the optimization target according to the desulfurization cost model;
[0056] Step S40, building a multi-objective optimization model based on the desulfurization cost and the desulfurization efficiency;
[0057] Step S50, based on the resource load time series data set, a hybrid optimization algorithm based on random forest and enumeration method is used to screen the optimal solution meeting the optimization target of the multi-objective optimization model.
[0058] In one or more embodiments, the historical data includes boiler load, desulfurization efficiency, raw flue gas flow, net flue gas flow, and inlet and outlet SO2 concentration. The pretreatment method includes methods such as outlier deletion, outlier point processing, and missing point filling. The original DCS historical data is converted into reliable high-quality data set through pretreatment to ensure the accuracy and integrity of the data, and form a complete resource load time series data set.
[0059] In one or more embodiments, a desulfurization cost model based on desulfurization material consumption cost and energy consumption cost is constructed, including:
[0060] A material consumption cost model is constructed, which is used to analyze the cost of desulfurization agent limestone, and determine the mapping relationship between unit SO2 removal and material consumption.
[0061] The material consumption cost model is specifically:
[0062]
[0063] In the formula, is the cost of limestone, η is the desulfurization efficiency, Q in is the raw flue gas flow, is the inlet SO2 concentration, Ca / S is the calcium sulfur ratio, λ is the absorbent purity, and β is the absorbent utilization rate, is the unit price of limestone, and B is the boiler load;
[0064] An energy consumption cost model is constructed, which is used to analyze the power consumption of the main desulfurization equipment, and determine the mapping relationship between unit SO2 removal and energy consumption.
[0065] The energy consumption cost model is specifically:
[0066]
[0067] In the formula, F dh is the power consumption cost, P d is the price of auxiliary power, U i is the equipment operating voltage, I i is the equipment operating current, cosθ i is the power factor.
[0068] Exemplarily, the optimization target is calculated by the material consumption cost model and the energy consumption cost model. It is calculated that the energy consumption cost accounts for more than 70% of the total cost, and according to the investigation, the slurry circulating pump group with the largest power consumption is determined as the optimization target.
[0069] In one or more embodiments, the process of building a multi-objective optimization model includes: based on the negative correlation between the cost and efficiency of the desulfurization system, the two are in conflict with each other, because the minimum desulfurization system cost (economy) and the maximum desulfurization efficiency (environmental protection) cannot be optimized at the same time, so the weighted summation method of multi-objective optimization is used to establish the objective function, and the multi-objective optimization objective function finds a compromise solution by weighing, specifically:
[0070]
[0071] In the formula, z i is the start-stop state of the i-th pump, z i = 0 indicates stop, z i = 1 indicates running, and α and β are weights reflecting the priority of the target. For example, α > β indicates that more attention is paid to economy, and vice versa, indicating that more attention is paid to environmental protection. Exemplarily, α is 1 and β is 10. i P is the power of the i-th pump, s λ s is the SO2 emission standard, λ 3 = 35 mg / m 3 .
[0072] In one or more embodiments, based on the resource load time series data set, a hybrid optimization algorithm based on random forest and enumeration method is used to screen the optimal solution of the optimization target of the multi-objective optimization model, including:
[0073] Step S501, a random forest regression prediction model is trained using a resource load time series data set. The random forest regression prediction model can mine the nonlinear mapping relationship between the start-stop state of the slurry circulating pump group and some important operating parameters (real-time power, inlet flue gas flow, original flue gas SO2 concentration, etc.) from historical data, and then predict the key performance indicators under different pump group configurations.
[0074] Step S502, the trained random forest regression prediction model is used to obtain the corresponding key indicators of various pump combinations.
[0075] Step S503, various pump combinations and corresponding key indicators are substituted into the multi-objective optimization model to obtain the target optimization value.
[0076] Step S504, the pump combination corresponding to the smallest target optimization value is taken as the optimal solution of the optimization target of the multi-objective optimization model.
[0077] Exemplarily, the enumeration method systematically traverses all possible circulating pump combinations. For each combination, firstly, its corresponding indicators are predicted by the random forest regression prediction model, and then it is evaluated based on the set objective function. The energy consumption and emissions are normalized by weighted summation, and the pump group configuration can be dynamically adjusted according to the actual emission requirements by modifying the weight coefficient.
[0078] The above optimization process ensures that the theoretically optimal solution, i.e., the circulating pump combination when the objective function is minimized, can be found through a global search strategy. Compared with traditional heuristic algorithms, the deterministic optimization method based on the enumeration method can avoid falling into local optimum.
[0079] As shown in Figure 2 Embodiment two of the present application also discloses a desulfurization system operation optimization device based on a random forest regression and enumeration hybrid algorithm. The device includes a preprocessing module 100, a cost construction module 200, a cost analysis module 300, an optimization model construction module 400, and an optimization solving module 500. The preprocessing module 100 is configured to preprocess the historical data of the desulfurization system to obtain a resource load time series data set. The cost construction module 200 is configured to construct a desulfurization cost model based on desulfurization material consumption cost and energy consumption cost. The cost analysis module 300 is configured to determine the optimization target according to the desulfurization cost model. The optimization model construction module 400 is configured to construct a multi-objective optimization model based on desulfurization cost and desulfurization efficiency. The optimization solving module 500 is configured to filter the optimal solution that meets the optimization target of the multi-objective optimization model based on the resource load time series data set using a hybrid optimization algorithm based on random forest and enumeration method. The detailed processing process of each module is described in Embodiment one and will not be repeated here.
[0080] In one or more embodiments, the computer-readable storage medium stores a desulfurization system operation optimization program based on a random forest regression and enumeration hybrid algorithm, which, when executed by a processor, implements the above-mentioned desulfurization system operation optimization method based on a random forest regression and enumeration hybrid algorithm.
[0081] In one or more embodiments, the computer device comprises a processor 12, an internal bus 13, a network interface 14, a computer readable storage medium 11. The processor 12 reads the corresponding computer program from the computer readable storage medium and then runs, and forms a request processing device at the logical level. Of course, in addition to the software implementation, one or more embodiments of the present specification do not exclude other implementations, such as logic devices or a combination of software and hardware, etc., that is, the execution subject of the following processing flow is not limited to each logical unit, but can also be hardware or logic devices. The computer readable storage medium 11 stores a desulfurization system operation optimization program based on a random forest regression and enumeration hybrid algorithm, and the desulfurization system operation optimization program based on the random forest regression and enumeration hybrid algorithm is executed by the processor to realize the above-mentioned desulfurization system operation optimization method based on the random forest regression and enumeration hybrid algorithm.
[0082] Wherein, the computer readable storage medium includes permanent and non-permanent, removable and non-removable media, which can realize information storage by any method or technology. The information can be computer readable instructions, data structures, program modules or other data. Examples of computer readable storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, disk storage, quantum memory, graphene-based storage medium or other magnetic storage device or any other non-transmission medium, which can be used to store information that can be accessed by a computing device.
[0083] As shown in the figure, after optimization by the desulfurization system operation optimization method based on the random forest regression and enumeration hybrid algorithm (hereinafter referred to as "the present scheme"), the energy consumption cost has a significant decline. Compared with the prior art, the present scheme has the following beneficial effects:
[0084] (1) The present scheme optimizes the operation of the desulfurization system of the thermal power plant, and due to the inaccuracy of the state prediction and the complexity of the massive historical data of the traditional control method, a random forest-enumeration hybrid algorithm is constructed. The random forest regression as a proxy model reduces the calculation cost of enumeration, can directly select the optimal configuration, and avoids local optimization. Based on the feature importance ranking method of the random forest, the input feature dimension is reduced, the training time is greatly shortened, and the real-time performance and energy efficiency of the desulfurization system pump group optimization are significantly improved.
[0085] (2) This scheme combines the predictive power of random forest regression with the global search characteristics of enumeration. It maintains the advantages of random forest in handling high-dimensional nonlinear problems, while avoiding the massive computation of traditional multi-objective optimization algorithms through the enumeration strategy, making it suitable for rapid optimization and decision-making in real-world scenarios. The advantages of this method are low computational cost, no need to calculate the Pareto front, faster prediction speed, and no need for additional algorithm design, only the number and depth of trees need to be adjusted.
[0086] (3) This scheme adopts a multi-objective optimization weighted summation method. The weighted summation method constructs a single objective function, simultaneously incorporating two objectives into the optimization search process. This ensures that the optimization algorithm can find a globally optimal compromise solution across the entire solution space, rather than a locally optimal or suboptimal solution. The final solution obtained is a systematic, globally optimal equilibrium point, with higher engineering practical value. The weighted summation method has extremely high flexibility. The weighting coefficients can be dynamically adjusted according to policy requirements, seasonal changes, or market electricity prices, quickly obtaining new optimal operating strategies.
[0087] The specific embodiments of the present invention have been described in detail above. Although some embodiments have been shown and described, those skilled in the art should understand that modifications and improvements can be made to these embodiments without departing from the principles and spirit of the present invention as defined by the claims and their equivalents, and such modifications and improvements should also be within the protection scope of the present invention.
Claims
1. A desulfurization system operation optimization method based on a random forest regression and enumeration hybrid algorithm, characterized in that, The method comprises: preprocessing historical data of a desulfurization system to obtain a resource load time series data set; constructing a desulfurization cost model based on desulfurization material consumption cost and energy consumption cost; determining an optimization target according to the desulfurization cost model analysis; constructing a multi-objective optimization model based on desulfurization cost and desulfurization efficiency; based on the resource load time series data set, using a hybrid optimization algorithm based on random forest and enumeration method to screen to obtain an optimal solution that meets the optimization target of the multi-objective optimization model.
2. The method for desulfurization system operation optimization based on random forest regression and enumeration hybrid algorithm according to claim 1, characterized in that, The historical data includes boiler load, desulfurization efficiency, raw flue gas flow, net flue gas flow, and inlet and outlet SO2 concentrations.
3. The method for desulfurization system operation optimization based on random forest regression and enumeration hybrid algorithm according to claim 2, characterized in that, The desulfurization cost model based on desulfurization material consumption cost and energy consumption cost comprises: constructing a material consumption cost model, which is used to analyze the cost of desulfurization agent limestone and determine the mapping relationship between unit SO2 removal and material consumption; constructing an energy consumption cost model, which is used to analyze the power consumption of main desulfurization equipment and determine the mapping relationship between unit SO2 removal and energy consumption.
4. The method for desulfurization system operation optimization based on random forest regression and enumeration hybrid algorithm according to claim 3, characterized in that, The material consumption cost model is specifically: wherein, Limestone cost, η is desulfurization efficiency, Q in Flue gas flow, Inlet SO2 concentration, Ca / S is calcium sulfur ratio, λ is absorbent purity, β is absorbent utilization rate, Limestone unit price, B is boiler load; The energy consumption cost model is specifically: In the formula, F dh is the electricity cost, P d is the price of auxiliary power, U i is the equipment operating voltage, I i is the equipment operating current, cosθ i is the power factor.
5. The method for desulfurization system operation optimization based on random forest regression and enumeration hybrid algorithm according to claim 1, characterized in that, The optimization target is a slurry circulating pump combination mode.
6. The method of claim 1, wherein the method is characterized by, The objective function of the multi-objective optimization model is: In the formula, z i is the start-stop state of the ith pump, z i = 0 indicates stop, z i = 1 indicates running, α and β are weights, P i is the power of the ith pump, is the net flue gas SO2 concentration, λ s is the SO2 emission standard.
7. The method of claim 1, wherein the method is characterized by, based on the resource load time series data set, using a hybrid optimization algorithm based on random forest and enumeration method to screen to obtain an optimal solution that meets the optimization target of the multi-objective optimization model, comprising: training a random forest regression prediction model using the resource load time series data set; using the trained random forest regression prediction model to predict the corresponding key indicators of various pump combinations; substituting various pump combinations and corresponding key indicators into the multi-objective optimization model to obtain target optimization values; the pump combination corresponding to the smallest target optimization value is used as the optimal solution that meets the optimization target of the multi-objective optimization model.
8. A desulfurization system operation optimization device based on a random forest regression and enumeration hybrid algorithm, characterized in that, The device comprises: a preprocessing module configured to preprocess historical data of a desulfurization system to obtain a resource load time series data set; a cost construction module configured to construct a desulfurization cost model based on desulfurization material consumption cost and energy consumption cost; a cost analysis module configured to determine an optimization target according to the desulfurization cost model analysis; an optimization model construction module configured to construct a multi-objective optimization model based on desulfurization cost and desulfurization efficiency; an optimization solving module configured to, based on the resource load time series data set, use a hybrid optimization algorithm based on random forest and enumeration method to screen to obtain an optimal solution that meets the optimization target of the multi-objective optimization model.
9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a desulfurization system operation optimization program based on a random forest regression and enumeration hybrid algorithm, and the desulfurization system operation optimization program based on the random forest regression and enumeration hybrid algorithm is executed by the processor to realize the desulfurization system operation optimization method based on the random forest regression and enumeration hybrid algorithm of any one of claims 1 to 7.
10. A computer device, comprising: The computer device comprises a computer readable storage medium, a processor and a desulfurization system operation optimization program based on a random forest regression and enumeration hybrid algorithm stored in the computer readable storage medium, and the desulfurization system operation optimization program based on the random forest regression and enumeration hybrid algorithm is executed by the processor to realize the desulfurization system operation optimization method based on the random forest regression and enumeration hybrid algorithm in any one of claims 1 to 7.