Safe and energy-saving organic waste gas purification system and control method

Abstract

The invention relates to the technical field of catalytic combustion control, and discloses a safe and energy-saving organic waste gas purification system and a control method. The server comprises a data acquisition module which is used for acquiring the current gas inlet flow and the current gas inlet temperature of corresponding organic waste gas in the catalytic combustion system in real time; the mechanism model calculation module is used for outputting first heating power required by the heater at the current moment based on the heating basic parameters and a heating mechanism model according to the current air inlet flow and the current air inlet temperature; the AI correction module is used for outputting first correction power corresponding to the heater at the current moment based on a pre-constructed AI power correction model according to the current air inlet flow, the current air inlet temperature, preset heating basic parameters and first heating power required by the heater at the current moment; and the control module is used for determining second heating power corresponding to the heater at the current moment according to the first heating power and the first correction power of the heater at the current moment.

Description

Technical Field

[0001] The present invention relates to the technical field of catalytic combustion control, and particularly relates to a safe and energy-saving organic waste gas purification system and a control method thereof. Background Art

[0002] In the process of treating organic waste gas by a catalytic combustion system, appropriately heating the organic waste gas is a key link to ensure efficient purification. The organic waste gas usually needs to be heated to a specific temperature to smoothly undergo a combustion reaction under the action of a catalyst, realizing the decomposition and transformation of harmful components. However, the existing organic waste gas heating technologies face many severe challenges.

[0003] The existing heating systems generally adopt traditional heating control or mechanism models for control;

[0004] Among them, it is difficult for the traditional heating system to adjust the heating power in real time and accurately according to these dynamic parameters. When the intake gas flow rate increases or the intake gas temperature decreases, if the heater power is insufficient, the waste gas cannot be heated to the appropriate reaction temperature, resulting in a significant decrease in the purification efficiency; conversely, when the intake gas flow rate decreases or the intake gas temperature increases, if the heater power fails to be adjusted downward in time, it will not only cause great waste of energy, but also may cause safety accidents such as equipment damage and fire due to local overheating.

[0005] And controlling the heating power through a mechanism model is difficult to adapt to complex and changeable actual working conditions. Relying solely on a fixed model to calculate the heating power cannot fully consider these complex factors, resulting in unstable heating effects and being difficult to meet the increasingly strict environmental protection and energy-saving requirements.

[0006] Based on this, there is an urgent need for a safe and energy-saving organic waste gas purification system and a control method thereof, which can solve the problems of low heating control accuracy and poor control timeliness in the heating process of the organic waste gas treated by the catalytic combustion system in the prior art, so as to realize the precision and high efficiency of the heating control of the organic waste gas. Summary of the Invention

[0007] One of the purposes of the present invention is to provide a safe and energy-saving organic waste gas purification system and a control method thereof, which can solve the problems of low heating control accuracy and poor control timeliness in the heating process of the organic waste gas treated by the catalytic combustion system in the prior art, so as to realize the precision and high efficiency of the heating control of the organic waste gas.

[0008] To achieve the above purpose, a safe and energy-saving organic waste gas purification system is provided, including a server;

[0009] The server includes:

[0010] A data acquisition module for real-time collection of the current intake flow rate and current intake temperature of the organic waste gas corresponding to the catalytic combustion system.

[0011] A mechanism model calculation module for outputting the first heating power required by the heater at the current moment based on the collected current intake flow rate and current intake temperature, and based on preset basic heating parameters and a pre-constructed heating mechanism model.

[0012] A control module for controlling the heater to operate at the first heating power according to the first heating power required by the heater at the current moment.

[0013] An AI correction module for obtaining the current intake flow rate, current intake temperature, and preset basic heating parameters corresponding at this time during the operation of the heater at the first heating power, and outputting the first correction power corresponding to the heater at this time based on a pre-constructed AI power correction model.

[0014] The control module is further configured to determine the second heating power corresponding to the heater at this time according to the first heating power and the first correction power, and control the heater to operate at the second heating power.

[0015] The technical principle and effect of this solution: In this solution, first, the current intake flow rate and current intake temperature of the organic waste gas in the catalytic combustion system are monitored in real time. These real-time data are the basic information sources for subsequent operations of the entire system. By continuously collecting the intake flow rate and temperature, the real-time state parameters of the organic waste gas can be accurately grasped.

[0016] Then, based on the collected current intake flow rate and temperature, combined with preset basic heating parameters (such as the specification parameters of the heater, the initial physical and chemical properties of the waste gas, etc.) and a pre-constructed heating mechanism model (this model may be based on principles such as heat transfer and chemical reaction kinetics, and describes the internal relationship between the intake flow rate, temperature, and required heating power), the first heating power required by the heater at the current moment is calculated. This step is to initially determine the heating power based on traditional physical and chemical principles and a preset parameter model.

[0017] The heater is controlled to operate according to the first heating power output by the mechanism model calculation module, so that the heater heats the organic waste gas at this power to meet the basic requirements of the catalytic combustion for the waste gas temperature.

[0018] After the heater operates at the first heating power, the current intake air flow rate, temperature, and preset basic heating parameters are obtained again. Based on the pre-constructed AI power correction model (which may be trained with a large amount of historical data and can take into account the influence of various complex factors on the heating power during actual operation, such as environmental temperature fluctuations and equipment aging, which are difficult to accurately describe with traditional mechanism models), the first corrected power is output. This step uses artificial intelligence technology to correct the power calculated based on the traditional mechanism model to adapt to the complex situations in actual operation.

[0019] Finally, the first heating power and the first corrected power are comprehensively considered to determine the second heating power that better meets the actual needs, and the heater is controlled to operate at this power. In this way, the dynamic and precise control of the heater power is achieved, making the heating process more adaptable to the real-time changes of organic waste gas.

[0020] In this solution, through the heating mechanism model, based on a deep analysis of the physical and chemical principles of the heating process, the complex relationship between the waste gas flow rate, temperature, and heating power is quantitatively presented through a mathematical model. Facing the dynamically changing intake air flow rate and temperature, it can give the power value that meets the basic heating requirements according to the established principles, initially solving the problem that the traditional system cannot accurately respond to the change of waste gas parameters, resulting in the heating power deviating from the actual needs and further affecting the purification efficiency. From the basic level, it ensures the accuracy of heating control and realizes the initial control of the heating power of the heater.

[0021] During the process of controlling the heater, the current intake air flow rate, temperature, and preset basic parameters are continuously obtained, and the first corrected power is output through the pre-constructed AI power correction model. It can analyze the actual operation data in real time and capture the subtle factors and non-linear relationships ignored by the traditional mechanism model under complex working conditions. When the waste gas parameters mutate or the complexity of the working conditions exceeds the preset range of the mechanism model, it can quickly adjust the power correction value and combine it with the first heating power to obtain the second heating power. This effectively solves the problem of poor timeliness of traditional technology control, enables the heating power to be dynamically and timely optimized according to the real-time state of the waste gas, realizes the precise and efficient heating control of organic waste gas, and also solves the problems of low heating control accuracy and poor control timeliness in the heating process of the organic waste gas processed by the catalytic combustion system in the existing technology, thus realizing the precision and efficiency of the heating control of organic waste gas.

[0022] Further, the server further includes:

[0023] A historical data acquisition module, which is used to collect the historical heating data corresponding to the organic waste gas of the catalytic combustion system and divide the collected historical heating data into a training set and a verification set;

[0024] The historical heating data includes historical intake air flow rate, historical intake air temperature, historical heating basic parameters, and the corresponding historical heating power;

[0025] A model construction module for constructing a corresponding AI power correction model;

[0026] A training module for using the historical intake air flow rate, historical intake air temperature, and historical heating basic parameters in the training set as input data and inputting them into the constructed AI power correction model to output the corresponding training prediction power. Based on the training prediction power, the historical heating power corresponding to the training set, and a preset model training strategy, the AI power correction model is trained, and the trained AI power correction model is output;

[0027] A verification module for inputting the historical intake air flow rate, historical intake air temperature, and historical heating basic parameters in the verification set into the trained AI power correction model to output the corresponding verification prediction power, and based on the verification prediction power and the historical heating power in the verification set, judging whether the corresponding trained AI power correction model is a feasible model according to a preset model verification strategy. If so, the corresponding AI power correction model is output; otherwise, the training module is executed again.

[0028] Beneficial effects: Abundant historical heating data is collected and divided into a training set and a verification set, providing a sufficient and scientific data basis for subsequent model training and verification. The training set data is used to repeatedly train the AI power correction model. By comparing the training prediction power and the historical heating power, the model parameters are continuously optimized according to the preset model training strategy. The verification module further uses the verification set data to test the trained model, and judges the feasibility of the model according to the verification prediction power and the historical heating power in the verification set according to the preset model verification strategy. This rigorous training and verification process can effectively improve the accuracy of the AI power correction model in predicting the heating power, enabling it to more accurately correct the heating power calculated by the mechanism model, thereby improving the performance of the entire purification system.

[0029] Further, the preset model training strategy is:

[0030] S10. According to the constructed AI power correction model, determine each hyperparameter corresponding to the model and the value range corresponding to each hyperparameter;

[0031] S20. According to the value range corresponding to each hyperparameter, randomly generate a number of hyperparameter groups to form AI power correction models corresponding to each hyperparameter group;

[0032] S30. Take the historical intake air flow rate, historical intake air temperature, and historical heating basic parameters in the training set as input data, and input them into the AI power correction models corresponding to each hyperparameter group to output the training prediction power corresponding to each hyperparameter group.

[0033] S40. Based on the training prediction power corresponding to each hyperparameter group and the historical heating power corresponding to the historical intake air flow rate, historical intake air temperature, and historical heating basic parameters in the training set, calculate the loss value corresponding to the AI power correction model of each hyperparameter group based on a preset loss value calculation formula.

[0034] The preset loss value calculation formula is:

[0035]

[0036] In the formula, L is the loss value corresponding to a certain hyperparameter group, P i is the historical heating power, is the training prediction power, w1 is the weighting coefficient corresponding to the i-th sample, λ1 and λ2 are regularization coefficients, θ j is the j-th parameter of the model, m is the number of model parameters, and n is the total number of samples.

[0037] S50. According to the loss values corresponding to each hyperparameter group, select the hyperparameter groups with loss values less than or equal to the preset first loss threshold to form a corresponding first hyperparameter group set.

[0038] S60. According to each hyperparameter group in the first hyperparameter group set, randomly set a random number between 0 and 1 for each hyperparameter in each hyperparameter group, and change the random numbers at a preset change frequency within a preset change time period. Determine whether the random number corresponding to the current moment is greater than the preset random threshold. If not, the hyperparameter corresponding to the random data is not replaced. If so, label the hyperparameter corresponding to the random number at the current moment with a first label, and determine whether there are also hyperparameters of the same type labeled with the first label at the current moment. If so, randomly exchange the hyperparameters of the same type labeled with the first label at the current moment to form the hyperparameter groups corresponding to the current moment. Until after the preset change time, output the updated second hyperparameter group set and restart S30 until the preset number of iterations is reached, and output the final corresponding hyperparameter groups.

[0039] S70. According to the finally corresponding hyperparameter groups output, randomly select a hyperparameter group and output the AI power correction model corresponding to the hyperparameter group as the trained AI power correction model.

[0040] Beneficial effects: Steps S10 to S20 can comprehensively explore the hyperparameter space by determining the hyperparameters of the AI power correction model and their value ranges, and randomly generating multiple hyperparameter groups. Different hyperparameter combinations will affect the learning ability and generalization performance of the model. This extensive exploration helps to find a better hyperparameter configuration, enabling the model to better fit the training data, thereby improving the accuracy of heating power prediction and enhancing the performance of the purification system.

[0041] Steps S30 to S50 calculate the loss value of the model corresponding to each hyperparameter group using a preset loss value calculation formula, and filter out the hyperparameter groups whose loss values are less than or equal to a preset first loss threshold to form a first hyperparameter group set. This process can effectively evaluate the fitting degree of the model to the training data and retain the hyperparameter groups with better performance. The loss value comprehensively considers the error between the predicted power and the historical heating power, as well as the regularization of the model parameters, which can prevent the model from overfitting, ensure that the model not only performs well on the training set but also has high accuracy and reliability in actual applications, providing a more accurate power correction basis for the purification system.

[0042] Step S60 provides an opportunity for the model to jump out of the local optimum by randomly adjusting the hyperparameters in the first hyperparameter group set. Within a preset change time period, the random number is changed at a preset frequency. When the random number is greater than the preset random threshold, the corresponding hyperparameter is operated on. This is like a "random jump" in the parameter space of the model. For example, if the model is hovering around a local optimum solution, by changing the values of the hyperparameters, the training direction of the model will change, and it may explore new and better parameter regions, thereby improving the model performance.

[0043] When it is determined that there are hyperparameters of the same type that are all labeled with the first label at the current moment, these hyperparameters are randomly exchanged. This operation can explore the synergistic effect between different hyperparameters. Different hyperparameters often interact with each other in the model. By randomly exchanging hyperparameters of the same type, different parameter combination methods can be tried to find the hyperparameter combinations that can cooperate with each other and improve the model performance.

[0044] As time goes by, the hyperparameters are continuously adjusted according to a preset rule, and the model can continuously learn and adapt to different training situations. Each adjustment is an exploration of the model's parameter space. As the number of iterations increases, the model gradually converges to a better parameter configuration. This continuous optimization process enables the model to better adapt to the complex and changeable working conditions in the organic waste gas treatment process, such as the irregular fluctuations of waste gas flow and temperature. It enables the model to encounter more diverse hyperparameter combinations during the training process, which helps to enhance the generalization ability of the model. The model is no longer limited to specific hyperparameter settings but learns to learn and predict under different parameter conditions.

[0045] Further, the preset change time period and change frequency in S60 are dynamically adjusted based on a preset dynamic change strategy;

[0046] The preset dynamic change strategy is as follows:

[0047] When the current iteration number is less than the first iteration threshold, the first time period is used as the corresponding preset change time period, and the corresponding change frequency is the first frequency;

[0048] When the current iteration number is greater than or equal to the first iteration threshold and less than the second iteration threshold, based on a preset time period and frequency change calculation formula, calculate the preset change time period and change frequency corresponding to the current iteration;

[0049] The preset time period and frequency change calculation formula is:

[0050]

[0051] In the formula, T(n) is the preset change time period corresponding to the current iteration, n is the current iteration number, n1 is the first iteration threshold, n2 is the second iteration threshold, T1 is the first time period, and f1 is the first frequency;

[0052] When the current iteration number is greater than the second iteration threshold, the second time period is used as the corresponding preset change time period, and the corresponding change frequency is the second frequency; where the first time period is greater than the second time period, and the first frequency is greater than the second frequency.

[0053] Beneficial effects: In the initial stage of model training, that is, when the current iteration number is less than the first iteration threshold, a longer first time period and a higher first frequency are adopted. The longer time period provides a time guarantee for the hyperparameters to fully explore the parameter space, and the higher frequency enables the model to try different hyperparameter combinations faster, so as to quickly find potential better parameter regions in the early stage of training and accelerate model convergence.

[0054] When the iteration number is between the first iteration threshold and the second iteration threshold, it is dynamically adjusted through a preset time period and frequency change calculation formula. As the iteration number increases, the preset change time period gradually shortens, and the change frequency gradually decreases. This is because as the training progresses, the model has initially explored some better parameter regions. At this time, it is necessary to reduce the drastic changes of the hyperparameters and focus more on fine-tuning within the explored regions to achieve precise convergence of the model.

[0055] When the number of iterations is greater than the second iteration threshold, a shorter second time period and a lower second frequency are adopted. At this time, the model is already close to the convergence state. The shorter time period and lower frequency can prevent the over-adjustment of hyperparameters from damaging the already optimized model structure, ensuring that the model continues to be fine-tuned under a stable parameter configuration to further improve the model performance.

[0056] Furthermore, the preset model verification strategy is as follows:

[0057] According to the verification prediction power and the historical heating power in the verification set, based on the preset verification error value calculation formula, calculate the verification error value corresponding to the trained AI power correction model;

[0058] The preset verification error value calculation formula is:

[0059]

[0060] In the formula, E is the verification error value corresponding to the trained AI power correction model, k is the total amount of data corresponding to the verification set, P1 j is the j-th historical heating power in the verification set, is the verification prediction power corresponding to the j-th historical heating power in the verification set, σ is the data standard deviation of the verification prediction power corresponding to each historical heating power in the verification set, N is the number of model parameters corresponding to the model; α is the corresponding weighting coefficient;

[0061] Judge whether the verification error value corresponding to the corresponding trained AI power correction model is less than the preset error threshold. If so, output the corresponding AI power correction model; otherwise, re-execute the training module.

[0062] Beneficial effects: The preset verification error value calculation formula comprehensively considers the deviation between the model prediction result and the actual historical heating power. The first term in the formula reflects the mean square of the direct error between the prediction power and the true power, intuitively reflecting the accuracy of the model prediction; the second term combines the number of model parameters and the data standard deviation of the prediction power, considering the complexity of the model and the dispersion degree of the prediction results. The weighting coefficient can flexibly adjust the proportion of these two parts of factors in the evaluation, so that the verification error value can accurately reflect the performance of the model in different application scenarios.

[0063] The present invention also provides a safe and energy-saving organic waste gas purification control method, using the above-mentioned safe and energy-saving organic waste gas purification system. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is the logic block diagram of the safe and energy-saving organic waste gas purification system in Embodiment 1 of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0065] The following is a further detailed description through specific embodiments:

[0066] Embodiment 1

[0067] A safe and energy-saving organic waste gas purification system is basically as Figure 1 shown, and the server includes:

[0068] A data acquisition module for real-time collecting the current intake air flow rate and the current intake air temperature corresponding to the organic waste gas in the catalytic combustion system;

[0069] A mechanism model calculation module for outputting the first heating power required by the heater at the current moment based on the collected current intake air flow rate and the current intake air temperature, and based on the preset basic heating parameters and the pre-constructed heating mechanism model; in this embodiment, the basic heating parameters include the preset outlet temperature, heating efficiency, and specific heat capacity of the gas. The heating mechanism model is constructed based on the law of conservation of energy and the principle of heat transfer.

[0070] A control module for controlling the heater to work according to the first heating power based on the first heating power required by the heater at the current moment;

[0071] An AI correction module for obtaining the current intake air flow rate, the current intake air temperature, and the preset basic heating parameters corresponding thereto during the process of the heater working according to the first heating power, and outputting the first correction power corresponding to the heater at this time based on the pre-constructed AI power correction model; in this embodiment, the pre-constructed AI power correction model is constructed by using the existing deep neural network technology.

[0072] The control module is further used for determining the second heating power corresponding to the heater at this time according to the first heating power and the first correction power, and controlling the heater to work according to the second heating power.

[0073] The server further includes:

[0074] A historical data acquisition module for collecting the historical heating data corresponding to the organic waste gas in the catalytic combustion system, and dividing the collected historical heating data into a training set and a validation set;

[0075] The historical heating data includes historical intake air flow rate, historical intake air temperature, historical basic heating parameters, and the corresponding historical heating power;

[0076] A model construction module for constructing the corresponding AI power correction model;

[0077] A training module, which is used to input historical intake air flow, historical intake air temperature, and historical heating basic parameters in a training set as input data into a constructed AI power correction model, output corresponding training prediction power, and based on the training prediction power, historical heating power corresponding to the training set, and a preset model training strategy, train the AI power correction model and output the trained AI power correction model;

[0078] The preset model training strategy is as follows:

[0079] S10. According to the constructed AI power correction model, determine each hyperparameter corresponding to the model and the value range corresponding to each hyperparameter;

[0080] S20. According to the value range corresponding to each hyperparameter, randomly generate several hyperparameter groups to form AI power correction models corresponding to each hyperparameter group;

[0081] S30. Input historical intake air flow, historical intake air temperature, and historical heating basic parameters in the training set as input data into the AI power correction models corresponding to each hyperparameter group, and output the training prediction power corresponding to each hyperparameter group;

[0082] S40. According to the training prediction power corresponding to each hyperparameter group, and the historical heating power corresponding to historical intake air flow, historical intake air temperature, and historical heating basic parameters in the training set, based on a preset loss value calculation formula, calculate the loss value corresponding to the AI power correction model corresponding to each hyperparameter group;

[0083] The preset loss value calculation formula is as follows:

[0084]

[0085] In the formula, L is the loss value corresponding to a certain hyperparameter group, P i is the historical heating power, is the training prediction power, w1 is the weighting coefficient corresponding to the i-th sample, λ1 and λ2 are regularization coefficients, θ j is the j-th parameter of the model, m is the number of model parameters, and n is the total number of samples;

[0086] S50. According to the loss values corresponding to each hyperparameter group, select the hyperparameter groups with loss values less than or equal to a preset first loss threshold to form a corresponding first hyperparameter group set;

[0087] S60. According to each hyperparameter group in the first set of hyperparameters, randomly set a random number between 0 and 1 for each hyperparameter in each hyperparameter group, and change each random number according to a preset change frequency within a preset change time period; determine whether the random number corresponding to the current moment is greater than a preset random threshold. If not, the hyperparameter corresponding to the random data is not replaced; if so, the hyperparameter corresponding to the random number at the current moment is labeled with a first label, and determine whether there are also hyperparameters of the same type labeled with the first label at the current moment. If so, randomly exchange the hyperparameters of the same type labeled with the first transformer at the current moment, and form each hyperparameter group corresponding to the current moment. Until the preset change time is reached, output the updated second set of hyperparameter groups, and repeat S30 until the preset number of iterations is reached, and output each corresponding hyperparameter group finally; for example, let the first set of hyperparameter groups be where H i represents the i-th hyperparameter group, (i = 1, 2,..., n).

[0088] Each hyperparameter group (H i ) contains m hyperparameters of the same type, that is, (H i = {h i1 , h i2 , …, h im}), where (h ij ) represents the j-th hyperparameter in the i-th hyperparameter group, (j = 1, 2,..., m).

[0089] Let (r ij (t)) represent the random number corresponding to the j-th hyperparameter in the i-th hyperparameter group at time t, (t = 0, 1, 2,..., T), and T is the total duration of the preset change time period.

[0090] Let (θ) be the preset random threshold.

[0091] Let (l ij (t)) be a label variable. When (r ij (t)>θ), (l ij (t) = 1) (indicating being labeled with the first label), otherwise (l ij (t) = 0). If at a certain sampling moment, only one hyperparameter in three hyperparameter groups is labeled with the first label, which are h i2 , h k2 , h q2 respectively, then when performing random replacement, it is possible that h i2 and h k2 perform value replacement of the corresponding hyperparameters, and then h k2 and h q2 perform value replacement of the corresponding hyperparameters, hi2 With h q2 No swapping is performed. At this time, the corresponding hyperparameter groups are respectively:

[0092] H i ={h i1 , h k2 , …, h im})

[0093] H1 k ={h k1 , h i2 , …, h km})

[0094] H2 k ={h k1 , h q2 , …, h km}}

[0095] H q ={h q1 , h k2 , …, h qm})

[0096] The preset change time period and change frequency in S60 are dynamically adjusted based on a preset dynamic change strategy;

[0097] The preset dynamic change strategy is:

[0098] When the current iteration number is less than the first iteration threshold, the first time period is used as the corresponding preset change time period, and the corresponding change frequency is the first frequency;

[0099] When the current iteration number is greater than or equal to the first iteration threshold and less than the second iteration threshold, based on a preset time period and frequency change calculation formula, calculate the preset change time period and change frequency corresponding to this iteration;

[0100] The preset time period and frequency change calculation formula is:

[0101]

[0102] In the formula, T(n) is the preset change time period corresponding to this iteration, n is the current iteration number, n1 is the first iteration threshold, n2 is the second iteration threshold, T1 is the first time period, and f1 is the first frequency;

[0103] When the current iteration number is greater than the second iteration threshold, the second time period is used as the corresponding preset change time period, and the corresponding change frequency is the second frequency; where the first time period is greater than the second time period, and the first frequency is greater than the second frequency.

[0104] S70. Randomly select a hyperparameter group according to the final corresponding hyperparameter group in the output, and output the AI power correction model corresponding to this hyperparameter group as the trained AI power correction model.

[0105] A verification module is used to input the historical intake air flow rate, historical intake air temperature, and historical heating basic parameters in the verification set into the trained AI power correction model, output the corresponding verification predicted power, and based on the preset model verification strategy, judge whether the corresponding trained AI power correction model is a feasible model according to the verification predicted power and the historical heating power in the verification set. If so, output the corresponding AI power correction model; otherwise, re - execute the training module.

[0106] The preset model verification strategy is as follows:

[0107] According to the verification predicted power and the historical heating power in the verification set, calculate the verification error value corresponding to the trained AI power correction model based on the preset verification error value calculation formula.

[0108] The preset verification error value calculation formula is:

[0109]

[0110] In the formula, E is the verification error value corresponding to the trained AI power correction model, k is the total amount of data corresponding in the verification set, P1 j is the j - th historical heating power in the verification set, is the verification predicted power corresponding to the j - th historical heating power in the verification set, σ is the data standard deviation of the verification predicted powers corresponding to each historical heating power in the verification set, N is the number of model parameters corresponding to the model; α is the corresponding weighting coefficient;

[0111] Judge whether the verification error value corresponding to the corresponding trained AI power correction model is less than the preset error threshold. If so, output the corresponding AI power correction model; otherwise, re - execute the training module.

[0112] This embodiment also discloses a safe and energy - saving organic waste gas purification control method, which uses the above - mentioned safe and energy - saving organic waste gas purification system.

[0113] The above are only embodiments of the present invention. Specific structures and characteristics and other common knowledge in the art are not described in detail here. Those of ordinary skill in the art know all the common general technical knowledge in the technical field to which the invention pertains before the filing date or the priority date, can learn all the prior arts in this field, and have the ability to apply the conventional experimental means before this date. Those of ordinary skill in the art can, under the inspiration given in this application, improve and implement this solution in combination with their own abilities. Some typical well-known structures or well-known methods should not become an obstacle for those of ordinary skill in the art to implement this application. It should be noted that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can also be made, which should also be regarded as the protection scope of the present invention, and these will not affect the implementation effect of the present invention and the practicality of the patent. The protection scope claimed in this application should be based on the content of its claims, and the specific implementation manners and other records in the specification can be used to interpret the content of the claims.

Claims

1. A safe and energy-saving organic waste gas purification system, characterized by: Including the server side; The server includes: The data acquisition module is used to collect the current intake air flow rate and current intake air temperature corresponding to the organic waste gas in the catalytic combustion system in real time; The mechanism model calculation module is used to output the first heating power required by the heater at the current moment according to the collected current intake air flow rate, the current intake air temperature, the preset basic heating parameters and the pre-built heating mechanism model; A control module, used for controlling the heater to work according to the first heating power required by the heater at a current moment; The AI ​​correction module is used to obtain the current intake air flow rate, the current intake air temperature, and the preset basic heating parameters corresponding to the heater when the heater is working at the first heating power, and output the first corrected power corresponding to the heater at this time based on the pre-built AI power correction model; The control module is further used to determine the second heating power corresponding to the heater at this time according to the first heating power and the first correction power, and control the heater to work according to the second heating power.

2. A safe and energy-saving organic waste gas purification system according to claim 1, characterized in that: The server also includes: A historical data acquisition module is used to collect historical heating data corresponding to the organic waste gas corresponding to the catalytic combustion system, and divide the collected historical heating data into a training set and a verification set; The historical heating data includes historical intake air flow, historical intake air temperature, historical basic heating parameters and corresponding historical heating power; A model building module, used to build the corresponding AI power correction model; A training module, which is used to input the historical intake flow, historical intake temperature and historical heating basic parameters in the training set as input data into the constructed AI power correction model, output the corresponding training predicted power, train the AI ​​power correction model based on the training predicted power, the historical heating power corresponding to the training set and the preset model training strategy, and output the trained AI power correction model; The verification module is used to input the historical intake flow, historical intake temperature and historical heating basic parameters in the verification set into the trained AI power correction model, output the corresponding verification predicted power, and based on the verification predicted power and the historical heating power in the verification set, based on the preset model verification strategy, determine whether the corresponding trained AI power correction model is a feasible model. If so, output the corresponding AI power correction model, otherwise, re-execute the training module.

3. A safe and energy-saving organic waste gas purification system according to claim 2, characterized in that: The preset model training strategy is: S10. According to the constructed AI power correction model, determine each hyperparameter corresponding to the model and the value range corresponding to each hyperparameter; S20. According to the value ranges corresponding to the respective hyperparameters, a number of hyperparameter groups are randomly generated to form an AI power correction model corresponding to each hyperparameter group. S30, taking the historical intake flow, historical intake temperature and historical heating basic parameters in the training set as input data, inputting them into the AI ​​power correction model corresponding to each hyperparameter group, and outputting the training predicted power corresponding to each hyperparameter group; S40, according to the training predicted power corresponding to each hyperparameter group, and the historical heating power corresponding to the historical intake flow, historical intake temperature, and historical heating basic parameters in the training set, based on a preset loss value calculation formula, calculate the loss value corresponding to the AI ​​power correction model corresponding to each hyperparameter group; The preset loss value calculation formula is: In the formula, L is the loss value corresponding to a certain hyperparameter group, P i is the historical heating power, is the training prediction power, w1 is the weight coefficient corresponding to the i-th sample, λ1, λ2 are regularization coefficients, θ j is the jth parameter of the model, m is the number of parameters of the model, and n is the total number of samples; S50, according to the loss values ​​corresponding to the respective hyperparameter groups, selecting a hyperparameter group whose loss value is less than or equal to a preset first loss threshold to form a corresponding first hyperparameter group set; S60, according to each hyperparameter group in the first hyperparameter group set, randomly set a random number from 0 to 1 for each hyperparameter in each hyperparameter group, and change each random number according to a preset change frequency within a preset change time period; determine whether the random number corresponding to the current moment is greater than a preset random threshold, if not, the hyperparameter corresponding to the random data is not changed; if so, a first label is attached to the hyperparameter corresponding to the random number corresponding to the current moment, and determine whether there is a hyperparameter of the same type also attached to the first label at the current moment, if so, randomly exchange the hyperparameters of the same type attached to the first label at the current moment, and form each hyperparameter group corresponding to the current moment, until the preset change time is reached, output the updated second hyperparameter group set, and repeat S30 until the preset number of iterations is reached, and output the corresponding hyperparameter groups in the end; S70. Randomly select a hyperparameter group according to the final corresponding hyperparameter group output, and output the AI ​​power correction model corresponding to the hyperparameter group as the trained AI power correction model.

4. A safe and energy-saving organic waste gas purification system according to claim 3, characterized in that: The preset change time period and change frequency in S60 are dynamically adjusted based on a preset dynamic change strategy; The preset dynamic change strategy is: When the number of iterations is less than the first iteration threshold, the first time period is used as the corresponding preset change time period, and the corresponding change frequency is the first frequency; When the number of iterations is greater than or equal to the first iteration threshold and less than the second iteration threshold, the preset change time period and change frequency corresponding to the iteration are calculated based on the preset time period and frequency change calculation formula; The preset time period and frequency change calculation formula are as follows: Wherein, T(n) is the preset change time period corresponding to this iteration, n is the number of iterations, n1 is the first iteration threshold, n2 is the second iteration threshold, T1 is the first time period, and f1 is the first frequency; When the number of iterations is greater than the second iteration threshold, the second time period is used as the corresponding preset change time period, and the corresponding change frequency is the second frequency; wherein the first time period is greater than the second time period, and the first frequency is greater than the second frequency.

5. A safe and energy-saving organic waste gas purification system according to claim 4, characterized in that: The preset model verification strategy is: According to the verification prediction power and the historical heating power in the verification set, the verification error value corresponding to the trained AI power correction model is calculated based on the preset verification error value calculation formula; The preset verification error value calculation formula is: Where E is the verification error value corresponding to the trained AI power correction model, k is the total amount of data corresponding to the verification set, and P1 j is the j-th historical heating power in the validation set, is the verification prediction power corresponding to the j-th historical heating power in the verification set, σ is the data standard deviation of the verification prediction power corresponding to each historical heating power in the verification set, N is the number of model parameters corresponding to the model; α is the corresponding weighting coefficient; Determine whether the verification error value corresponding to the trained AI power correction model is less than the preset error threshold. If so, output the corresponding AI power correction model; otherwise, re-execute the training module.

6. A safe and energy-saving organic waste gas purification control method, characterized in that: A safe and energy-saving organic waste gas purification system using any one of claims 1 to 5 above.

Images