CN117726257A - Industrial park hazardous waste treatment method based on artificial intelligence big data model - Google Patents
Industrial park hazardous waste treatment method based on artificial intelligence big data model Download PDFInfo
- Publication number
- CN117726257A CN117726257A CN202311759161.4A CN202311759161A CN117726257A CN 117726257 A CN117726257 A CN 117726257A CN 202311759161 A CN202311759161 A CN 202311759161A CN 117726257 A CN117726257 A CN 117726257A
- Authority
- CN
- China
- Prior art keywords
- storage bin
- dangerous
- waste
- storage
- hazardous waste
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 239000002920 hazardous waste Substances 0.000 title claims abstract description 52
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000013499 data model Methods 0.000 title claims abstract description 17
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 16
- 239000002699 waste material Substances 0.000 claims abstract description 110
- 238000004519 manufacturing process Methods 0.000 claims abstract description 31
- 238000004458 analytical method Methods 0.000 claims abstract description 26
- 238000004806 packaging method and process Methods 0.000 claims abstract description 19
- 230000002159 abnormal effect Effects 0.000 claims abstract description 12
- 238000012549 training Methods 0.000 claims description 48
- 238000003066 decision tree Methods 0.000 claims description 28
- 238000002372 labelling Methods 0.000 claims description 26
- 238000003062 neural network model Methods 0.000 claims description 18
- 230000005856 abnormality Effects 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 12
- 238000010276 construction Methods 0.000 claims description 10
- 238000007637 random forest analysis Methods 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 3
- 239000013064 chemical raw material Substances 0.000 claims description 3
- 230000003203 everyday effect Effects 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 3
- 239000000575 pesticide Substances 0.000 claims description 3
- 239000003208 petroleum Substances 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 239000003171 wood protecting agent Substances 0.000 claims description 3
- 238000007726 management method Methods 0.000 abstract description 8
- 238000012544 monitoring process Methods 0.000 abstract description 4
- 230000008901 benefit Effects 0.000 description 6
- 238000013459 approach Methods 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 231100001261 hazardous Toxicity 0.000 description 2
- 238000007789 sealing Methods 0.000 description 2
- 239000003905 agrochemical Substances 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000002906 medical waste Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 239000002994 raw material Substances 0.000 description 1
- 239000002910 solid waste Substances 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Landscapes
- Processing Of Solid Wastes (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of hazardous waste treatment, in particular to an industrial park hazardous waste treatment method based on an artificial intelligence big data model. According to the invention, the management party of the industrial park is taken as an operation main body, and no matter how many hazardous waste manufacturers exist in the industrial park, the whole-flow management and risk monitoring can be realized; aiming at the storage flow of dangerous wastes which can be mistakenly made manually, judging the production condition of a source manufacturer and judging whether the false packaging is likely to occur or not through an abnormal analysis model; during storage, whether the storage bin leaks or not can be judged, so that accident risk is remarkably reduced; aiming at the dangerous waste transportation flow, various characteristics of dangerous waste and a storage bin are integrated, the transportation priority is guided in real time, and the accident risk is further reduced on the basis of saving the transportation cost.
Description
Technical Field
The invention relates to the technical field of hazardous waste treatment, in particular to an industrial park hazardous waste treatment method based on an artificial intelligence big data model.
Background
Hazardous waste refers to waste with hazardous characteristics listed in the national hazardous waste list, including medical waste, waste pharmaceutical, agricultural chemical waste, etc., all of which need to be subjected to strict treatment.
According to laws and regulations such as the national control method for solid waste pollution environment of the people's republic of China, environmental protection conditions of Chongqing city, technical standards for dangerous waste collection, storage and transportation, the method for managing the transfer of dangerous waste, the method for managing the management of dangerous waste management license and the like, the whole process of dangerous waste production, storage, transportation and treatment is required to be subjected to standardized management, and meanwhile, the whole process of dangerous waste is also required to be monitored, so that immeasurable loss and casualties caused by the fact that the dangerous waste is not managed in place are prevented.
The current management flow is relatively perfect, and the full flow tracking of dangerous wastes is realized mostly by various automatic monitoring and automatic statistics means, but the method is insufficient in that: 1. the personnel participation is not lacked in the whole flow, and the possibility of errors, such as misencapsulation, delayed transportation, storage leakage, abnormal production of factories and the like, cannot be effectively identified by the current supervision means, and has a large risk; 2. the hazardous waste is more, and every mill has independent supervision software and hardware, can't carry out effectual supervision to all factories in the garden to large-scale industrial garden.
Disclosure of Invention
The invention discloses an industrial park hazardous waste treatment method based on an artificial intelligence big data model, which takes an industrial park as a main body, can independently complete the omnibearing supervision of hazardous waste aiming at different factories in the industrial park, and remarkably improves the treatment safety of the hazardous waste.
The technical scheme is that a plurality of storage bins for storing dangerous wastes of different categories are built in the industrial park, and the dangerous waste treatment method comprises the following steps:
labeling the type, packaging date, source manufacturer, weight and volume information of the dangerous waste aiming at the packaged dangerous waste;
storing the marked hazardous wastes in storage bins corresponding to various types;
transporting the hazardous waste in each storage bin to a corresponding disposal place by standardization;
judging the production condition of a source manufacturer through an anomaly analysis model when dangerous wastes enter a storage bin;
judging whether the dangerous waste is packaged by mistake when the dangerous waste enters a storage bin;
corresponding sensors for detecting leakage of different dangerous wastes are arranged in the storage bin, and data acquired by the sensors judge whether leakage occurs in the storage bin or not through a leakage prediction model;
and predicting the future storage quantity of each storage bin through a storage quantity prediction model, calculating the transportation demand value of each storage bin, and arranging a transportation vehicle to sequentially transport dangerous wastes in different storage bins according to the transportation demand value.
The method has the advantages that the management party of the industrial park is taken as an operation main body, respective monitoring models are respectively established for different storage bins, and no matter how many hazardous waste manufacturers exist in the industrial park, the whole-flow management and risk monitoring can be realized; aiming at the storage flow of dangerous wastes which can be mistakenly made manually, judging the production condition of a source manufacturer and judging whether the false packaging is likely to occur or not through an abnormal analysis model; during storage, whether the storage bin leaks or not can be judged, so that accident risk is remarkably reduced; aiming at the dangerous waste transportation flow, various characteristics of dangerous waste and a storage bin are integrated, the transportation priority is guided in real time, and the accident risk is further reduced on the basis of saving the transportation cost.
Optionally, the hazardous waste comprises any one of the following and more:
pharmaceutical manufacturing waste, pesticide manufacturing waste, wood preservative manufacturing waste, refined petroleum waste, gas production waste, and base chemical raw material manufacturing waste.
Further, the production condition of the source manufacturer is judged through an anomaly analysis model, and the specific method is as follows:
according to source manufacturers of dangerous wastes stored in the storage bin, selecting an abnormality analysis model of the corresponding manufacturer;
the day of the abnormal analysis is called, and the corresponding manufacturer stores the types of the dangerous wastes and the corresponding weights of each type in various dangerous waste storage bins respectively;
inputting various dangerous waste types and corresponding weights of each type into an abnormality analysis model;
the abnormality analysis model outputs the day of the abnormality analysis, and whether the manufacturer produces the abnormality exists.
Further, the anomaly analysis model is a random forest model, and the construction method comprises the following steps:
for the total training set T, N samples are shared, N samples are randomly selected each time with a replacement, each sample comprises the same random feature, the number k of the random features is smaller than the total feature number d, and therefore the N samples construct tree training samples of each decision tree;
when each tree training sample has M attributes, randomly selecting M attributes from the M attributes when each node of the decision tree needs to be split, satisfying the condition M < M, and selecting 1 split attribute belonging to the node from the M attributes by adopting information gain g (D, A) =H (D) -H (D|A) as a basis;
until each decision tree is no longer split;
the voting result of each decision tree is synthesized, and the category with the largest final voting result is finally output by the random forest model;
in the random forest model construction process, a total training set T stores the weight of each dangerous waste in a factory history and whether production marked by people is abnormal or not; n samples are weights of each dangerous waste stored by manufacturers in the historical N days every day, and the random characteristic k is the weight of each dangerous waste;
therefore, the tree training sample of each decision tree is obtained by randomly and replaceably obtaining N times of historical single-day data in the total training set T, and each time of obtained historical single-day data comprises random and same dangerous waste types and weights.
The method has the advantages that an abnormality analysis model is independently customized for each manufacturer in the park, whether the production condition of the manufacturer is abnormal or not can be detected from the delivery quantity of dangerous waste, and the control force of the park supervisor on the park is further improved; the random forest algorithm can be used as an abnormal classification algorithm to fuse big data characteristics, has the advantages of extremely high accuracy, difficult fitting, strong noise resistance, capability of processing high-dimensional data and the like, and is optimal for different special conditions of various manufacturers.
Further, when dangerous waste is put into the storage bin, judging whether the packaging is wrong or not, wherein the specific method is as follows:
reading type information marked by the packaged hazardous waste, and judging that the packaging is in error packaging or error storage if the type information marked by the packaged hazardous waste is different from the storage type information stored in the storage bin;
reading the labeling weight and the labeling volume information of the labeling of the packaged hazardous waste, and calculating the theoretical density according to the labeling weight and the labeling volume;
actual weight of the packaged hazardous waste is measured, and reference density is calculated according to the actual weight and the marked volume;
and taking the difference value between the marked weight and the actual weight, the marked volume and the difference value between the theoretical density and the reference density as inputs, and outputting whether the error package exists or not through a decision tree.
Further, the decision tree construction method comprises the following steps:
acquiring a plurality of groups of historical data, wherein the historical data comprise the difference value between the historical labeling weight and the actual weight, the historical labeling volume and the difference value between the historical theoretical density and the reference density, and labeling whether each group of historical data is packaged by mistake or not;
calculating the information entropy of the root node, H (D) = - Σp (i). Log 2 p (i), where p (i) is the probability of mis-packaging or non-mis-packaging;
calculating information gain of each attribute, g (D, a) =h (D) -H (d|a);
selecting the attribute with the maximum information gain as a dividing attribute;
and repeatedly calculating the information gain and selecting the maximum information gain attribute until the decision tree cannot be split any more.
The embodiment has the advantages that although the dangerous waste is the same in various types, the density of the dangerous waste can be changed correspondingly along with the change of the process, the raw materials and the like, so that compared with the threshold judgment which is easy to think, the decision tree algorithm can be adopted to remarkably improve the accuracy of the error packaging judgment, and is relatively simple, and the calculation force can be saved.
Further, judging whether leakage occurs in the storage bin or not through a leakage prediction model, wherein the concrete method comprises the following steps of:
setting one or more gas sensors corresponding to possible leaked gas according to different storage bins and dangerous waste types;
corresponding dangerous wastes with different weights are respectively stored in the storage bin, the storage bin door is closed, and the increment of leakage gas during closing of the bin door is obtained through the gas sensor after preset time;
training data are constructed for different weights and the leakage gas increment corresponding to the different weights;
training a neural network model with training data;
before the storage bin is closed each time, the actual weight of dangerous waste in the storage bin is obtained, leaked gas in the storage bin is emptied, after the preset time, the increment of the leaked gas obtained by the gas sensor and the actual weight of the dangerous waste are input into the neural network model, and the judgment of whether the leakage occurs in the storage bin is output.
The embodiment has the advantages that the leakage prediction model discharges leakage gas as much as possible before the sealing of the bin for the same sealing time, and the prediction accuracy can be remarkably improved.
Further, the future storage quantity of each storage bin is predicted by a storage quantity prediction model, and the specific method is as follows:
acquiring historical storage data of each storage bin, and taking the historical storage data as training data;
training a neural network model with training data;
and predicting future storage quantity of each storage bin by using a neural network model.
Further, the training method of the neural network model is as follows:
initializing an input layer, a hidden layer and an output layer;
initializing neuron weights and bias values of an input layer, a hidden layer and an output layer;
defining a learning rate, training cycle number and test cycle number;
setting an activation function as a sigmoid function and a loss function as an MSE mean square error function;
calculating a current loss by forward propagation;
calculating an update weight by back propagation;
until the neural network model converges, or the training times are reached.
Further, the transportation demand value of each storage bin is calculated as follows:
in the formula dv i For the transport demand value, ω, of the ith bin i Dangerous weight, rd, of dangerous waste stored in the ith storage bin i Predicting the number of days which can be stored for the current remaining space of the ith storage bin in combination with the future storage amount, td i To predict the number of days that can be stored in combination with the future stored amount assuming that the ith storage bin is currently empty, th i Dnc for presetting the maximum upper limit days for no transportation of the ith storage bin i Actual days of last shipment for the ith bin distance.
The advantage of this embodiment is that too many transport vehicles may lead to transport unsaturation, wasting transport costs, too few transport vehicles may extend the transport time of hazardous waste, increasing the handling risk; the embodiment considers the dangerous degree of dangerous waste, the capacity of a storage bin and the frequency of transportation, when no vehicle is transported for a long time, the denominator approaches zero, and the transportation demand value approaches infinity; the future predicted deposit amount is unchanged when the storage bin is empty, the total number of days in which the storage bin can store is unchanged, and the larger the number of days in which the storage bin can store is, the smaller the molecule is, and the lower the transportation requirement value of the dangerous waste is.
Drawings
The drawings of the present invention are described below.
Figure 1 is a schematic diagram of the internal construction of an industrial park.
FIG. 2 is a schematic diagram of a hazardous waste treatment process.
Detailed Description
The invention is further described below with reference to the drawings and examples.
The industrial park is internally provided with a plurality of storage bins for storing different types of dangerous wastes, the jurisdiction of the industrial park also comprises a plurality of source factories capable of generating various types of dangerous wastes, and each source factory conveys various types of dangerous wastes to the corresponding storage bin for centralized treatment, and the internal structure of the industrial park is shown in figure 1.
An industrial park hazardous waste treatment method based on an artificial intelligence big data model is shown in fig. 2, and comprises the following specific steps:
s1, labeling the types, the packaging dates, the source manufacturers, the weight and the volume information of the dangerous wastes aiming at the packaged dangerous wastes in each park.
Specifically, the hazardous waste includes wastes having hazardous characteristics listed in the national hazardous waste list, such as pharmaceutical manufacturing wastes, pesticide manufacturing wastes, wood preservative manufacturing wastes, refined petroleum wastes, gas production wastes, and basic chemical raw material manufacturing wastes.
S2, the industrial park stores the marked hazardous waste in corresponding storage bins.
S21, judging the production condition of a source manufacturer through an anomaly analysis model when the dangerous waste is put into the storage bin.
S211, selecting an abnormality analysis model of a corresponding manufacturer according to a source manufacturer of the dangerous waste stored in the storage bin.
S212, calling the day of the abnormality analysis, wherein the corresponding manufacturer stores the types and the corresponding weights of the types of the dangerous wastes in various dangerous waste storage bins respectively.
S213, inputting various dangerous waste types and corresponding weights of each type into an anomaly analysis model.
The anomaly analysis model is a random forest model, and the construction method comprises the following steps:
s2131, for the total training set T, N samples are selected randomly each time, wherein N samples are replaced randomly, each sample comprises the same random characteristics, the number k of the random characteristics is smaller than the total characteristic number d, and therefore the N samples construct tree training samples of each decision tree;
s2132, when each tree training sample has M attributes, randomly selecting M attributes from the M attributes when each node of the decision tree needs to be split, meeting the condition M < M, and selecting 1 split attribute belonging to the node from the M attributes by adopting information gain g (D, A) =H (D) -H (D|A) as a basis;
s2133, until each decision tree is no longer split;
s2134, synthesizing voting results of each decision tree, and finally outputting the class with the largest voting result, namely the random forest model.
In the random forest model construction process, a total training set T stores the weight of each dangerous waste in a factory history and whether production marked by people is abnormal or not; n samples are weights of each dangerous waste stored by manufacturers in the historical N days every day, and the random characteristic k is the weight of each dangerous waste;
therefore, the tree training sample of each decision tree is obtained by randomly and replaceably obtaining N times of historical single-day data in the total training set T, and each time of obtained historical single-day data comprises random and same dangerous waste types and weights.
For example:
wherein T is i Is the total training set of the ith source manufacturer, x m For the weight of the m-th hazardous waste on the nth day, 1 and 0 are respectively the normal production condition and the abnormal production condition of the artificial mark.
Wherein the tree training samples of a tree generated randomly are expressed as:
wherein t is k Training samples for the k-th decision tree, the total number of rows is N.
S214, outputting an abnormality analysis day by the abnormality analysis model, and judging whether the manufacturer produces the abnormality or not.
In step S21, a deep learning big data model is constructed by using historical data of the source manufacturer, so that whether the manufacture of the manufacturer is abnormal or not can be judged from the weight of dangerous waste produced from the source manufacturer, and risk prevention and control are facilitated.
S22, judging whether the dangerous waste is packaged by mistake or not when the dangerous waste enters the storage bin.
Specifically, judging whether the package is in error or not, the specific method is as follows:
s221, reading type information marked by the packaged hazardous waste, and judging that the packaging is in error or the storage is in error if the type information marked by the packaged hazardous waste is different from the storage type information stored in the storage bin;
s222, reading the labeling weight and the labeling volume information of the packaged hazardous waste labeling, and calculating the theoretical density according to the labeling weight and the labeling volume;
s223, actually measuring the actual weight of the packaged hazardous waste, and calculating the reference density according to the actual weight and the marked volume;
s224, taking the difference value between the marked weight and the actual weight, the marked volume and the difference value between the theoretical density and the reference density as inputs, and outputting whether the error package exists or not through a decision tree.
The decision tree construction method comprises the following steps:
s2241, a plurality of groups of historical data are obtained, the historical data comprise the difference value between the historical labeling weight and the actual weight, the historical labeling volume and the difference value between the historical theoretical density and the reference density, and whether the packaging errors exist in each group of historical data is labeled;
s2242, calculating information entropy of the root node, H (D) = - Σp (i). Log 2 p (i), where p (i) is the probability of mis-packaging or non-mis-packaging;
s2243, information gains of the respective attributes are calculated, g (D, a) =h (D) -H (d|a);
s2244, selecting the attribute with the maximum information gain as the dividing attribute;
s2245, the information gain is repeatedly calculated, and the information gain maximum attribute is selected until the decision tree can not be split any more.
In step S22, the decision tree automatically combines the difference between the labeled weight and the actual weight, the difference between the labeled volume and the theoretical density and the reference density to determine whether the false package occurs, so that the repeated design of the determination threshold value is avoided for different dangerous waste types, and the decision tree is more suitable for false package determination.
S23, corresponding sensors for detecting leakage of different dangerous wastes are arranged in the storage bin, and whether leakage occurs in the storage bin or not is judged by data collected by the sensors through a leakage prediction model.
S231, judging whether leakage occurs in the storage bin or not through a leakage prediction model, wherein the concrete method comprises the following steps of:
s232, setting one or more gas sensors corresponding to possible leaked gas according to different storage bins and dangerous waste types;
s233, respectively storing corresponding dangerous wastes with different weights in the storage bin, closing the storage bin door, and acquiring the increment of leakage gas during closing the bin door through a gas sensor after preset time;
s234, training data is constructed for leakage gas increment corresponding to different weights;
s235, training a neural network model by training data;
s236, before the storage bin is closed each time, acquiring the actual weight of the dangerous waste in the storage bin, evacuating the leaked gas in the storage bin, inputting the increment of the leaked gas and the actual weight of the dangerous waste acquired by the gas sensor into the neural network model after preset time, and outputting the judgment of whether the leakage occurs in the storage bin.
S3, carrying out normalized transportation in the industrial park, and arranging a transportation vehicle to transport the dangerous wastes in each storage bin to a corresponding disposal place.
S31, predicting future storage quantity of each storage bin through a storage quantity prediction model, wherein the method specifically comprises the following steps of:
s311, acquiring history storage data of each storage bin, and taking the history storage data as training data;
s312, training a neural network model by training data;
s313, predicting future storage quantity of each storage bin by using a neural network model.
In step S31 and step S23, the training method of the neural network model is as follows:
initializing an input layer, a hidden layer and an output layer;
initializing neuron weights and bias values of an input layer, a hidden layer and an output layer;
defining a learning rate, training cycle number and test cycle number;
setting an activation function as a sigmoid function and a loss function as an MSE mean square error function;
calculating a current loss by forward propagation;
calculating an update weight by back propagation;
until the neural network model converges, or the training times are reached.
S32, calculating the transportation demand value of each storage bin, and arranging a transportation vehicle to sequentially transport dangerous wastes in different storage bins according to the transportation demand value from high to low.
The transportation demand value of each storage bin is calculated as follows:
in the formula dv i For the transport demand value, ω, of the ith bin i Dangerous weight, rd, of dangerous waste stored in the ith storage bin i Predicting the number of days which can be stored for the current remaining space of the ith storage bin in combination with the future storage amount, td i To predict the number of days that can be stored in combination with the future stored amount assuming that the ith storage bin is currently empty, th i Dnc for presetting the maximum upper limit days for no transportation of the ith storage bin i Actual days of last shipment for the ith bin distance.
In step S32, it is assumed that the future storage amount of the bin i is predicted by the neural network as shown in the following table:
| t 1 | t 2 | t 3 | t 4 | t 5 | t 6 | |
| storage bin i | 3 | 4 | 3 | 6 | 4 | 2 |
The total storage capacity of the storage bin i is 20, the storage capacity is 8 currently, and the last transportation is 3 days later, dnc i 3, a maximum upper limit number th of non-transportation days is preset i 10, the dangerous weight omega of the stored dangerous wastes i 1.1;
the method can be calculated as follows: the current empty allowance of the storage bin is 12, the total storage amount of three days in the future is 10, and the storage amount of the storage bin is less than the storage amount of the fourth day, rd i 3; the total empty allowance of the storage bin is 20, the total storage amount of five future days is 20, td i 5;
from which the transportation demand value can be calculated
The current transportation demand value of each storage bin is calculated, and vehicles are arranged to be transported from high to low according to the transportation demand value.
Finally, it should be noted that: the above embodiments are only for illustrating the technical aspects of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the above embodiments, it should be understood by those of ordinary skill in the art that: modifications and equivalents may be made to the specific embodiments of the invention without departing from the spirit and scope of the invention, which is intended to be covered by the claims.
Claims (10)
1. The industrial park hazardous waste treatment method based on the artificial intelligence big data model is characterized in that a plurality of storage bins for storing hazardous wastes of different categories are built in the industrial park, and the hazardous waste treatment method comprises the following steps:
labeling the type, packaging date, source manufacturer, weight and volume information of the dangerous waste aiming at the packaged dangerous waste;
storing the marked hazardous wastes in storage bins corresponding to various types;
transporting the hazardous waste in each storage bin to a corresponding disposal place by standardization;
judging the production condition of a source manufacturer through an anomaly analysis model when dangerous wastes enter a storage bin;
judging whether the dangerous waste is packaged by mistake when the dangerous waste enters a storage bin;
corresponding sensors for detecting leakage of different dangerous wastes are arranged in the storage bin, and data acquired by the sensors judge whether leakage occurs in the storage bin or not through a leakage prediction model;
and predicting the future storage quantity of each storage bin through a storage quantity prediction model, calculating the transportation demand value of each storage bin, and arranging a transportation vehicle to sequentially transport dangerous wastes in different storage bins according to the transportation demand value.
2. The industrial park hazardous waste treatment method based on the artificial intelligence big data model according to claim 1, wherein the hazardous waste comprises any one of the following and above:
pharmaceutical manufacturing waste, pesticide manufacturing waste, wood preservative manufacturing waste, refined petroleum waste, gas production waste, and base chemical raw material manufacturing waste.
3. The industrial park hazardous waste treatment method based on the artificial intelligence big data model according to claim 1, wherein the production condition of a source manufacturer is judged through an anomaly analysis model, and the specific method is as follows:
according to source manufacturers of dangerous wastes stored in the storage bin, selecting an abnormality analysis model of the corresponding manufacturer;
the day of the abnormal analysis is called, and the corresponding manufacturer stores the types of the dangerous wastes and the corresponding weights of each type in various dangerous waste storage bins respectively;
inputting various dangerous waste types and corresponding weights of each type into an abnormality analysis model;
the abnormality analysis model outputs the day of the abnormality analysis, and whether the manufacturer produces the abnormality exists.
4. The industrial park hazardous waste treatment method based on the artificial intelligence big data model according to claim 3, wherein the anomaly analysis model is a random forest model, and the construction method is as follows:
for the total training set T, N samples are shared, N samples are randomly selected each time with a replacement, each sample comprises the same random feature, the number k of the random features is smaller than the total feature number d, and therefore the N samples construct tree training samples of each decision tree;
when each tree training sample has M attributes, randomly selecting M attributes from the M attributes when each node of the decision tree needs to be split, satisfying the condition M < M, and selecting 1 split attribute belonging to the node from the M attributes by adopting information gain g (D, A) =H (D) -H (D|A) as a basis;
until each decision tree is no longer split;
the voting result of each decision tree is synthesized, and the category with the largest final voting result is finally output by the random forest model;
in the random forest model construction process, a total training set T stores the weight of each dangerous waste in a factory history and whether production marked by people is abnormal or not; n samples are weights of each dangerous waste stored by manufacturers in the historical N days every day, and the random characteristic k is the weight of each dangerous waste;
therefore, the tree training sample of each decision tree is obtained by randomly and replaceably obtaining N times of historical single-day data in the total training set T, and each time of obtained historical single-day data comprises random and same dangerous waste types and weights.
5. The industrial park hazardous waste treatment method based on the artificial intelligence big data model according to claim 1, wherein when hazardous waste is put into a storage bin, judging whether there is a false package or not, the specific method is as follows:
reading type information marked by the packaged hazardous waste, and judging that the packaging is in error packaging or error storage if the type information marked by the packaged hazardous waste is different from the storage type information stored in the storage bin;
reading the labeling weight and the labeling volume information of the labeling of the packaged hazardous waste, and calculating the theoretical density according to the labeling weight and the labeling volume;
actual weight of the packaged hazardous waste is measured, and reference density is calculated according to the actual weight and the marked volume;
and taking the difference value between the marked weight and the actual weight, the marked volume and the difference value between the theoretical density and the reference density as inputs, and outputting whether the error package exists or not through a decision tree.
6. The industrial park hazardous waste treatment method based on the artificial intelligence big data model according to claim 5, wherein the decision tree construction method is as follows:
acquiring a plurality of groups of historical data, wherein the historical data comprise the difference value between the historical labeling weight and the actual weight, the historical labeling volume and the difference value between the historical theoretical density and the reference density, and labeling whether each group of historical data is packaged by mistake or not;
calculating the information entropy of the root node, H (D) = - Σp (i). Log 2 p (i), where p (i) is the probability of mis-packaging or non-mis-packaging;
calculating information gain of each attribute, g (D, a) =h (D) -H (d|a);
selecting the attribute with the maximum information gain as a dividing attribute;
and repeatedly calculating the information gain and selecting the maximum information gain attribute until the decision tree cannot be split any more.
7. The industrial park hazardous waste treatment method based on the artificial intelligence big data model according to claim 1, wherein whether leakage occurs in the storage bin is judged by a leakage prediction model, and the specific method is as follows:
setting one or more gas sensors corresponding to possible leaked gas according to different storage bins and dangerous waste types;
corresponding dangerous wastes with different weights are respectively stored in the storage bin, the storage bin door is closed, and the increment of leakage gas during closing of the bin door is obtained through the gas sensor after preset time;
training data are constructed for different weights and the leakage gas increment corresponding to the different weights;
training a neural network model with training data;
before the storage bin is closed each time, the actual weight of dangerous waste in the storage bin is obtained, leaked gas in the storage bin is emptied, after the preset time, the increment of the leaked gas obtained by the gas sensor and the actual weight of the dangerous waste are input into the neural network model, and the judgment of whether the leakage occurs in the storage bin is output.
8. The industrial park hazardous waste treatment method based on the artificial intelligence big data model according to claim 1, wherein the future deposit quantity of each storage bin is predicted by a deposit quantity prediction model, and the method is as follows:
acquiring historical storage data of each storage bin, and taking the historical storage data as training data;
training a neural network model with training data;
and predicting future storage quantity of each storage bin by using a neural network model.
9. The industrial park hazardous waste treatment method based on the artificial intelligence big data model according to claim 7 or 8, wherein the training method of the neural network model is as follows:
initializing an input layer, a hidden layer and an output layer;
initializing neuron weights and bias values of an input layer, a hidden layer and an output layer;
defining a learning rate, training cycle number and test cycle number;
setting an activation function as a sigmoid function and a loss function as an MSE mean square error function;
calculating a current loss by forward propagation;
calculating an update weight by back propagation;
until the neural network model converges, or the training times are reached.
10. The industrial park hazardous waste treatment method based on the artificial intelligence big data model according to claim 1, wherein the transportation requirement value of each storage bin is calculated as follows:
in the formula dv i For the transport demand value, ω, of the ith bin i Dangerous weight, rd, of dangerous waste stored in the ith storage bin i Predicting the number of days which can be stored for the current remaining space of the ith storage bin in combination with the future storage amount, td i To predict the number of days that can be stored in combination with the future stored amount assuming that the ith storage bin is currently empty, th i Dnc for presetting the maximum upper limit days for no transportation of the ith storage bin i Actual days of last shipment for the ith bin distance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311759161.4A CN117726257B (en) | 2023-12-20 | 2023-12-20 | Industrial park hazardous waste treatment method based on artificial intelligence big data model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311759161.4A CN117726257B (en) | 2023-12-20 | 2023-12-20 | Industrial park hazardous waste treatment method based on artificial intelligence big data model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117726257A true CN117726257A (en) | 2024-03-19 |
| CN117726257B CN117726257B (en) | 2024-07-05 |
Family
ID=90204965
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311759161.4A Active CN117726257B (en) | 2023-12-20 | 2023-12-20 | Industrial park hazardous waste treatment method based on artificial intelligence big data model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117726257B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117236750A (en) * | 2023-09-08 | 2023-12-15 | 重庆环问问科技有限公司 | Full-period carbon emission calculation method based on difference data analysis |
| CN118506341A (en) * | 2024-05-08 | 2024-08-16 | 北京互链时空数字科技有限公司 | Unattended weighing management system and method |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108154347A (en) * | 2018-01-15 | 2018-06-12 | 上海电气集团股份有限公司 | A kind of hazardous waste operation management system |
| US20180232690A1 (en) * | 2017-02-14 | 2018-08-16 | United Parcel Service Of America, Inc. | Dangerous goods shipping management systems |
| CN110242865A (en) * | 2019-07-09 | 2019-09-17 | 北京讯腾智慧科技股份有限公司 | A kind of gas leakage detection determination method and system being easy to Continuous optimization |
| CN111256913A (en) * | 2020-03-18 | 2020-06-09 | 江苏警官学院 | Chemical dangerous article leakage detection method for laboratory |
| CN114239385A (en) * | 2021-11-30 | 2022-03-25 | 南京邮电大学 | Intelligent decision making system and method for warehouse resource allocation |
| CN114912787A (en) * | 2022-05-06 | 2022-08-16 | 南京大学 | Intelligent assessment method for enterprise dangerous waste concealing, reporting and missing reporting risks |
| CN115132389A (en) * | 2022-06-29 | 2022-09-30 | 华能核能技术研究院有限公司 | Method, device, equipment and storage medium for predicting seawater leakage of condenser of nuclear power plant |
| CN115641044A (en) * | 2022-11-21 | 2023-01-24 | 四川易链科技有限公司 | Internet of things technology-based hazardous waste disposal monitoring system and method |
| WO2023045829A1 (en) * | 2021-09-24 | 2023-03-30 | 中兴通讯股份有限公司 | Service abnormality prediction method and device, storage medium, and electronic device |
-
2023
- 2023-12-20 CN CN202311759161.4A patent/CN117726257B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180232690A1 (en) * | 2017-02-14 | 2018-08-16 | United Parcel Service Of America, Inc. | Dangerous goods shipping management systems |
| CN108154347A (en) * | 2018-01-15 | 2018-06-12 | 上海电气集团股份有限公司 | A kind of hazardous waste operation management system |
| CN110242865A (en) * | 2019-07-09 | 2019-09-17 | 北京讯腾智慧科技股份有限公司 | A kind of gas leakage detection determination method and system being easy to Continuous optimization |
| CN111256913A (en) * | 2020-03-18 | 2020-06-09 | 江苏警官学院 | Chemical dangerous article leakage detection method for laboratory |
| WO2023045829A1 (en) * | 2021-09-24 | 2023-03-30 | 中兴通讯股份有限公司 | Service abnormality prediction method and device, storage medium, and electronic device |
| CN114239385A (en) * | 2021-11-30 | 2022-03-25 | 南京邮电大学 | Intelligent decision making system and method for warehouse resource allocation |
| CN114912787A (en) * | 2022-05-06 | 2022-08-16 | 南京大学 | Intelligent assessment method for enterprise dangerous waste concealing, reporting and missing reporting risks |
| CN115132389A (en) * | 2022-06-29 | 2022-09-30 | 华能核能技术研究院有限公司 | Method, device, equipment and storage medium for predicting seawater leakage of condenser of nuclear power plant |
| CN115641044A (en) * | 2022-11-21 | 2023-01-24 | 四川易链科技有限公司 | Internet of things technology-based hazardous waste disposal monitoring system and method |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117236750A (en) * | 2023-09-08 | 2023-12-15 | 重庆环问问科技有限公司 | Full-period carbon emission calculation method based on difference data analysis |
| CN117236750B (en) * | 2023-09-08 | 2024-07-05 | 重庆环问问科技有限公司 | Full-period carbon emission calculation method based on difference data analysis |
| CN118506341A (en) * | 2024-05-08 | 2024-08-16 | 北京互链时空数字科技有限公司 | Unattended weighing management system and method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117726257B (en) | 2024-07-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN117726257B (en) | Industrial park hazardous waste treatment method based on artificial intelligence big data model | |
| CN107220734A (en) | CNC Lathe Turning process Energy Consumption Prediction System based on decision tree | |
| CN106991437A (en) | The method and system of sewage quality data are predicted based on random forest | |
| CN112183709B (en) | Method for predicting and early warning excessive dioxin in waste incineration gas | |
| CN111489046A (en) | Regional food safety evaluation model based on supply chain and BP neural network | |
| CN117312879B (en) | Injection molding machine production data supervision and early warning method, system and medium | |
| CN110980050B (en) | Garbage disposal method, storage medium and garbage storage equipment | |
| CN111860701A (en) | Denitration system working condition discrimination preprocessing method based on clustering method | |
| CN116523428B (en) | Logistics management method, device, equipment and storage medium of box-type logistics system | |
| CN114023399A (en) | Air particulate matter analysis early warning method and device based on artificial intelligence | |
| CN111191855B (en) | Water quality abnormal event identification and early warning method based on pipe network multi-element water quality time sequence data | |
| CN116485020A (en) | Supply chain risk identification early warning method, system and medium based on big data | |
| CN118228606A (en) | Online quick pollution prediction method for polluted land block based on machine learning, electronic equipment and storage medium | |
| CN111832993B (en) | Predictive maintenance method for warehouse logistics system and related components | |
| CN113657023A (en) | Near-surface ozone concentration inversion method based on combination of machine learning and deep learning | |
| CN116432831A (en) | Prediction method for abnormal event of chemical risk type | |
| CN116522065A (en) | Coal mill health degree assessment method based on deep learning | |
| CN117763341A (en) | Internet hazardous chemical substance safety production monitoring platform constructed by utilizing big data technology | |
| CN114969068B (en) | Urban pressure pipe network real-time flow monitoring data analysis method and system | |
| CN116090710A (en) | Management method, system, electronic equipment and medium for enterprise pollution discharge permission | |
| CN118392255B (en) | Method, system and equipment for monitoring operation condition of electromechanical equipment of water plant | |
| CN118626911A (en) | Port handling equipment health index construction method, device and storage medium | |
| CN118410312B (en) | Intelligent control method for dry bulk cargo wharf sewage treatment equipment | |
| CN117408128B (en) | Air quality simulation and observation machine learning NO2Coupling forecasting method | |
| CN118674596A (en) | Urban net zero emission tracking management method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |