CN116432831A - Prediction method for abnormal event of chemical risk type - Google Patents
Prediction method for abnormal event of chemical risk type Download PDFInfo
- Publication number
- CN116432831A CN116432831A CN202310273471.9A CN202310273471A CN116432831A CN 116432831 A CN116432831 A CN 116432831A CN 202310273471 A CN202310273471 A CN 202310273471A CN 116432831 A CN116432831 A CN 116432831A
- Authority
- CN
- China
- Prior art keywords
- monitoring
- risk data
- chemical
- risk
- monitoring risk
- 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.)
- Pending
Links
- 239000000126 substance Substances 0.000 title claims abstract description 80
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 63
- 238000000034 method Methods 0.000 title claims abstract description 51
- 238000012544 monitoring process Methods 0.000 claims abstract description 156
- 238000012549 training Methods 0.000 claims abstract description 19
- 238000013528 artificial neural network Methods 0.000 claims abstract description 17
- 238000003062 neural network model Methods 0.000 claims abstract description 6
- 230000008569 process Effects 0.000 claims description 16
- 238000012360 testing method Methods 0.000 claims description 11
- 230000004913 activation Effects 0.000 claims description 6
- 238000007621 cluster analysis Methods 0.000 claims description 4
- 238000009499 grossing Methods 0.000 claims description 3
- 238000012935 Averaging Methods 0.000 claims description 2
- 239000007789 gas Substances 0.000 description 32
- 230000006870 function Effects 0.000 description 16
- 231100000331 toxic Toxicity 0.000 description 12
- 230000002588 toxic effect Effects 0.000 description 12
- 230000007613 environmental effect Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 238000004590 computer program Methods 0.000 description 5
- 238000003860 storage Methods 0.000 description 5
- 238000004880 explosion Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 239000007788 liquid Substances 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 230000005856 abnormality Effects 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 2
- 239000013043 chemical agent Substances 0.000 description 2
- 238000002485 combustion reaction Methods 0.000 description 2
- 230000006378 damage Effects 0.000 description 2
- 238000009792 diffusion process Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 231100000572 poisoning Toxicity 0.000 description 2
- 230000000607 poisoning effect Effects 0.000 description 2
- 231100000419 toxicity Toxicity 0.000 description 2
- 230000001988 toxicity Effects 0.000 description 2
- 208000012260 Accidental injury Diseases 0.000 description 1
- 108010001267 Protein Subunits Proteins 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 231100000481 chemical toxicant Toxicity 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 230000001939 inductive effect Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000000197 pyrolysis Methods 0.000 description 1
- 230000002269 spontaneous effect Effects 0.000 description 1
- 239000002341 toxic gas Substances 0.000 description 1
- 239000003440 toxic substance Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Educational Administration (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for predicting an abnormal event of a chemical risk type, which comprises the following steps: acquiring monitoring risk data and taking the monitoring risk data as an observation value; calculating the predicted value and the observed value of the same class of monitoring risk data at the previous moment to obtain the predicted value of each class of monitoring risk data at the current moment by weighted average; inputting the observed values of various monitoring risk data into an LSTM neural network for training to obtain the trained LSTM neural network; and inputting the observed values of various monitoring risk data into the LSTM neural network model for prediction to obtain a first predicted value, replacing the predicted value with the first predicted value, and if the deviation between the first predicted value and the observed value meets the corresponding condition, considering that an abnormal event occurs, and calculating the probability of system risk. The invention realizes the predictive early warning of the parameter level and the system level, and can also calculate the specific probability.
Description
Technical Field
The invention relates to the technical field of dangerous chemical prediction and early warning, in particular to a method for predicting a chemical risk type abnormal event.
Background
In warehouses or other storage areas where chemicals (e.g., chemical agents) are stored, casualties and property damage (e.g., damage to chemical agents) often occur with leakage of chemicals due to instability and toxicity of the chemicals. The chemical can be called as dangerous chemical due to instability and toxicity, the dangerous chemical has the properties of poisoning, corrosion, explosion, combustion supporting and the like, and has the characteristics of extremely toxic chemical and other chemicals which are harmful to human bodies, facilities and environment, various dangerous chemical types, various deliquescence, pyrolysis, spontaneous combustion and explosion reactions of leakage matters in accident sites, complex products and various physical and chemical properties.
There is no good prediction method in the prior art as to whether or not dangerous chemical is dangerous and what type of dangerous is happening during normal storage, transportation, etc., that is, whether or not (how likely) an abnormal event of chemical is happening and what type of abnormal event is happening. As a relatively close prior art, chinese patent publication No. CN217443286U, which verifies the type and concentration of dangerous gas based on gas detection to realize dangerous early warning and failure early warning of dangerous chemicals, but like other prior art, there are the following disadvantages and drawbacks:
(1) The method is characterized in that the method is used for judging dangerous chemicals only through gas detection, does not consider factors such as environmental weather, tank field technology and the like, and does not form a dynamic early warning mode;
(2) The gas leakage condition of the dangerous chemical which occurs at present can be solved, and the advanced early warning condition of the dangerous chemical risk state is not solved for the dangerous chemical, namely the prediction early warning can not be carried out in advance.
Disclosure of Invention
In view of the shortcomings of the prior art, it is an object of the present invention to provide a method for predicting an abnormal event of the risk type of chemicals, which can solve the problems mentioned in the background art.
The technical scheme for realizing the purpose of the invention is as follows: a method of predicting a chemical risk type anomaly event, comprising the steps of:
step 1: obtaining different kinds of monitoring risk data of the chemical and taking the monitoring risk data as an observed value, wherein the monitoring risk data comprises time information;
step 2: giving weight to each kind of monitoring risk data, calculating the predicted value of each kind of monitoring risk data at the current moment according to the weight of each kind of monitoring risk data, the predicted value and the observed value of the same kind of monitoring risk data at the previous moment, thereby obtaining the predicted value of each kind of monitoring risk data,
and comparing the predicted value and the observed value of the same class at the same moment, judging whether the monitoring risk data of the same class at the corresponding moment is abnormal according to the comparison result, and if so, judging that the chemical product is about to generate abnormal events of the monitoring risks at the moment, thereby judging that the chemical product is about to generate abnormal events of the monitoring risks of different classes at different moments.
Further, in step 1, obtaining different kinds of monitoring risk data of the chemical means obtaining different kinds of monitoring risk data of the chemical of the target area.
Further, the target area is a pre-delimited area.
Further, the current area includes at least two distinct sub-unit areas, and the same type of monitoring risk data for the same type of monitoring risk data is the same type of monitoring risk data of one of the sub-units, or an average or weighted average of the same type of monitoring risk data of a plurality of the sub-units.
Further, the two different subunit areas are respectively an enterprise area and a public area, the enterprise area is the enterprise area where the chemical storage place is located or the area within a certain range of the enterprise, the public area is the public area formed by transport passing areas where the chemical is transported out from the enterprise area, monitoring risk data of the enterprise area comprise combustible gas concentration, toxic and harmful gas concentration, device medium pressure, device medium liquid level, device medium temperature, environment temperature and environment humidity, and monitoring risk data of the public area comprise the combustible gas concentration, the toxic and harmful gas concentration, the environment temperature and the environment humidity.
Further, the monitoring risk data, the corresponding monitoring terminal and the monitoring range are shown in the following table:
| parameter type | Monitoring terminal | Monitoring range |
| Concentration of combustible gas | Combustible gas detector | Combustible gas concentration for public areas and enterprise hazard sources |
| Concentration of toxic and harmful gases | Toxic and harmful gas detector | Concentration of toxic and harmful gases in public areas and enterprise hazard sources |
| Wind speed/direction | Anemometer and anemometer | Environmental wind speed and wind direction parameters of chemical industry park |
| Ambient temperature/humidity | Temperature and humidity sensor | Environmental temperature and humidity of chemical industry park |
| Device medium pressure | Pressure sensor | Pressure of medium for dangerous device of enterprise |
| Device medium level | Liquid level sensor | Medium level of enterprise dangerous device |
| Device medium temperature | Temperature sensor | Medium temperature of enterprise dangerous device |
。
Further, in step 2, the specific implementation process of the weighting value given to each type of monitoring risk data includes the following steps:
firstly projecting the k-th monitoring risk data in the optimal projection direction according to a projection pursuit cluster analysis method, wherein k is more than or equal to 1, and obtaining a corresponding projection component omega k Then calculating the weight alpha of the kth monitoring risk data according to the following formula k ：
Wherein n represents the total class number of various monitoring risk data, namely the data quantity of the monitoring risk data.
Further, in step 2, the predicted values of the various types of monitoring risk data at the current moment are calculated according to the weight of each type of monitoring risk data, the predicted values and the observed values of the same type of monitoring risk data at the previous moment, and the average value obtained by the average or the weighted average is used as the predicted value of each type of monitoring risk data at the corresponding moment.
Further, the weighted average is performed by using an exponential smoothing model as shown in formula (1):
Further, the predicted value of the kth monitoring risk data of the latest i (i is more than or equal to 2) times is recursively substituted into the formula (1) to obtain the formula (2):
weighted average is performed according to formula (2).
Further, in step 2, the same kind of predicted value and observed value at the same moment are compared, and whether the monitoring risk data at the corresponding moment is abnormal or not is judged according to the comparison result, if so, the abnormal event of the monitoring risk of the chemical at the moment is judged, and the specific implementation process comprises the following steps:
if the deviation between the observed value and the predicted value
And the chemical is considered to have an abnormal event of the kth monitoring risk at the time t, so that the prediction and early warning of the parameter level are realized.
Further, after step 2, the method further comprises the following steps:
step 3: dividing the observation values of various monitoring risk data into two groups, wherein one group is used as a training set, the other group is used as a test set, inputting the training set into the LSTM neural network for training, and testing by using the test set, so that the trained LSTM neural network is obtained;
step 4: inputting the observed values of various monitoring risk data into the LSTM neural network model obtained in the step 3 for prediction to obtain predicted values of various monitoring parameters at various moments, marking the predicted values as first predicted values, replacing the predicted values in the step 2 with the first predicted values,
if the first predicted value X of the kth monitoring parameter at the moment t t,k Observations of kth-class monitoring risk data with time t
Deviation of->
It is considered that the kth monitoring parameter will have an abnormal event at time t.
Further, will
Substituting x into s igmod to activate the function, and taking the obtained result as the probability that the kth monitoring parameter will generate an abnormal event at the time t, namely calculating according to the following formula:
f (x) is the probability that the kth monitoring parameter will have an abnormal event at time t.
Further, the average value of the probabilities of various monitoring parameters of abnormal events at the same moment or the average value after weighted average is used as the probability of system risks at corresponding moments, and the probability of the system risks represents the comprehensive probability of the chemical occurrence of the abnormal events comprising various monitoring risks.
Further, when the probability of the system risk exceeds a preset threshold, judging that the chemical is at risk.
The beneficial effects of the invention are as follows: according to the method, a process of sudden accumulation of dangerous chemicals from risks to abnormal events is regarded as a process of increasing risk values, and fluctuation conditions of situation elements such as combustible gas, environmental meteorological parameters and tank farm technological parameters are utilized to build a dangerous chemical parameter risk model and a system risk model, the dangerous chemical parameter risk values and the system risk values are respectively predicted and judged to identify dangerous chemical risk situations, two overall risk situation researching and judging mechanisms are provided, early warning situations are realized when the system risk probability is larger than a set threshold value, prediction of abnormality of various risk parameters is realized, prediction of overall risks is also realized, namely prediction of system risks is also realized, specific probability can be calculated, and accordingly prediction and early warning of parameter levels and system levels are realized. Meanwhile, a multivariate LSTM algorithm is introduced to predict a risk entropy time sequence of the dangerous chemical system, a system risk entropy prediction model based on the multivariate LSTM is established, an optimal parameter combination of the LSTM model is searched by using a cross verification algorithm, and the prediction model obtained by Gaussian radial basis functions and cross verification has good generalization capability and higher prediction accuracy.
Drawings
FIG. 1 is a schematic flow chart of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present application more apparent, the following detailed description of specific embodiments thereof is given with reference to the accompanying drawings. It is to be understood that the specific embodiments described herein are merely illustrative of the application and not limiting thereof. It should be further noted that, for convenience of description, only some, but not all of the matters related to the present application are shown in the accompanying drawings. Before discussing exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although a flowchart depicts operations (or steps) as a sequential process, many of the operations can be performed in parallel, concurrently, or at the same time. Furthermore, the order of the operations may be rearranged. The process may be terminated when its operations are completed, but may have additional steps not included in the figures. The processes may correspond to methods, functions, procedures, subroutines, and the like.
As shown in fig. 1, a method for predicting a chemical risk type abnormal event includes the steps of:
step 1: different kinds of monitoring risk data of the chemicals in the target area are obtained, wherein the monitoring risk data comprise time information, namely, various monitoring risk data corresponding to different time are obtained through acquisition, and the obtained monitoring risk data are observation values.
The target area may be pre-defined and in this embodiment includes a common area comprising an enterprise area or areas within a certain range of the enterprise where the chemicals are stored and/or transport-through areas where the chemicals are transported out of the enterprise area. Areas like tank farm, production site, storage area etc. may be considered as the enterprise area. The public area is a necessary area for carrying out logistics transportation between enterprises and the outside in the park, particularly a factory boundary and a trunk road, the enterprise area and the public area can be defined in advance according to the actual condition of the area where the chemical is located, so that different areas are obtained, each area corresponds to one subunit area respectively, and a plurality of different subunit areas can be obtained.
In practical application, the target area can be divided and selected according to practical situations.
The monitoring risk data includes physicochemical property parameters of combustible gas concentration, toxic and harmful gas concentration, device medium pressure, device medium level, device medium temperature, ambient humidity and the like which influence the stability of chemicals and can trigger abnormal events (such as explosion, leakage and other accidents) of the chemicals. Typically, the monitoring risk data for the enterprise area includes a combustible gas concentration, a toxic and harmful gas concentration, a device medium pressure, a device medium level, a device medium temperature, an ambient humidity, etc., and the monitoring risk data for the public area includes a combustible gas concentration, a toxic and harmful gas concentration, an ambient temperature, an ambient humidity, etc.
Each of the above monitoring risk data is a type of monitoring parameter, for example, the concentration of the combustible gas is the type 1 monitoring risk data, the concentration of the toxic and harmful gas is the type 2 monitoring risk data, and so on, and other monitoring risk data are also corresponding types of monitoring parameters.
The risk data monitoring in the enterprise area mainly aims at medium leakage events caused by failure of dangerous devices, the medium leakage usually causes gas leakage diffusion and environmental pollution events, a flammable gas leakage electrode is easy to induce fire explosion accidents, and toxic gas leakage is easy to cause personnel poisoning, atmospheric pollution and the like. Corresponding sensing terminals need to be deployed near significant sources of risk to the enterprise (tank farm, production site, storage area) to monitor changes in process parameters (i.e., parameters that can affect chemicals).
The public area is the necessary area that the chemical is deposited in the garden that enterprise and external carry out commodity circulation transportation, especially factory boundary and trunk road need bear the multiple accident risk pressures such as removal danger source, public piping lane facility on the one hand, and on the other hand still probably receive the accident injury of enterprise beyond factory boundary scope to go through, and the dynamic characteristic of its risk is more obvious. The public area needs to timely sense the medium leakage diffusion event and monitor the environmental meteorological parameters, so that the influence of inducing domino effect accidents on more enterprises is avoided.
In an alternative embodiment, the monitoring risk data, the corresponding monitoring terminals and the monitoring ranges are shown in the following table:
| parameter type | Monitoring terminal | Monitoring range |
| Concentration of combustible gas | Combustible gas detector | Combustible gas concentration for public areas and enterprise hazard sources |
| Concentration of toxic and harmful gases | Toxic and harmful gas detector | Concentration of toxic and harmful gases in public areas and enterprise hazard sources |
| Wind speed/direction | Anemometer and anemometer | Environmental wind speed and wind direction parameters of chemical industry park |
| Ambient temperature/humidity | Temperature and humidity sensor | Environmental temperature and humidity of chemical industry park |
| Device medium pressure | Pressure sensor | Pressure of medium for dangerous device of enterprise |
| Device medium level | Liquid level sensor | Medium level of enterprise dangerous device |
| Device medium temperature | Temperature sensor | Medium temperature of enterprise dangerous device |
。
The detection range is a certain area in the target area.
Step 2: each class of monitoring risk data is given a weight.
In an optional embodiment, the monitoring risk data of the kth (k is greater than or equal to 1) class is projected in the optimal projection direction according to a projection pursuit cluster analysis method to obtain a corresponding projection component omega k Then calculating the weight alpha of the kth monitoring risk data according to the following formula k ：
Wherein n represents the total class number of various monitoring risk data, namely the data quantity of the monitoring risk data.
The projection pursuit cluster analysis method is a data processing method in the fields of statistics and computer technology, so that it is not repeated here how to perform projection according to an optimal projection direction to obtain a projection component of the optimal projection direction.
After the weight of each type of monitoring risk data is obtained, calculating the predicted value of each type of monitoring risk data at the current moment according to the weight of each type of monitoring risk data, the predicted value of the same type of monitoring risk data at the previous moment and the observed value of the same type of monitoring risk data at the previous moment, and carrying out weighted average on the predicted values of each type of monitoring risk data at the current moment to obtain the predicted value of each type of monitoring risk data at the current moment, wherein the observed value is the value obtained by actual collection, namely the monitoring risk data obtained in the step 1.
In an alternative embodiment, the weighted average is performed by an exponential smoothing model. Specifically, the predicted value of the monitoring risk data for the kth class at the current time t
As shown in formula (1):
wherein H is t-1,k Representing the observed value of the kth monitoring risk data at the time (t-1), namely the observed value of the monitoring risk data of the same class at the last time,
a predicted value of the monitoring risk data of the kth class at the time (t-1) (i.e., the last time).
In step 1, various monitoring risk data of a target area (enterprise area and/or public area) are collected within a period of time, so that various monitoring risk data at different moments are obtained, and the monitoring risk data can be used as observation values at corresponding moments.
In an alternative embodiment, if the k-th monitored risk data includes enterprise area and public area monitoring, the data collected by the two areas are averaged or weighted according to the weight.
In an alternative manner, the predicted value of the kth monitoring risk data of the latest i (i.gtoreq.2) times is recursively substituted into formula (1) to obtain formula (2):
in an alternative embodiment, various types of monitoring risk data satisfy a normal distribution, so that abnormal data can be screened, the abnormal data reflects that abnormal events possibly occur in the chemical, and abnormal data can be screened according to the 3 sigma principle of the normal distribution, and the abnormal data possibly means that the chemical is dangerous, namely the abnormal events possibly occur.
I.e. when the deviation between the observed value and the predicted value
When the method is used, the k-th monitoring risk data can be considered to be abnormal, the risk is large, and an abnormal event occurs, namely the chemical is considered to be the abnormal event of the k-th monitoring risk at the time t, so that the predictive early warning of the parameter level is realized.
The above steps can realize predictive early warning of parameter level, but cannot realize predictive early warning of system level including all kinds of monitoring risks and probability of occurrence of system risks. For this reason, the following steps need to be further performed.
Step 3: the observation values of various monitoring risk data are divided into two groups, one group is used as a training set, the other group is used as a test set, the training set is input into the LSTM neural network for training, and the test set is used for testing, so that the trained LSTM neural network is obtained.
In an alternative embodiment, the observed values of various monitoring risk data at the last moment are taken as a test set, and the rest data are taken as a training set. Or selecting a plurality of monitoring risk data from the rest data, wherein each type of monitoring risk data at least comprises one training set, and the length of the training set is marked as T which is less than or equal to n-1.
Taking into consideration the correlation of the sequence formed by the observation values of the monitoring risk data in the time dimension, carrying out delay reconstruction on the sequence, constructing a data set through a sliding time window and inputting an LSTM neural network, and assuming that the window length is d, inputting the actual length L=T-d of the data, wherein at the moment of T > d, the input of the LSTM neural network is H t-d,k ,H t-d+1,k ,H t-1,k ...H t,k K.epsilon.1, 2,3. I.e. the input is a value representing each point in time of each monitored risk data parameter within the time interval of dThe matrix model of its training set sliding time window is as follows:
wherein, t=d+1, d+2, …, l+1, thereby simultaneously considering the input-output range of the nonlinear activation function in the LSTM neural network, the training efficiency of the LSTM neural network, and other factors.
In an alternative embodiment, to reduce the impact of training set orders on the LSTM neural network while accelerating the training process of the LSTM neural network, the test set and training set data are normalized, e.g., normalized to the range of 0-1, i.e., mapped to [0,1 ]]In the range, wherein the j-th data is normalized data x j The method is calculated according to the following formula:
in the above, H t,j Represents the j-th data at time t, min (H t ， j ) Represents H t,j Minimum value of monitoring risk data of the same kind, max (H t ， j ) Represents H t,j The maximum value of risk data is monitored in the same class.
In the training process of the LSTM neural network, super parameters such as the number of hidden layers, the number of neurons, the sliding time window length and the like of the LSTM neural network are adjusted until an optimal LSTM neural network model is obtained, and a final LSTM neural network model is obtained.
Step 4: and (3) inputting the observed values of various monitoring risk data into the LSTM neural network model obtained in the step (3) for prediction, and obtaining the predicted values of various monitoring parameters at various moments. If the predicted value X of the kth monitoring parameter at the moment t tk Observations of kth-class monitoring risk data with time t
Deviation of->
It is considered that the kth monitoring parameter will have an abnormal event, i.e. a hazard, at time t, thereby enabling an abnormal prediction of the chemical.
After judging that the kth monitoring parameter is abnormal at the time t, the system can also send out an alarm reminding.
In step 2 and step 4, the comparison between the predicted value and the observed value may be performed to determine whether the monitored parameter of the corresponding class is abnormal, that is, whether an abnormal event occurs, which is different from the way in which the predicted value is calculated in step 2 and step 4. In actual use, the predicted value obtained by using the LSTM neural network in the step 4 is used for system level prediction, and the predicted value in the step 2 is used for parameter level prediction.
In an alternative embodiment, the foregoing merely indicates that it may have an abnormal event, but the probability of occurrence of the abnormal event cannot be calculated yet. For this purpose, the s igmod activation function that is normally used in artificial intelligence is applied to this embodiment. The s igmod activation function is adopted because the s igmoid function can 'compress' the value of an input sample to between 0 and 1, the output range is limited and extremely stable, the value range of x is between minus infinity and plus infinity, the value of x is monotonously continuous, the larger the obtained function value is, the inventor finds that the s igmod activation function is very suitable for the probability characteristic of an abnormal event of a chemical, and therefore, the s igmod activation function is adopted to calculate the probability.
Will be
Substituting x into s igmod to activate the function, and obtaining a result which is the probability that the kth monitoring parameter will generate an abnormal event at the time t. Namely, the method is calculated according to the following formula:
f (x) is the probability that the kth monitoring parameter will have an abnormal event at time t.
In an alternative embodiment, an average value of probabilities of abnormal events occurring at the same time or an average value after weighted averaging of various monitoring analysis parameters is used as a probability of system risks occurring at corresponding times, and when the probability of the system risks exceeds a preset threshold, the system risks are considered to occur and an alarm is given, so that predictive early warning is realized. The probability of the system risk characterizes the comprehensive probability of the chemical occurrence of abnormal events comprising various monitoring risks, namely the probability of the chemical occurrence of the abnormal events caused by the comprehensive action of various monitoring risks on the chemical, at the moment, the chemical may possibly generate one or a plurality of types of abnormal events of the monitoring risks, and the abnormal events may possibly synchronously occur.
According to the method, a process of sudden accumulation of dangerous chemicals from risks to abnormal events is regarded as a process of increasing risk values, and fluctuation conditions of situation factors such as combustible gas, environmental meteorological parameters and tank farm technological parameters are utilized to build a dangerous chemical parameter risk model and a system risk model, the dangerous chemical parameter risk values and the system risk values are respectively predicted and judged to identify dangerous chemical risk situations, two overall risk situation researching and judging mechanisms are provided, early warning is achieved when the system risk probability is larger than a set threshold value, prediction of abnormality of various risk parameters is achieved, prediction of overall risk is achieved, namely prediction of system risk is achieved, specific probability can be calculated, and accordingly prediction early warning of parameter level and system level is achieved.
The invention also introduces a multivariable LSTM algorithm to predict the risk entropy time sequence of the dangerous chemical system, establishes a multivariable LSTM-based system risk entropy prediction model, searches the optimal parameter combination of the LSTM model by using a cross verification algorithm, and has good generalization capability and higher prediction accuracy by the prediction model obtained by Gaussian radial basis functions and cross verification.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. It is therefore intended that the following claims be interpreted as including the preferred embodiments and all such alterations and modifications as fall within the scope of the invention.
It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.
Claims (10)
1. A method for predicting a chemical risk type anomaly, comprising the steps of:
step 1: obtaining different kinds of monitoring risk data of the chemical and taking the monitoring risk data as an observed value, wherein the monitoring risk data comprises time information;
step 2: giving weight to each kind of monitoring risk data, calculating the predicted value of each kind of monitoring risk data at the current moment according to the weight of each kind of monitoring risk data, the predicted value and the observed value of the same kind of monitoring risk data at the previous moment, thereby obtaining the predicted value of each kind of monitoring risk data,
and comparing the predicted value and the observed value of the same class at the same moment, judging whether the monitoring risk data of the same class at the corresponding moment is abnormal according to the comparison result, and if so, judging that the chemical product is about to generate abnormal events of the monitoring risks at the moment, thereby judging that the chemical product is about to generate abnormal events of the monitoring risks of different classes at different moments.
2. The method for predicting an abnormal event of a chemical risk type according to claim 1, wherein in step 2, the specific implementation process of weighting each type of monitoring risk data comprises the steps of:
firstly projecting the k-th monitoring risk data in the optimal projection direction according to a projection pursuit cluster analysis method, wherein k is more than or equal to 1, and obtaining a corresponding projection component omega k Then calculating the weight alpha of the kth monitoring risk data according to the following formula k ：
Wherein n represents the total class number of various monitoring risk data, namely the data quantity of the monitoring risk data.
3. The method according to claim 1, wherein in step 2, the predicted values of each type of monitoring risk data at the current time are calculated from the weight of each type of monitoring risk data, the predicted values and the observed values of the same type of monitoring risk data at the previous time, and the average or weighted average is performed on the predicted values of each type of monitoring risk data at the current time, and the average value obtained by the average or weighted average is used as the predicted value of each type of monitoring risk data at the corresponding time.
4. A method of predicting a chemical risk type anomaly as recited in claim 3, wherein the weighted average is weighted average using an exponential smoothing model as shown in equation (1):
5. The method for predicting an abnormal event of a chemical risk type according to claim 4, wherein the predicted value of the kth-class monitoring risk data of the latest i (i.gtoreq.2) times is recursively substituted into the formula (1) to obtain the formula (2):
weighted average is performed according to formula (2).
6. The method for predicting an abnormal event of a chemical risk type according to claim 4 or 5, wherein in step 2, the predicted value and the observed value of the same class at the same time are compared, and whether the monitored risk data at the corresponding time is abnormal is determined according to the comparison result, and if so, it is determined that the abnormal event of the monitored risk of the chemical will occur at the time, and the specific implementation process includes the following steps:
if the deviation between the observed value and the predicted value
And the chemical is considered to have an abnormal event of the kth monitoring risk at the time t, so that the prediction and early warning of the parameter level are realized.
7. The method of predicting a chemical risk type anomaly of claim 6, further comprising, after step 2, the steps of:
step 3: dividing the observation values of various monitoring risk data into two groups, wherein one group is used as a training set, the other group is used as a test set, inputting the training set into the LSTM neural network for training, and testing by using the test set, so that the trained LSTM neural network is obtained;
step 4: inputting the observed values of various monitoring risk data into the LSTM neural network model obtained in the step 3 for prediction to obtain predicted values of various monitoring parameters at various moments, marking the predicted values as first predicted values, replacing the predicted values in the step 2 with the first predicted values,
if the first predicted value X of the kth monitoring parameter at the moment t t,k Observations of kth-class monitoring risk data with time t
Deviation of->
Then consider the kth class of supervisionThe measured parameter will have an abnormal event at time t.
8. The method for predicting an abnormal event of a chemical risk type as set forth in claim 7, wherein the method comprises the steps of
Substituting x into a sigmod activation function, and taking the obtained result as the probability that the kth monitoring parameter will generate an abnormal event at the time t, namely calculating according to the following formula:
f (x) is the probability that the kth monitoring parameter will have an abnormal event at time t.
9. The method for predicting an abnormal event of a chemical risk type according to claim 8, wherein an average value of probabilities of occurrence of the abnormal event of each type of monitoring parameters at the same time or an average value after weighted averaging is used as a probability of occurrence of a system risk at a corresponding time, and the probability of the system risk characterizes a comprehensive probability of occurrence of the abnormal event including each type of monitoring risk.
10. The method of claim 9, wherein the chemical is judged to be at risk of the system when the probability of the risk exceeds a predetermined threshold.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310273471.9A CN116432831A (en) | 2023-03-20 | 2023-03-20 | Prediction method for abnormal event of chemical risk type |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310273471.9A CN116432831A (en) | 2023-03-20 | 2023-03-20 | Prediction method for abnormal event of chemical risk type |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116432831A true CN116432831A (en) | 2023-07-14 |
Family
ID=87078871
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310273471.9A Pending CN116432831A (en) | 2023-03-20 | 2023-03-20 | Prediction method for abnormal event of chemical risk type |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116432831A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117688505A (en) * | 2024-02-04 | 2024-03-12 | 河海大学 | Prediction method and system for vegetation large-range regional negative abnormality |
-
2023
- 2023-03-20 CN CN202310273471.9A patent/CN116432831A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117688505A (en) * | 2024-02-04 | 2024-03-12 | 河海大学 | Prediction method and system for vegetation large-range regional negative abnormality |
| CN117688505B (en) * | 2024-02-04 | 2024-04-19 | 河海大学 | Prediction method and system for vegetation large-range regional negative abnormality |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wu et al. | An intelligent tunnel firefighting system and small-scale demonstration | |
| CN116975378B (en) | Equipment environment monitoring method and system based on big data | |
| CN111210128B (en) | Wetland early warning method based on artificial intelligence and random self-adaptive threshold | |
| CN103745279A (en) | Method and device for monitoring energy consumption abnormity | |
| CN116432831A (en) | Prediction method for abnormal event of chemical risk type | |
| CN111125641A (en) | Atmospheric pollution anomaly identification and cause analysis method, device and storage medium | |
| CN114023399A (en) | Air particulate matter analysis early warning method and device based on artificial intelligence | |
| CN113848494B (en) | On-line monitoring method for temperature of power battery and vehicle-mounted T-BOX | |
| Morabito et al. | Fuzzy neural identification and forecasting techniques to process experimental urban air pollution data | |
| CN111680712B (en) | Method, device and system for predicting oil temperature of transformer based on similar time in day | |
| CN114638114A (en) | Water pollution tracing-oriented emergency monitoring method and tracing method | |
| CN112229952A (en) | Method for detecting toxic and harmful gas in chemical industrial park | |
| CN112085869A (en) | Civil aircraft flight safety analysis method based on flight parameter data | |
| CN111191855B (en) | Water quality abnormal event identification and early warning method based on pipe network multi-element water quality time sequence data | |
| Chen et al. | A bayesian network for the transportation accidents of hazardous materials handling time assessment | |
| CN115358647A (en) | Hydrogen energy industry chain risk monitoring system and monitoring method based on big data | |
| CN112418529B (en) | Outdoor advertisement online collapse prediction method based on LSTM neural network | |
| CN116384255B (en) | Park dangerous situation perception method and system based on multi-source data fusion | |
| Vairoac et al. | A Bayesian belief network for local air quality forecasting | |
| CN117370919A (en) | Remote monitoring system for sewage treatment equipment | |
| CN115296933B (en) | Industrial production data risk level assessment method and system | |
| CN113688506B (en) | Potential atmospheric pollution source identification method based on multi-dimensional data such as micro-station and the like | |
| CN114263855B (en) | Natural gas transportation pipeline leakage prediction method and application thereof | |
| CN114969068A (en) | Method and system for analyzing real-time flow monitoring data of urban pressure pipe network | |
| Wang et al. | Interval estimation of urban ozone level and selection of influential factors by employing automatic relevance determination model |
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 |