CN109890002B - Method and device for predicting environmental risk based on knowledge reasoning - Google Patents
Method and device for predicting environmental risk based on knowledge reasoning Download PDFInfo
- Publication number
- CN109890002B CN109890002B CN201910007843.7A CN201910007843A CN109890002B CN 109890002 B CN109890002 B CN 109890002B CN 201910007843 A CN201910007843 A CN 201910007843A CN 109890002 B CN109890002 B CN 109890002B
- Authority
- CN
- China
- Prior art keywords
- node
- environment information
- risk
- current
- determining
- 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.)
- Active
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
- Alarm Systems (AREA)
Abstract
The invention discloses a method and a device for predicting environmental risk based on knowledge reasoning, wherein the method comprises the following steps: determining the prediction environment information of the environment where the current node is located based on historical environment information and a predetermined time sequence prediction algorithm; determining a probability distribution function of the environment information obtained at each moment corresponding to the historical environment information; determining deviation environment information according to the current environment information and the probability distribution function; and determining the risk degree of the environment where the current node is located according to the current environment information, the prediction environment information and the deviation environment information. The wireless node has the capability of sensing and reasoning aiming at the occurrence of a certain abnormal event, the processing process is executed by each node, the processing efficiency is higher, a master controller does not need to wait for the return of a calculation result, and the mode of processing by the node can greatly reduce the number of messages responsible in a network and improve the system performance.
Description
Technical Field
The invention relates to the field of communication, in particular to a method and a device for predicting environmental risk based on knowledge reasoning.
Background
At present, the internet of things monitoring environment of a wireless sensor network comprises a plurality of wireless nodes (namely sensors) which can sense the occurrence of certain abnormal events (such as fire), most of mainstream knowledge reasoning systems based on the wireless sensor network are centralized systems, and a master controller intensively converges and processes the environment collected by each sensor to complete knowledge reasoning, so that the knowledge reasoning efficiency is difficult to ensure, and the number of messages loaded in the network is overlarge; in addition, the mainstream system at present relies heavily on the type I fuzzy logic model, because the model has obvious flexibility defect when being applied to dynamic environment or when the construction of the fuzzy logic rule contains the non-determinism caused by local knowledge.
Disclosure of Invention
The invention provides a method and a device for predicting environmental risk based on knowledge reasoning, which are used for solving the following problems in the prior art: the existing knowledge reasoning system is used for intensively converging and processing the environment collected by each sensor by a master controller to finish knowledge reasoning, so that the knowledge reasoning efficiency is difficult to ensure, and the quantity of messages loaded in a network is overlarge.
In order to solve the above technical problem, in one aspect, the present invention provides a method for predicting environmental risk based on knowledge inference, including: determining the prediction environment information of the environment where the current node is located based on historical environment information and a predetermined time sequence prediction algorithm; determining a probability distribution function of the environment information obtained at each moment corresponding to the historical environment information; determining deviation environment information according to the current environment information and the probability distribution function; and determining the risk degree of the environment where the current node is located according to the current environment information, the prediction environment information and the deviation environment information.
Optionally, after determining the risk of the environment in which the current node is located according to the current environment information, the predicted environment information, and the deviation environment information, the method further includes: detecting whether the risk exceeds a predetermined risk threshold; and sending the risk degree and/or a warning message to a central node of a predetermined cluster where the current node is located under the condition that the risk degree exceeds the predetermined risk degree threshold value.
Optionally, after sending the risk and/or the warning message to the central node of the predetermined cluster where the current node is located, the method further includes: and the central node sends the risk degree of each node in the preset cluster to a master controller.
Optionally, determining deviation environment information according to the current environment information and the probability distribution function includes: determining expected environmental information of the current time point according to the probability distribution function; and determining the absolute value of the difference value between the expected environmental information and the current environmental information as the deviation environmental information.
Optionally, after determining the risk of the environment in which the current node is located according to the current environment information, the predicted environment information, and the deviation environment information, the method further includes: and the current node sends the risk degree of the current node to the adjacent node.
Optionally, the method further includes: calculating the risk degree of each node in the preset cluster according to a preset time interval; and determining the node with the highest risk as a new central node in all the risks.
Optionally, the method further includes: each node sets a selection probability according to the risk degree of the node, wherein the selection probability indicates the probability of the node becoming a central node; and determining the node with the maximum selection probability as a new central node in all the selection probabilities.
In another aspect, the present invention further provides an apparatus for predicting environmental risk based on knowledge inference, including: the first determining module is used for determining the prediction environment information of the environment where the current node is located based on the historical environment information and a predetermined time sequence prediction algorithm; a second determining module, configured to determine a probability distribution function of the environment information obtained at each time corresponding to the historical environment information; the third determining module is used for determining deviation environment information according to the current environment information and the probability distribution function; and the execution module is used for determining the risk degree of the environment where the current node is located according to the current environment information, the prediction environment information and the deviation environment information.
Optionally, the method further includes: the detection module is used for detecting whether the risk degree exceeds a preset risk degree threshold value; and the sending module is used for sending the risk and/or the warning message to the central node of the preset cluster where the current node is located under the condition that the risk exceeds the preset risk threshold.
Optionally, the third determining module is specifically configured to: determining expected environmental information of the current time point according to the probability distribution function; and determining the absolute value of the difference value between the expected environmental information and the current environmental information as the deviation environmental information.
The wireless node of the embodiment of the invention has the capability of sensing and reasoning for the occurrence of a certain abnormal event, the risk degree of the environment where the current node is located is determined according to the current environment information, the prediction environment information and the deviation environment information, the processing process is executed by each node, the processing efficiency is higher, the total controller is not required to wait for returning a calculation result, and the processing mode of the node can greatly reduce the number of messages responsible in the network and improve the system performance.
Drawings
FIG. 1 is a flow chart of a method of predicting environmental risk based on knowledge-based reasoning in a first embodiment of the invention;
FIG. 2 is a schematic diagram of a second embodiment of the present invention, illustrating an apparatus for predicting environmental risk based on knowledge inference;
FIG. 3 is a schematic diagram of another configuration of the apparatus for predicting environmental risk based on knowledge-based reasoning in the second embodiment of the present invention;
fig. 4 is a schematic system architecture of a wireless sensor network according to a third embodiment of the present invention.
Detailed Description
In order to solve the following problems in the prior art: the prior instruction reasoning system completes knowledge reasoning by intensively converging and processing the environment acquired by each sensor through a master controller, thereby not only being difficult to ensure the knowledge reasoning efficiency, but also causing the overlarge message quantity loaded in a network; the invention provides a method and a device for predicting environmental risk based on knowledge reasoning, which are further described in detail in the following by combining the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the invention.
The first embodiment of the present invention provides a method for predicting environmental risk based on knowledge inference, the flow of the method is shown in fig. 1, and the method includes steps S101 to S104:
s101, determining the prediction environment information of the environment where the current node is located based on historical environment information and a predetermined time sequence prediction algorithm;
s102, determining a probability distribution function of the environment information obtained at each moment corresponding to the historical environment information;
s103, determining deviation environment information according to the current environment information and the probability distribution function;
and S104, determining the risk degree of the environment where the current node is located according to the current environment information, the prediction environment information and the deviation environment information.
The wireless node of the embodiment of the invention has the capability of sensing and reasoning for the occurrence of a certain abnormal event, the risk degree of the environment where the current node is located is determined according to the current environment information, the prediction environment information and the deviation environment information, the processing process is executed by each node, the processing efficiency is higher, the total controller is not required to wait for returning a calculation result, and the processing mode of the node can greatly reduce the number of messages responsible in the network and improve the system performance.
After determining the risk of the environment where the current node is located according to the current environment information, the predicted environment information and the deviation environment information, whether the risk exceeds a preset risk threshold value needs to be detected; if the risk degree exceeds a preset risk degree threshold value, indicating that a larger risk hidden danger exists in the current environment, and needing to send the risk degree and/or an alarm message to a central node of a preset cluster where the current node is located; if the risk does not exceed the preset risk threshold, the risk and/or the alarm message may not be reported to the central node, and the current node only sends the risk of the current node to the adjacent node, so that the adjacent node can subsequently achieve consensus of the risk condition with the current node.
After the current node sends the risk degree and/or the warning message to the central node of the predetermined cluster where the current node is located, the central node can send the risk degree of each node in the predetermined cluster to the master controller at regular time. Through the interaction mode of the central node of the preset cluster and the general controller, the general controller can know the risk degree of each node, the load condition of the general controller is greatly reduced, and the system performance is improved.
In the process of determining the deviation environment information according to the current environment information and the probability distribution function, the expected environment information of the current time point may be determined according to the probability distribution function, and then the expected environment information and the current environment information are subtracted, and the absolute value of the difference is determined as the deviation environment information.
In the embodiment of the invention, the central node of each preset cluster can be changed in real time. For example, the risk of each node in the predetermined class cluster is counted according to a predetermined time interval, and the node with the highest risk is determined as a new central node in all the risk. For another example, each node sets a selection probability according to the risk degree of itself, wherein the selection probability indicates the probability that itself becomes the central node, and the node with the highest selection probability is determined as a new central node in all the selection probabilities. According to the embodiment of the invention, the node in the most dangerous condition at present can interact with the master controller in a mode of adjusting the central node according to the degree of danger, so that more attention can be paid to the master controller.
The second embodiment of the present invention provides an apparatus for predicting environmental risk based on knowledge inference, which can be disposed in each node including a central node in a predetermined class cluster, and the structure of the apparatus is schematically shown in fig. 2, and the apparatus includes:
a first determining module 10, configured to determine, based on historical environment information and a predetermined time sequence prediction algorithm, prediction environment information of an environment where a current node is located; a second determining module 20, coupled to the first determining module 10, for determining a probability distribution function of the obtained environment information at each time corresponding to the historical environment information; a third determining module 30, coupled to the second determining module 20, for determining deviation environment information according to the current environment information and the probability distribution function; and the executing module 40 is coupled with the third determining module 30 and is used for determining the risk of the environment where the current node is located according to the current environment information, the predicted environment information and the deviation environment information.
The wireless node of the embodiment of the invention has the capability of sensing and reasoning for the occurrence of a certain abnormal event, the risk degree of the environment where the current node is located is determined according to the current environment information, the prediction environment information and the deviation environment information, the processing process is executed by each node, the processing efficiency is higher, the total controller is not required to wait for returning a calculation result, and the processing mode of the node can greatly reduce the number of messages responsible in the network and improve the system performance.
The above apparatus may also be as shown in fig. 3, comprising: a detection module 50, coupled to the execution module 40 and the sending module 60, for detecting whether the risk exceeds a predetermined risk threshold; a sending module 60, configured to send the risk and/or the warning message to the central node of the predetermined cluster where the current node is located when the risk exceeds the predetermined risk threshold.
Since the central node of the embodiment of the present invention is changeable, the sending module 60 of each node can also be used to send the risk of each node in the predetermined cluster to the master controller. Through the interaction mode of the central node of the preset cluster and the general controller, the general controller can know the risk degree of each node, the load condition of the general controller is greatly reduced, and the system performance is improved.
The sending module 60 may also send its own risk to the neighboring nodes. The current node sends the risk degree of the current node to the adjacent node, so that the adjacent node can subsequently achieve consensus of the risk condition with the current node.
In a specific implementation, the sending functions may also be set to functions of different modules, not all of which are set in the sending module 60, and those skilled in the art may set the functions according to actual situations, which is not described herein again.
The third determining module is specifically configured to: determining expected environment information of the current time point according to the probability distribution function; and determining the absolute value of the difference value between the expected environment information and the current environment information as deviation environment information.
The device can also comprise an adjusting module, wherein the adjusting module is used for counting the risk degree of each node in the preset cluster according to a preset time interval, and determining the node with the maximum risk degree as a new central node in all the risk degrees; or, the node with the highest selection probability is determined as a new central node in all the selection probabilities according to the selection probabilities of the risk degrees set by the nodes, wherein the selection probability indicates the probability that the node becomes the central node. According to the embodiment of the invention, the node in the most dangerous condition at present can interact with the master controller in a mode of adjusting the central node according to the degree of danger, so that more attention can be paid to the master controller.
In view of the shortcomings in the prior art, a third embodiment of the present invention is to provide a distributed knowledge inference technology (i.e., a process of performing prediction through data) in a wireless sensor network environment, so as to implement distributed abnormal event (e.g., fire) inference based on local environment knowledge, and significantly reduce message load in the wireless sensor network on the premise of improving instantaneity and accuracy of the inference process. In order to achieve the above objects and other related objects, the present invention provides a first distributed knowledge inference scheme based on local fuzzy logic and knowledge-driven clustering, i.e., a method for predicting environmental risk based on knowledge inference and a system applying the method. The system comprises a plurality of nodes (sensors), each node observes the same abnormal event (taking fire as an example) and infers whether the abnormal event occurs or not by receiving environmental information data. The method and the system call the probability degree of the abnormal event occurrence, which is inferred by each node, as the risk degree.
The system integrates the current environment information, the predicted environment information and the deviation environment information of a certain node, so that each node needs to be comprehensively considered: (1) whether obvious deviation exists between the current environment information and the predicted environment information or not; (2) the degree of deviation of the current environmental information from the statistical distribution pattern of the node to the current position. The former uses short-term knowledge to make inference, and the latter uses long-term knowledge to make statistical rules of inference nodes. The fusion of the short-term knowledge and the long-term knowledge can generate a more complex and fine reasoning result aiming at a certain event.
The architecture of the system is schematically shown in fig. 4, in this embodiment, a cluster in a wireless sensor network and a master controller are used as an example for description, and the process for predicting the environmental risk based on knowledge reasoning includes the following main steps:
step 1, predicting the localized environment information.
Based on the historical environmental information, a time series prediction algorithm is used to predict short-term upcoming environmental information (i.e., predicted environmental information).
And 2, learning the increment of the probability distribution function.
Based on the historical environmental information, a probability distribution function of the environmental information obtained in the past at each time is incrementally learned. The probability distribution function of the environment information of a certain node can measure the deviation between the current environment information and the past environment information, and further generate the deviation environment information.
And 3, reasoning the risk.
And fusing the output of the localized environment information prediction process (step 1) and the output of the probability distribution function incremental learning process (step 2) based on a II type fuzzy logic model, wherein the goal is to generate a risk degree for each node. The risk of a node provides local reasoning for the monitored abnormal events based on the current environment information of the node, the deviation of the current environment information from an expected value (extracted from the probability distribution function), and the predicted environment information. When the risk degree exceeds a certain preset threshold value, the node confirms the occurrence of a certain abnormal event and then alarms to the cluster-like central node.
And 4, automatically clustering knowledge driving nodes.
And finishing a knowledge-driven node clustering process based on the neighbor nodes, wherein the process is used for enhancing the local reasoning effect and reducing the network environment load. The danger degree value generated by a certain node is only transmitted to the neighbor node of the node to further enhance the local knowledge of the neighbor node, and the danger degree transmission process stimulates the automatic clustering process of the node. The class clusters derived by the process contain nodes that hold a similar "view (i.e., local context information inference result)" for some exceptional event. The invention executes a multi-round knowledge driving node automatic clustering process, and after each round of clustering process is finished, each cluster center node sends the aggregated local knowledge (including the knowledge inference result of the cluster center node) to a master controller.
The above method is described in detail below with reference to two specific examples.
In the first embodiment of the present invention, in step 1, the linear time prediction algorithm used is the levinson durbin algorithm. For a node, the algorithm weights and sums the historical environmental information values of the node to generate predicted environmental information. Due to the requirement of the monitoring environment of the internet of things on the instantaneity of signal processing, a prediction result needs to be generated in the shortest time, and the linear time sequence prediction algorithm selected by the invention can control the time overhead to be in the linear complexity.
In the first embodiment of the present invention, in step 2, the incremental learning algorithm used is the Kernel sensitivity Estimator algorithm. The algorithm is able to generate an implicit distribution of context information describing the given node's past each time instance. Through incremental learning of the probability distribution function, the node can judge whether the current environment information has a significant deviation from the expected value of the environment information so far, namely whether the instantaneous current environment information does not accord with the inherent statistical distribution mode (namely the probability distribution function) of the node, and the difference between the current environment information and the expected value of the environment information is called deviation environment information.
In the first embodiment of the present invention, in step 3, the current environment information, the predicted environment information and the deviation environment information are fused to complete inference on a certain abnormal event, and generate a risk (the higher the risk value, the higher the risk level). The fusion process is based on fuzzy inference rules, and the invention designs and realizes a II type fuzzy logic model. For the application scenario of fire alarm, under the guidance of the field expert experience, the fuzzy inference rule (as shown in table 1) designed by the invention uses three logic terms: low (value close to 0), medium (value close to 0.5), high (value close to 1). Wherein each term corresponds to a numerical interval; the current environment information, the predicted environment information and the deviation environment information are mapped to corresponding logical terms through a mapping function.
TABLE 1
In the first embodiment of the present invention, in step 4, the knowledge-driven clustering process refers to a process of generating different clusters for all nodes based on the risk of the nodes. Member nodes in the class cluster hold a similar view of whether an exception event occurred. The invention predefines a plurality of clustering time intervals, and a knowledge-driven clustering process is carried out in each clustering time interval. In each class cluster, one node is selected to be a class cluster central node, and other nodes are selected to be non-class cluster central nodes (i.e. class cluster member nodes). The cluster center node is responsible for aggregating the danger degree information of the cluster member nodes and then communicating with the master controller. Thus, under this communication strategy of the present invention, the number of messages streamed throughout the network is greatly reduced, as it is not necessary for each node to communicate with the overall controller. The basic idea of the selection process of the cluster-like central node is that if the generated risk of a certain node is higher than that of its neighbor nodes, the node is considered as the cluster-like central node. Once a certain node is determined to be a cluster-like central node, the cluster-like member nodes send respective danger degree messages to the cluster-like central node.
The automatic clustering process of the knowledge-driven nodes is described as follows:
4-1, initially randomly selecting some nodes as cluster-like central nodes;
4-2, dynamically changing cluster center nodes in each iteration;
4-3, when the number of iterative processes (or the number of messages exchanged) reaches a preset threshold, the cluster center selection process terminates.
Wherein, the dynamic change cluster center node process (step 4-2) is described as follows: for each node, the selection process requires a series of iterations; at each iteration, the node sends and receives risk messages to and from its neighbor nodes. Before a certain node starts a selection process, a selection probability of becoming a cluster-like central node is configured for the node, and the selection probability is obtained by selecting a maximum value from the global minimum selection probability and the risk degree value of the node. For a certain node, selecting the node with the probability higher than the self selection probability as a cluster-like central node by comparing the self selection probability with the selection probabilities of other nodes: the node with relatively high selection probability sends a selection probability message to the neighbor and becomes a cluster-like central node; and the node with the smaller selection probability also sends a selection probability message to the neighbor, and if the received messages from other nodes indicate that the selection probability of other nodes is higher than that of the node, the node is determined to be a member node of the cluster.
In summary, the distributed knowledge inference method and system of the invention effectively fuses the current environment information, the predicted environment information and the deviation environment information through the local fuzzy logic and the knowledge-driven clustering, can generate a more complex and more precise knowledge inference result aiming at a certain event, remarkably reduces the number of messages carried by the wireless sensor network environment, and improves the instantaneity and accuracy of knowledge inference. Therefore, the invention effectively overcomes various defects in the prior art and has high industrial utilization value.
As described above, the distributed knowledge inference method and system of the present invention have the following beneficial effects: the fusion of the short-term knowledge and the long-term knowledge can generate a more complex and fine reasoning result aiming at an abnormal event. The invention reduces the number of messages borne by the wireless sensor network environment, solves the problem of low processing efficiency of the mainstream centralized knowledge reasoning system, and improves the instantaneity and accuracy of knowledge reasoning; by dynamically changing the cluster center nodes, the service life of the wireless sensor network can be prolonged, and the energy consumption of the knowledge reasoning process and the message transmission process is further balanced.
In a second embodiment of the present invention, in step 1, the time-series prediction algorithm used is a time prediction model (i.e. LSTM model), and the main steps include: (1) setting LSTM model parameters (an activation function, an activation function of a full-connection artificial neural network for receiving LSTM output, rejection rate of each layer of network nodes, an error calculation mode, an iterative update mode of weight parameters and the like); (2) the LSTM model is trained using the training set data and the prediction is performed using the trained model.
In the second embodiment of the present invention, in step 2, the incremental learning algorithm used is the Kernel sensitivity Estimator algorithm. The algorithm is able to generate an implicit distribution of context information describing the given node's past each time instance. Through incremental learning of the probability distribution function, the node can judge whether the current environmental information has a significant deviation from the expected value of the environmental information so far, that is, whether the current environmental information at the moment does not conform to the inherent statistical distribution pattern (i.e., the probability distribution function) of the node, wherein the difference between the current environmental information and the expected value of the environmental information is called deviation environmental information.
In the second embodiment of the present invention, in step 3, the current environment information, the predicted environment information and the deviation environment information are fused to complete inference on a certain abnormal event, and generate a risk (the higher the risk value, the higher the risk level). The fusion process is based on fuzzy inference rules, and the invention designs and realizes a II type fuzzy logic model. For the application scenario of fire alarm, under the guidance of the field expert experience, the fuzzy inference rule (as shown in table 1) designed by the invention uses three logic terms: low (value close to 0), medium (value close to 0.5), high (value close to 1). Wherein each term corresponds to a numerical interval; the current environment information, the predicted environment information and the deviation environment information are mapped to corresponding logical terms through a mapping function.
In the second embodiment of the present invention, in step 4, the knowledge-driven clustering process refers to a process of generating different clusters for all nodes based on the risk of the nodes. Member nodes in the class cluster hold a similar view of whether an exception event occurred. The invention predefines a plurality of clustering time periods, and a knowledge-driven clustering process is carried out in each clustering time period. In each class cluster, one node is selected to be a class cluster central node, and other nodes are selected to be non-class cluster central nodes (i.e. class cluster member nodes). The cluster center node is responsible for aggregating the danger degree information of the cluster member nodes and then communicating with the master controller. Thus, under this communication strategy of the present invention, the number of messages streamed throughout the network is greatly reduced, as it is not necessary for each node to communicate with the overall controller. The basic idea of the selection process of the cluster-like central node is that if the generated risk of a certain node is higher than that of its neighbor nodes, the node is considered as the cluster-like central node. Once a certain node is determined to be a cluster-like central node, the cluster-like member nodes send respective danger degree messages to the cluster-like central node.
The automatic clustering process of the knowledge-driven nodes is described as follows:
4-1, initially randomly selecting some nodes as cluster-like central nodes;
4-2, dynamically changing cluster center nodes in each iteration;
4-3, when the number of iterative processes (or the number of messages exchanged) reaches a preset threshold, the cluster center selection process terminates.
Wherein, the dynamic change cluster center node process (step 4-2) is described as follows: for each node, the selection process requires a series of iterations; at each iteration, the node sends and receives risk messages to and from its neighbor nodes. For a certain node, comparing the risk of the node with the risk of other nodes, selecting the node with the risk higher than the risk of the node as a cluster-like central node, wherein the node with the highest risk becomes the cluster-like central node: the node with relatively high risk sends a risk message to the neighbor and becomes a cluster-like central node; and the node with lower risk sends a risk message to the neighbor, and if the received messages from other nodes indicate that the risk of other nodes is higher than that of the node, the node is determined to be a member node of the cluster. In a certain iteration, a certain node determines whether to become a cluster-like central node or not according to the risk degree.
The wireless nodes of the monitoring environment of the Internet of things in the embodiment of the invention can sense and reason the occurrence of a certain abnormal event (such as a fire), each node has sensing and calculating capabilities of reasoning localization knowledge (such as the occurrence of the fire and confidence), and transmits a localization reasoning result to the network environment under the condition of no manual intervention, namely, the localization reasoning result is transmitted to a neighbor node or a centralized information processing system (namely, a master controller).
The embodiment of the invention fuses the current environment information, the predicted environment information and the deviation environment information, fuses the short-term knowledge and the long-term knowledge, and can generate a more complex and more precise reasoning result aiming at a certain event; by adopting the II-type fuzzy logic model, the defects of flexibility and accuracy generated when the I-type fuzzy logic model models the non-deterministic local knowledge are avoided; by adopting a knowledge-driven node automatic clustering strategy and introducing a selection probability mechanism, the network load can be reduced and the modeling capability of the non-determinacy is improved; the method has wide applicable range and can be widely applied to the interconnection environment of the intelligent equipment.
Although the preferred embodiments of the present invention have been disclosed for illustrative purposes, those skilled in the art will appreciate that various modifications, additions and substitutions are possible, and that the scope of the present invention is not limited to the embodiments disclosed.
Claims (8)
1. A method for predicting environmental risk based on knowledge inference, comprising:
determining the prediction environment information of the environment where the current node is located based on historical environment information and a predetermined time sequence prediction algorithm;
determining a probability distribution function of the environment information obtained at each moment corresponding to the historical environment information;
determining deviation environment information according to the current environment information and the probability distribution function;
determining the risk degree of the environment where the current node is located according to the current environment information, the prediction environment information and the deviation environment information;
the method further comprises the following steps:
counting the risk degree of each node in the preset cluster according to a preset time interval;
determining the node with the maximum risk degree as a new central node in all the risk degrees;
or determining the node with the maximum selection probability as a new central node in all the selection probabilities according to the selection probabilities of the risk degrees set by each node; wherein the selection probability indicates a probability of itself becoming a central node; and the central node sends the risk degree of each node in the preset cluster to the master controller at regular time.
2. The method of claim 1, wherein after determining the risk of the environment in which the current node is located based on the current environment information, the predicted environment information, and the biased environment information, further comprising: detecting whether the risk exceeds a predetermined risk threshold;
and sending the risk degree and/or a warning message to a central node of a preset cluster where the current node is located under the condition that the risk degree exceeds the preset risk degree threshold value.
3. The method according to claim 2, wherein after sending the risk and/or warning message to the central node of the predetermined cluster class in which the current node is located, further comprising: and the central node sends the risk degree of each node in the preset cluster to a master controller.
4. The method of claim 1, wherein determining deviation context information based on current context information and the probability distribution function comprises: determining expected environment information of the current time point according to the probability distribution function;
and determining the absolute value of the difference value between the expected environmental information and the current environmental information as the deviation environmental information.
5. The method of claim 1, wherein after determining the risk of the environment in which the current node is located based on the current environment information, the predicted environment information, and the biased environment information, further comprising: and the current node sends the risk degree of the current node to the adjacent node.
6. An apparatus for predicting environmental risk based on knowledge inference, comprising:
the first determining module is used for determining the prediction environment information of the environment where the current node is located based on the historical environment information and a predetermined time sequence prediction algorithm;
a second determining module, configured to determine a probability distribution function of the environment information obtained at each time corresponding to the historical environment information;
the third determining module is used for determining deviation environment information according to the current environment information and the probability distribution function;
the execution module is used for determining the risk degree of the environment where the current node is located according to the current environment information, the prediction environment information and the deviation environment information;
the adjusting module is used for counting the risk degree of each node in the preset cluster according to a preset time interval;
determining the node with the maximum risk degree as a new central node in all the risk degrees;
or, the node with the highest selection probability is determined as a new central node in all the selection probabilities according to the selection probability of the risk degree set by each node, wherein the selection probability indicates the probability of the node becoming the central node;
and the central node sends the risk degree of each node in the preset cluster to the master controller at regular time.
7. The apparatus of claim 6, further comprising: the detection module is used for detecting whether the risk degree exceeds a preset risk degree threshold value;
and the sending module is used for sending the risk and/or the warning message to the central node of the preset cluster where the current node is located under the condition that the risk exceeds the preset risk threshold.
8. The apparatus of claim 6 or 7, wherein the third determining module is specifically configured to: determining expected environmental information of the current time point according to the probability distribution function; and determining the absolute value of the difference value between the expected environmental information and the current environmental information as the deviation environmental information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910007843.7A CN109890002B (en) | 2019-01-04 | 2019-01-04 | Method and device for predicting environmental risk based on knowledge reasoning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910007843.7A CN109890002B (en) | 2019-01-04 | 2019-01-04 | Method and device for predicting environmental risk based on knowledge reasoning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109890002A CN109890002A (en) | 2019-06-14 |
| CN109890002B true CN109890002B (en) | 2022-08-16 |
Family
ID=66925571
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910007843.7A Active CN109890002B (en) | 2019-01-04 | 2019-01-04 | Method and device for predicting environmental risk based on knowledge reasoning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109890002B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110766140A (en) * | 2019-10-22 | 2020-02-07 | 电子科技大学 | XGboost and LSTM-based multi-sensor real-time threat level classification method |
| CN111325347B (en) * | 2020-02-19 | 2023-04-11 | 山东大学 | Automatic danger early warning description generation method based on interpretable visual reasoning model |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104469875A (en) * | 2014-11-26 | 2015-03-25 | 北京邮电大学 | Prediction-based target tracking method and system in wireless sensor network |
| CN107071848A (en) * | 2017-05-27 | 2017-08-18 | 河南交通职业技术学院 | A kind of target tracking algorism based on cluster structured reduction energy consumption |
| WO2018098754A1 (en) * | 2016-11-30 | 2018-06-07 | 深圳天珑无线科技有限公司 | Cluster head selection method for distributed network, node and system |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3109801A1 (en) * | 2015-06-26 | 2016-12-28 | National University of Ireland, Galway | Data analysis and event detection method and system |
| CN106302487B (en) * | 2016-08-22 | 2019-08-09 | 中国农业大学 | Agriculture internet of things data throat floater real-time detection processing method and processing device |
| CN106872657B (en) * | 2017-01-05 | 2018-12-14 | 河海大学 | A kind of multivariable water quality parameter time series data accident detection method |
| CN107092582B (en) * | 2017-03-31 | 2021-04-27 | 江苏方天电力技术有限公司 | Online abnormal value detection and confidence evaluation method based on residual posterior |
| CN108052528B (en) * | 2017-11-09 | 2019-11-26 | 华中科技大学 | A kind of storage equipment timing classification method for early warning |
| CN108089962A (en) * | 2017-11-13 | 2018-05-29 | 北京奇艺世纪科技有限公司 | A kind of method for detecting abnormality, device and electronic equipment |
-
2019
- 2019-01-04 CN CN201910007843.7A patent/CN109890002B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104469875A (en) * | 2014-11-26 | 2015-03-25 | 北京邮电大学 | Prediction-based target tracking method and system in wireless sensor network |
| WO2018098754A1 (en) * | 2016-11-30 | 2018-06-07 | 深圳天珑无线科技有限公司 | Cluster head selection method for distributed network, node and system |
| CN107071848A (en) * | 2017-05-27 | 2017-08-18 | 河南交通职业技术学院 | A kind of target tracking algorism based on cluster structured reduction energy consumption |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109890002A (en) | 2019-06-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108418841B (en) | Next-generation key message infrastructure network Security Situation Awareness Systems based on AI | |
| US6389331B1 (en) | Technique for monitoring performance of a facility management system | |
| CN108197845A (en) | A kind of monitoring method of the transaction Indexes Abnormality based on deep learning model LSTM | |
| CN109890002B (en) | Method and device for predicting environmental risk based on knowledge reasoning | |
| KR101994465B1 (en) | Method for the computer-assisted monitoring of the operation of a technical system, particularly of an electrical energy-generating installation | |
| CN109977621A (en) | A kind of air-conditioning failure prediction method based on deep learning | |
| CN109194534B (en) | Scheduling and management method for Internet of things equipment group | |
| CN101907685A (en) | Method for checking correctness of operating parameters of generator set capable of adapting to change of operating conditions | |
| CN113341813B (en) | Urban gas medium-low pressure pipe network detection method and system | |
| CN113033772A (en) | Multi-equipment state monitoring method based on federal learning | |
| CN104615122A (en) | Industrial control signal detection system and detection method | |
| KR101468142B1 (en) | Prediction method of plant health status, and a computer-readable storage medium having program to perform the same | |
| CN110298741A (en) | A kind of Financial Fraud risk recognition system | |
| Jose | Dynamic fuzzy cognitive maps for the supervision of multiagent systems | |
| Dai et al. | Performance adjustable event-triggered synchronization policies to nonlinear multiagent systems | |
| Goller et al. | Towards a continuous assessment of stability in (self-) adaptation behaviour | |
| CN108459991B (en) | Method for obtaining equipment reliability data | |
| Voropai et al. | A suite of intelligent tools for early detection and prevention of blackouts in power interconnections | |
| CN110889646A (en) | Method for checking correctness of operating parameters of generator set adapting to working condition change | |
| CN111210361B (en) | Power communication network routing planning method based on reliability prediction and particle swarm optimization | |
| KR101856552B1 (en) | Wireless sensor network system with enhanced event detection efficiency | |
| CN114897196B (en) | Operation management method, equipment and medium for office building water supply network | |
| CN111382946A (en) | Autonomous evaluation method and system for health state of equipment and industrial internet equipment | |
| Hamid et al. | A reliable structural health monitoring protocol using wireless sensor networks | |
| CN112883581B (en) | Hydropower station production control processing method and system |
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 |