CN105844401A - Case-based reasoning-based lake and reservoir water bloom control complex dynamic correlation model and decision making method - Google Patents

Abstract

The invention discloses a case-based reasoning-based lake and reservoir water bloom control complex dynamic correlation model and decision making method and belongs to the environmental engineering technical field. According to the method of the invention, based on case-based reasoning, the architecture framework of a case-based reasoning-based lake and reservoir water bloom decision making system is provided according to a 4R model; a water bloom body is constructed with the protege tool so as to be adopted as the base of case representation; a case library is built; case retrieval is divided into case reasoning and into matching; preliminary screening and case matching are completed, a matched case is obtained; and after being adjusted by domain experts, the matched case is applied to a target case and is saved to the case library. With the method of the invention adopted, the thinking process of expert experience can be imitated, artificial intelligent decision making can be carried out, and therefore, a decision making result more accords with actual lake and reservoir water bloom control situations, the accuracy and reliability of the decision making result can be obviously improved; and decision making efficiency can be improved, and decision making time can be shortened.

Description

Case reasoning-based lake and reservoir water bloom treatment complex dynamic correlation model and decision method

Technical Field

The invention relates to the establishment of a complex dynamic correlation model among a plurality of attributes and the research of a corresponding decision method problem in the lake and reservoir water bloom treatment process based on case-based reasoning, belonging to the technical field of environmental engineering. Specifically, on the basis of deep research on known various lake and reservoir water bloom treatment schemes, by analyzing the characteristics presented in different states in the lake and reservoir water bloom outbreak process and the relationship between the characteristics and a plurality of associated attributes, a feasible treatment method suitable for a target case is solved by case reasoning in a dynamic environment. The system learning and the comprehensive utilization of the past knowledge of the water bloom treatment in the lakes and reservoirs are achieved, the decision-making method of the water bloom treatment in the lakes and reservoirs suitable for the actual condition of the water environment is effectively explored, and the auxiliary effect of an information system on the decision-making of the water bloom treatment is better exerted.

Background

In recent years, the water pollution and ecological problems faced by China are increasingly prominent, and according to statistics, eutrophic lakes account for 69.8% of main lakes in China, and eutrophic reservoirs account for 41.95% of main reservoirs in China. The water bloom phenomenon of partial water areas frequently occurs, so that the normal order of production and life of people is seriously damaged. The water bloom phenomenon is the prominent manifestation of water pollution, algae can release various algal toxins in the metabolic death process, seriously threaten the survival of aquatic organisms, remarkably reduce the species abundance and cause the imbalance of an aquatic ecosystem. Under the promotion of the action plan for preventing and treating water pollution, which is discharged from the state academy in 17.4.2015, the water bloom is taken as one of important water pollution phenomena and is already brought into the national key treatment project, so that the water bloom is effectively prevented and treated, a feasible treatment method is rapidly provided for the outbreak water bloom phenomenon, and the method has very important significance for providing decision bases for water environment protection departments.

At present, research aiming at lake and reservoir water bloom treatment decisions has been accumulated to a certain extent, real and effective treatment cases provide basic experience and characteristic advantages supporting the lake and reservoir water bloom treatment decisions, and if the formed cases can be used for assisting the water bloom decisions in high quality, the existing water bloom treatment methods can be preferably inherited, so that the emergency situations can be responded, and practical and feasible comprehensive treatment measures can be rapidly provided.

As the water bloom outbreak is a complex ecological problem, aiming at the treatment decision process of the water bloom outbreak, which relates to multiple factors such as water quality factors, economic factors, environmental factors and the like, the existing research is difficult to comprehensively consider the multiple factors, and further difficult to carry out quality analysis on an optional biological treatment method, a physical treatment method, a chemical treatment method and other specific treatment methods according to decision attributes, and the fuzzy problem and the random problem which occur in the decision process can not be reasonably treated, so that the deviation of the decision result is large.

The case reasoning emergency decision method is a method in the field of artificial intelligence, is derived from the cognitive activities and the psychological activities of human beings, and can effectively relieve the bottleneck of the human beings in knowledge acquisition. Case reasoning is applied to the field of water bloom, cases with similar attributes can be extracted and summarized, and the cases are applied to a target case and stored in a case library again after the existing decision results are analyzed and adjusted. The case reasoning fully combines the characteristics of the dynamic knowledge base and the incremental learning of the case reasoning, and can accurately judge the serious conditions of the lake and reservoir water bloom after the water bloom outbreak, thereby forming a decision for treating the lake and reservoir water bloom according with the actual condition.

Disclosure of Invention

The invention provides a case-reasoning-based lake and reservoir water bloom treatment complex dynamic correlation model and a decision method, wherein the method is based on case reasoning, analyzes the task of water bloom treatment decision, considers that the influence of the structural dynamic change of a plurality of related attributes on the state of water bloom outbreak is complex, and firstly provides a case-reasoning-based lake and reservoir water bloom decision system architecture framework according to a 4R model for more comprehensive research, wherein the 4R model refers to four modules of case representation, case retrieval, case reuse and case learning. Secondly, constructing a water bloom ontology by using a prot g e tool as a basis for case representation, then collecting cases of water bloom treatment decisions of lakes and reservoirs in recent years, and organizing and summarizing to form a case library; then dividing case retrieval into two parts of case reasoning and case matching, completing case reasoning by setting case screening rules for specific attributes, and calculating contribution degree after classifying the existing attributes according to properties in the case matching part so as to retrieve the optimal matching case of the target case (the case to be managed), and calling a management scheme and a result of the optimal matching case; and the target case is adjusted by the domain expert and then is applied to the target case and is stored in the case base.

The water bloom treatment process follows the natural law and the social law, and based on the purpose of restoring the ecological function health of the water body, specific environmental ecological engineering measures are applied and are divided into five stages, namely a black and odorous stage, an algae type turbid water stable stage, an algae-grass competition stage, a grass-algae coexistence stage and a grass type clear water stable stage, so that the water bloom treatment needs to be repaired for a long time and is a long-term task. In order to define the application range of the method, the relevant lake and reservoir water bloom treatment measures are all aimed at the black and odorous stage of the lake and reservoir water bloom outbreak, and the later ecological restoration is not considered.

The invention provides a case-reasoning-based complex dynamic correlation model and a decision method for governing lake and reservoir water bloom, which mainly comprise the following five steps:

step one, designing a general ontology model for governing decision of water bloom in lakes and reservoirs;

the ontology model can provide a general model in the field, and as a powerful knowledge description means, the ontology model can assist decision support of the lake and reservoir water bloom treatment model, so that a treatment decision case representation is constructed in an ontology mode. The general Ontology model for the lake and reservoir water bloom treatment decision defines quintuple according to the Ontology construction Rules of the framework method and the seven-step method, and the form is P _ Ontology [ < P _ Consortions, P _ relationships, P _ Indviduals, P _ Restriction, P _ Rules >, which are respectively expressed as Concepts, Relations, examples, constraints and Rules related to the lake and reservoir water bloom treatment decision. The body construction tool Prot g e tool is adopted for constructing the lake and reservoir water bloom treatment decision body, and the body language OWL is adopted for coding realization, so that a general body model for the lake and reservoir water bloom treatment decision is formed.

Step two, constructing a lake and reservoir water bloom treatment decision case library;

and constructing a water bloom treatment decision case library by sorting and collecting on the basis of the lake and reservoir water bloom treatment decision body case representation. The case base comprises the outbreak conditions of water blooms in various lakes, reservoirs, rivers and sea areas at home and abroad in recent years in different time periods. And recording the water bloom treatment decision cases in the case library according to attributes, wherein the recorded attribute types are divided into three types of data, namely numerical data, Boolean data and option data, and the data are connected with the database by utilizing a Prot é tool, so that the lake and reservoir water bloom treatment decision case library is formed.

Step three, a decision case reasoning engine for water bloom treatment in lakes and reservoirs;

case retrieval of the lake and reservoir water bloom treatment decision should firstly utilize a rule reasoning method, and the realization of the rule reasoning method adopts a water bloom treatment decision case reasoning engine to obtain a preliminary screening scheme. All indexes of the water bloom treatment decision cases listed in the case representation are regarded as nodes of the inference network, the nodes are mutually related, restriction Slots (Constraint Slots) and condition Slots (conditional Slots) are defined according to the special properties of indexes such as ecological safety, water bloom area and the like, the screening of partial cases can be realized according to whether the attributes in the restriction Slots and the condition Slots meet specific conditions, and the inference process is represented by a rule inference method.

Step four, matching water bloom treatment decision cases;

4.1 water bloom governing decision correlation analysis and empowerment calculation based on a complex dynamic network model;

for the attributes in the constraint slot and the condition slot that have played a role in step three, the contribution weight is set to 0; and carrying out dynamic network association analysis and empowerment calculation on other attributes expressed in the water bloom treatment decision case ontology, wherein the dynamic empowerment process provided by the invention adopts a complex network modeling method. Firstly, a complex dynamic network model of a water bloom treatment decision process is constructed, the complex dynamic network model is divided into a total network model and a sub-network model according to the relation of decision attributes, characteristic parameters and an importance evaluation matrix of the complex dynamic network model are calculated, and an optimization criticality model of a construction node is utilized to represent dynamic weight in a case matching process.

4.2 case matching based on the intuitive fuzzy rough set;

according to the dynamic weight, other attributes (which means that the attribute with contribution degree of 0 is excluded) of the case to be managed are respectively matched with the cases preliminarily screened in the case base, each attribute is respectively matched with the attribute of each case in the case base according to three types of numerical type, Boolean type and option type, if the attribute values are the same according to the actual attribute values of the cases, the contribution degree is 1, otherwise, the contribution degree is 0. And the numerical contribution degree is calculated by introducing an intuitive fuzzy rough set method to calculate the similarity of each attribute, an index operator is introduced in the similarity calculation to improve the intuitive fuzzy rough set method, and finally the contribution degree is obtained after weight processing. Setting the contribution degree of the cases in the case base to be zero according to the condition that the attribute values of the cases are missing, forming comprehensive contribution degree by all the contribution degrees in the cases, and taking the case with the maximum comprehensive contribution degree as a standby case of the case to be treated.

Step five, determining, reusing and correcting a water bloom treatment decision case;

for the standby case, if the comprehensive contribution degree exceeds a matching threshold value, the standby case is taken as a matching case, and a treatment method is directly applied to the lake and reservoir; if the comprehensive contribution degree does not exceed the threshold value, the treatment method of the standby case is not completely matched with the case to be treated, and the applicability of the decision result is uncertain, the decision result needs to consult domain experts, and the domain experts directly give treatment suggestions to form a final application case. Finally, after the application cases are subjected to decision application practice, the application cases, the treatment mode and the treatment effect are stored in a case library, so that case reference is provided for a water bloom treatment method of the lake and reservoir to be treated later.

The invention has the advantages that:

1. the invention provides and constructs a general ontology model for lake and reservoir water bloom treatment decision-making for the first time, summarizes important concepts, relations and rules in lake and reservoir water bloom outbreak aiming at the process of water bloom treatment decision-making, forms a basic framework of water bloom treatment decision-making, and makes the way of constructing cases of water bloom in the expression form of ontology more clear and clearer.

2. The invention collects and organizes the outbreak situation of the lake and reservoir water bloom of the water bodies at home and abroad in recent years to form a case reservoir, the actual situation of the outbreak of the lake and reservoir water bloom is relatively comprehensive, the index coverage range is relatively wide, the invention not only comprehensively summarizes the relevant cases of the lake and reservoir water bloom treatment, but also provides reference for the decision of the lake and reservoir water bloom comprehensive treatment. The acquired water bloom treatment measures of the lakes and reservoirs are more in line with the requirements of factors such as complex environment, economy and the like, and the decision result is more applicable.

3. Limiting grooves and condition grooves are defined in the lake and reservoir water bloom treatment decision case reasoning engine, a plurality of decision modes can be preliminarily screened, and the operation process of case similarity with large calculation amount is reduced as much as possible, so that the decision efficiency is improved, and the calculation time is shortened.

4. Because the situation in the lake and reservoir water bloom treatment is complex and changeable, the invention provides that each attribute is weighted by adopting the correlation analysis of a complex dynamic network when the cases are matched, which is beneficial to weighting the change according to the relation between the attributes and different situations of each case, thereby more objectively representing the influence of each attribute on the decision result. In addition, the water bloom treatment decision correlation analysis of a complex dynamic network is introduced, the self importance degree formula of the node is optimized, and the case retrieval precision is also improved.

5. The invention provides a method for calculating contribution degrees in the case matching process to replace the conventional similarity calculation, and can solve the attribute missing situation in the case collecting and sorting process. Therefore, if the case added into the case base cannot avoid the phenomenon of attribute value loss, the influence caused by the loss of the attribute can be eliminated by calculating the contribution degree instead of the similarity, and meanwhile, the threshold value of the contribution degree is set to avoid the extreme situation, so that the decision result is more in line with the actual water bloom management situation of the lake and the reservoir.

6. The invention utilizes an intuitive fuzzy rough set method to process numerical attributes, and introduces an intuitive operator to balance the calculation of uncertainty, thereby effectively improving the calculation precision of contribution and ensuring the accuracy of a matching result.

7. The case reuse and case learning in the invention have the characteristics of self-adaptation and self-learning, the adjustment of the expert opinions to the alternative decisions can improve the accuracy and reliability of the decision result, and the case learning can continuously enlarge the scale of the establishment of the case base, thereby realizing the automatic adjustment of the case base.

8. The lake and reservoir water bloom treatment decision method based on case reasoning can simulate the thinking process of expert experience to carry out artificial intelligent decision, and obviously improves the accuracy and reliability of decision results.

Drawings

FIG. 1 is a flow chart of a complex dynamic correlation model and a decision method for governing lake and reservoir water bloom based on case-based reasoning provided by the invention.

FIG. 2 is a case-based reasoning model for water bloom management decisions.

FIG. 3 is a generic ontology representation of lake reservoir water bloom remediation decisions.

FIG. 4 is a flow chart of case reasoning rules for water bloom management decisions.

FIG. 5 is a water bloom treatment decision complex dynamic network model.

FIG. 6 is a complex dynamic network model of a water bloom outbreak scenario.

Fig. 7 is a water quality parameter complex dynamic network model.

FIG. 8 is an economic factor complex dynamic network model.

FIG. 9 is a complex dynamic network model of a human context environment.

FIG. 10 is a natural environment complex dynamic network model.

Detailed Description

The present invention will be described in further detail below with reference to the accompanying drawings.

The invention provides a case-reasoning-based complex dynamic correlation model and a decision method for governing lake and reservoir water bloom, which have the flow shown in figure 1 and specifically comprise the following steps:

step one, designing a general ontology model for governing decision of water bloom in lakes and reservoirs;

case reasoning is an important part in the field of artificial intelligence, and records experience and knowledge of past problems in a case mode, simulates human reasoning modes and thinking processes, and can provide a solution for current water bloom treatment. The lake and reservoir water bloom treatment decision model based on case reasoning is divided into four parts, namely water bloom treatment case representation, case retrieval, case reuse and case learning, which are shown in figure 2. The case representation adopts a method for constructing a general ontology model to extract characteristic factors of a new case to form a case library; case retrieval utilizes a mode of combining rule reasoning and nearest neighbor strategies to rapidly retrieve cases with guiding function from a water bloom treatment decision case library as matching cases; and case reuse adopts the latest similar case solution of the matched case, adjusts the solution by means of expert opinions according to actual requirements, continuously revises the case for implementing the decision result, applies the final scheme adjusted by the expert to the water bloom treatment process to be solved, learns the case as the application case, and stores the application case to a case library. Aiming at the complex system problem of water bloom treatment, the case reasoning method has the characteristics of association, comparison and induction, and is wide in applicable field, high in problem solving capability and easy to accept results.

The decision case ontology for governing the water bloom in the lake or the reservoir is a concept system for classifying and describing the experience and knowledge of governing the water bloom, so that the commonly recognized concepts in the water environment field can be determined according to the experience and knowledge of governing the water bloom in the lake or the reservoir, and the mutual relation among the concepts is clearly defined. In the invention, quintuple P _ Ontology of the general Ontology model for defining lake and reservoir water bloom governing decision is defined as < P _ Concepts, P _ relationships, P _ Indvials, P _ Restriction and P _ Rules > to describe the water bloom case.

(1) P _ Concepts represents a concept set in the field of water bloom treatment in lakes and reservoirs, including cyanobacterial water blooms, treatment plans, and the like.

(2)P_Relations＝{r(c₁,c₂) And expressing the relationship among concepts and attributes in the decision-making field of lake and reservoir water bloom treatment. Wherein c is₁,c₂Belonging to a set of concepts, r being concept c₁,c₂Direct relationship names. For example, eutrophic contamination is a type of contamination.

(3) P _ Individuals ═ { Individuals } represents a set of instances of the lake and reservoir water bloom management areas, for example, the water bloom outbreak in the 8 th-month tai lake basin in 2010 is an example of the water bloom outbreak.

(4) P _ Restriction represents a constraint condition between concepts in the field of lake and reservoir water bloom treatment, for example, if no secondary pollution is required in the treatment process, the acid-base neutralization method is not acceptable.

(5) P _ Rules ═ { rule } represents a set of Rules in the field of water bloom treatment in lakes and reservoirs, for example: biological methods are under the control scheme.

And (3) constructing a general ontology model for the lake and reservoir water bloom treatment by using a Prot g é tool, wherein a relational graph of a part of the ontology model is shown in FIG. 3, a rectangle represents a concept entity, a label on a connecting line represents a relation, and an ellipse represents an attribute. "Kind of" indicates "membership" relationship, "Related to" indicates "relevance" relationship, "Component of" indicates "composition" relationship, and "Subclas" indicates "Subclass" relationship. The encoding representation is performed by using an OWL ontology language, and the representation of the class is shown as follows.

<owl:Class rdf:ID＝"Characteristic parameter"/＞

＜owl:Class rdf:ID＝"COD"/＞

＜rdfs subClassOf＞

＜owl:Class rdf:about＝"#Characteristic parameter"/＞

＜/rdfs:subClassOf＞

＜/owl:Class＞

Step two, constructing a lake and reservoir water bloom treatment decision case library;

the case of the lake reservoir water bloom treatment decision corresponds to the general water bloom treatment ontology model, and on the basis of the concept and the class of the general water bloom treatment ontology, the attributes of the water bloom treatment decision case (the case for short) in the case reservoir are divided into case basic information, water quality parameters, natural environment, human environment, water bloom outbreak scene, economic factors, treatment conditions and the like. The case basic information comprises water body names, outbreak time and the like, the water quality parameters comprise total phosphorus, total nitrogen and the like, the natural environment comprises air temperature, humidity and the like, the human environment comprises population mobility and water body utilization rate, the water bloom outbreak scene comprises surface color, smell and the like, the economic factors comprise water treatment investment, ecological safety, secondary pollution and the like, and the treatment condition comprises prevention measures, treatment effects and the like. The attribute types of the cases are divided into three types, namely a boolean type, an option type and a numerical type, wherein the boolean type attribute comprises ecological safety, secondary pollution and the like, the option type attribute comprises surface color, smell and the like, and the numerical type attribute comprises total phosphorus, total nitrogen and the like.

Step three, a decision case reasoning engine for water bloom treatment in lakes and reservoirs;

the attribute concepts in the case of lake and reservoir water bloom governing decision-making are not isolated from each other, but are organically connected into a unified system by some inherent factor, either tightly or loosely. Therefore, the invention defines the water bloom governing decision ontology network, the nodes on the water bloom governing decision ontology network are the attribute concepts related to the water bloom, namely the attributes of the cases listed in the case library, and the interrelation among the nodes corresponds to the interrelation among the attributes.

In the decision-making body network for governing the water bloom in the lakes and reservoirs, the attribute represented by certain specific nodes has specificity, called as a groove point, and has a direct effect on the selection of the governing method for the water bloom in the lakes and reservoirs. Therefore, according to the nature of these slot points, there are two categories, namely, restriction Slots (Constraint Slots) and conditional Slots (conditional Slots). The limiting slot represents a limiting condition of the attribute on the decision result, and when the attribute value meets the requirement, the selection of the treatment method is screened by using an elimination method according to the limiting slot; and the condition slot represents the action condition of the attribute on the decision result, and when the attribute value meets the requirement, the treatment method directly selects according to the condition slot.

The attributes of the restriction tanks include bloom area, water treatment investment, long-term treatment, secondary pollution and ecological safety, restriction conditions and restriction results of the attributes of the five restriction tanks are sequentially expressed according to the existing research and expert experience summary, as shown in fig. 4, the rules of the restriction tanks are respectively as follows:

(1) if the area of the bloom is more than 30 square kilometers, physical methods such as water flushing, manual aeration, mechanical algae removal and the like are eliminated.

(2) If the investment of water treatment is less than 500 yuan/square meter, physical methods such as water flushing, artificial aeration, mechanical algae removal and the like and an activated carbon adsorption method are eliminated.

(3) If the long-term treatment belongs to the requirement condition, physical methods such as water diversion scouring, manual aeration, mechanical algae removal and the like are eliminated.

(4) If the secondary pollution is forced to be negative, chemical methods such as an acid-base neutralization method and a chemical algae removal method are eliminated.

(5) If ecological safety is compulsorily required, the biological prevention and control technology and the biological algae inhibiting technology of the source nutritive salt are eliminated.

Taking the constraint slot rule (1) as an example, the inference rule can be converted into an inference language as follows.

IF Bloom area>30km²,THEN D21,D22,D23are excluded.

Wherein D21 represents water-diversion scouring, D22 represents artificial aeration, and D23 represents mechanical algae removal method.

The attributes of the condition tanks comprise total phosphorus, total nitrogen, nitrogen-phosphorus ratio, dissolved oxygen, pH and effective time, the condition requirements and action results of the attributes of the six condition tanks are sequentially expressed according to the existing research and expert experience summary, and the rule of the condition tanks is as follows:

(1) if the total phosphorus concentration is more than 2.00mg/L or the total nitrogen concentration is more than 2.00mg/L or the mass ratio of nitrogen to phosphorus is between 32:1 and 64:1, the biological control technology of the source nutritive salt, the chemical sedimentation method, the passivation method and the water hyacinth purification method are selected.

(2) If the concentration of the dissolved oxygen is less than 5mg/L, a water-diversion flushing method and a manual aeration method are selected.

(3) If the pH value is between 7.9 and 8.1, acid-base neutralization and chemical algae removal are selected.

(4) If the quick effect is required, the water flushing, the manual aeration and the mechanical algae removal are selected.

Taking the condition slot rule (1) as an example, the inference rule is converted into an inference language as follows.

IF TP>2.00mg/L OR TN>2.00mg/L,OR 32:1<N/P<64:1,THEN D11,D31,D32andD42are choosed.

Wherein D11 represents biological control technology of source nutritive salt, D31 represents chemical sedimentation method, D32 represents passivation method, and D42 represents purification of water hyacinth.

Step four, matching water bloom treatment decision cases;

1. constructing a complex dynamic network model of the water bloom treatment decision;

the decision-making result of the lake reservoir water bloom treatment can be influenced by multiple aspects such as the basic condition of a water body, water quality parameters, a human environment, economic factors, a natural environment, water bloom outbreak scene and the like, so that the decision-making problem of the water environment is a complex system problem, various influencing factors are abstracted into a complex dynamic network to obtain the key degree for accurately and quantitatively describing the importance degree of decision attributes, and the interaction of all entities in the water environment complex system is reflected through the interaction among nodes.

According to the characteristics of case attributes in the water bloom treatment decision, a complex dynamic network model is constructed and divided into a main network and sub-networks, the nodes of the main network comprise six nodes including water quality parameters, human environments, economic factors, natural environments and water bloom outbreak scenes, and each node of the main network can be subdivided into sub-nodes according to the content of the node, so that the sub-networks are formed. The sub-nodes of the water quality parameters comprise Total Phosphorus (TP), Total Nitrogen (TN), nitrogen-phosphorus ratio (N/P), Dissolved Oxygen (DO), Chemical Oxygen Demand (COD), Biological Oxygen Demand (BOD), pH value (pH), conductivity (EC), chlorophyll a (Chl _ a) and Water Temperature (WT), the sub-nodes of the human environment comprise water body utilization rate, population mobility, agricultural sewage discharge amount, industrial sewage discharge amount and domestic sewage discharge amount, the sub-nodes of the economic factors comprise water treatment investment, long-term treatment, quick response time, ecological safety and secondary pollution, the sub-nodes of the natural environment comprise water body area, Air Temperature (AT), humidity, illumination intensity, geographical position, wind speed, surrounding environment and lake type, the sub-nodes of the water bloom outbreak scene comprise water bloom area, eutrophication level, algae type, surface color, odor type, and Water Temperature (WT) and the sub-nodes of the water bloom outbreak scene comprise water bloom area, The surface properties of the water body.

The total network and the sub-networks are both a network point set v (v ═ v { (v))₁,v₂,…,v_i,…,v_n}) and a network edge set e (e ═ e₁,e₂,…e_j,…,e_mH) wherein v is_iFor identifying the ith node, e_jIs used for identifying the firstj edges, n represents the number of nodes of the network, and m represents the number of edge sets of the network.

2. Calculating characteristic parameters of the complex dynamic network;

node degree representation complex dynamic network with ith node v_iNumber of nodes directly connected, by λ_iIndicating that the node degree includes an in-degree lambda_i ⁺And out degree lambda_i ^-The node degree is the sum of the in-degree and out-degree, and the node degree is lambda_iCan directly reflect the node v_iThe influence on adjacent nodes shows the effect in the whole complex dynamic network, and the undirected network does not distinguish the in-degree and the out-degree and is directly represented by the node degree.

Node distance d_ijThe number of edges, node v, used to represent the shortest path between two nodes_iAnd node v_jThe shortest path between them is d_ijIf d is_ijAnd → infinity indicates that no path exists between the two nodes.

The average difficulty of network information circulation is represented by network efficiency E, and the calculation formula of the network efficiency is shown in formula (1).

$E=\frac{1}{n(n-1)}\underset{i\&NotEqual;j}{\&Sigma;}\frac{1}{d_{ij}}---\left(1\right)$

The network efficiency is used for describing the difficulty degree of information flow transfer in the network, and the larger E is, the better the circulation is. The efficiency of each node can be calculated according to the average difficulty degree of network information circulation, and the calculation formula is shown as a formula (2).

$I_k=\frac{1}{n}\underset{i=1,i\&NotEqual;k}{\overset{n}{\&Sigma;}}\frac{1}{d_{ki}}---\left(2\right)$

Wherein, I_kRepresenting a node v_kEfficiency of d_kiRepresenting a node v_kTo node v_iDistance of nodes between, I_kThe larger the value, the higher the transmission efficiency of the node, and the more important the node is in the network, it can be understood that removing the node will hinder the smooth flow of information.

3. Constructing an importance evaluation matrix;

the mutual association among the nodes can influence the load and the importance of the nodes, and an importance contribution relation topology is established according to the topology mapping of the actual network. The importance contribution relationship of the node comes from the adjacent node, and therefore, the importance contribution relationship can be represented by a mapping matrix of an adjacency matrix, namely a node importance contribution matrix, which is denoted as H_IC。

$H_{IC}=\left[\begin{array}{cccc}1& {\mathrm{\&delta;}}_{12}{\mathrm{\&lambda;}}_2/<k{>}^2& \mathrm{...}& {\mathrm{\&delta;}}_{1n}{\mathrm{\&lambda;}}_n/<k{>}^2\\ {\mathrm{\&delta;}}_{21}{\mathrm{\&lambda;}}_1/<k{>}^2& 1& \mathrm{...}& {\mathrm{\&delta;}}_{2n}{\mathrm{\&lambda;}}_n/<k{>}^2\\ .& .& & .\\ .& .& \mathrm{...}& .\\ .& .& & .\\ {\mathrm{\&delta;}}_{n1}{\mathrm{\&lambda;}}_1/<k{>}^2& {\mathrm{\&delta;}}_{n2}{\mathrm{\&lambda;}}_2/<k{>}^2& \mathrm{...}& 1\end{array}\right]---\left(3\right)$

Wherein,<k > represents the mean value of the undirected network, λ_i/<k>²Representing a node v_iWill directly act on the associated neighbouring node, H_ICIt represents the importance contribution matrix of all nodes to their neighbors._ijThe elements of the complex dynamic network connection matrix represent the connection condition between network nodes, if the node v_iAnd node v_jAre connected with each other, then_ij1, otherwise_ijWhen the value is equal to 0, then_ijReferred to as contribution allocation parameters.

In the undirected network, the node degree does not distinguish the in degree and the out degree, and the lambda is considered when the importance degree of the node per se is considered_i/<k>²However, in the directed network, the node with the out degree attribute has stronger action than the node with the in degree attribute, and the node v in the formula (3) is used for representing the influence of the out degree and the in degree on the importance of the node_iOf self importance λ_i/<k>²By the formulaOptimizing so that the optimized node importance contribution matrix H'_ICAs follows.

${H^{\&prime;}}_{IC}=\left[\begin{array}{cccc}1& {\mathrm{\&delta;}}_{12}{\left({{\mathrm{\&lambda;}}_2}^{{{\mathrm{\&lambda;}}_2}^{-1}/{\mathrm{\&lambda;}}_2}\right)}^2/<k{>}^2& \mathrm{...}& {\mathrm{\&delta;}}_{1n}{\left({{\mathrm{\&lambda;}}_n}^{{{\mathrm{\&lambda;}}_n}^{-1}/{\mathrm{\&lambda;}}_n}\right)}^2/<k{>}^2\\ {\mathrm{\&delta;}}_{21}{\left({{\mathrm{\&lambda;}}_1}^{{{\mathrm{\&lambda;}}_1}^{-1}/{\mathrm{\&lambda;}}_1}\right)}^2/<k{>}^2& 1& \mathrm{...}& {\mathrm{\&delta;}}_{2n}{\left({{\mathrm{\&lambda;}}_n}^{{{\mathrm{\&lambda;}}_n}^{-1}/{\mathrm{\&lambda;}}_n}\right)}^2/<k{>}^2\\ .& .& & .\\ .& .& \mathrm{...}& .\\ .& .& & .\\ {\mathrm{\&delta;}}_{n1}{\left({{\mathrm{\&lambda;}}_1}^{{{\mathrm{\&lambda;}}_1}^{-1}/{\mathrm{\&lambda;}}_1}\right)}^2/<k{>}^2& {\mathrm{\&delta;}}_{n2}{\left({{\mathrm{\&lambda;}}_2}^{{{\mathrm{\&lambda;}}_2}^{-1}/{\mathrm{\&lambda;}}_2}\right)}^2/<k{>}^2& \mathrm{...}& 1\end{array}\right]---\left(4\right)$

However, the importance of the node depends not only on the functions of the adjacent nodes, but also on the transmission efficiency of the node, so the importance of the node is represented by two indexes, namely position information and adjacent information, and the importance evaluation matrix of the node can be represented in a form of formula (5).

$H_E=\left[\begin{array}{cccc}{I}_1& {\mathrm{\&delta;}}_{12}{\left({{\mathrm{\&lambda;}}_2}^{{{\mathrm{\&lambda;}}_2}^{-}/{\mathrm{\&lambda;}}_2}\right)}^2{I}_2/<k{>}^2& \mathrm{...}& {\mathrm{\&delta;}}_{1n}{\left({{\mathrm{\&lambda;}}_n}^{{{\mathrm{\&lambda;}}_n}^{-}/{\mathrm{\&lambda;}}_n}\right)}^2{I}_n/<k{>}^2\\ {\mathrm{\&delta;}}_{21}{\left({{\mathrm{\&lambda;}}_1}^{{{\mathrm{\&lambda;}}_1}^{-}/{\mathrm{\&lambda;}}_1}\right)}^2{I}_1/<k{>}^2& I_2& \mathrm{...}& {\mathrm{\&delta;}}_{2n}{\left({{\mathrm{\&lambda;}}_n}^{{{\mathrm{\&lambda;}}_n}^{-}/{\mathrm{\&lambda;}}_n}\right)}^2{I}_n/<k{>}^2\\ .& .& & .\\ .& .& \mathrm{...}& .\\ .& .& & .\\ {\mathrm{\&delta;}}_{n1}{\left({{\mathrm{\&lambda;}}_1}^{{{\mathrm{\&lambda;}}_1}^{-}/{\mathrm{\&lambda;}}_1}\right)}^2{I}_1/<k{>}^2& {\mathrm{\&delta;}}_{n2}{\left({{\mathrm{\&lambda;}}_2}^{{{\mathrm{\&lambda;}}_2}^{-}/{\mathrm{\&lambda;}}_2}\right)}^2{I}_2/<k{>}^2& \mathrm{...}& I_n\end{array}\right]---\left(5\right)$

Wherein H_EFor importance evaluation matrix, its matrix element H_EijRepresenting a node v_jTo node v_iThe importance contribution value of. The importance of a node needs to be integrated with the global importance and the local importance of the node, and thus is defined as the integration of the efficiency of the node itself and the efficiency of the neighboring nodes, and the formula of the importance of the node is shown in (6).

$C_i=I_i\&times;\underset{j=1,j\&NotEqual;i}{\overset{n}{\&Sigma;}}{\mathrm{\&delta;}}_{ij}{\left({{\mathrm{\&lambda;}}_j}^{{{\mathrm{\&lambda;}}_j}^{-}/{\mathrm{\&lambda;}}_j}\right)}^2{I}_j/<k{>}^2---\left(6\right)$

Wherein, I_iAnd I_jRespectively represent nodes v_iAnd node v_jThe efficiency of (c).

The importance of the node is normalized to obtain the normalized attribute weight

4. Calculating a fuzzy rough set of numerical attributes;

a set of fuzzy rough sets is defined as S,upper and lower approximations in (A) are denoted by X, respectively⁺And X^-Then one intuitive fuzzy rough set (RFIS) a in fuzzy rough set S is as follows.

$A=\{<x_A,u_{A^{-}}\left(x_A\right),u_{A^{+}}\left(x_A\right),{\mathrm{\&gamma;}}_{A^{-}}\left(x_A\right),{\mathrm{\&gamma;}}_{A^{+}}\left(x_A\right)>|\&ForAll;x_A\&Element;X\}---\left(7\right)$

Wherein,lower approximate membership function of AIndicating the degree of negative impact of the case attributes,the influence of the severity of the attribute on the decision process of water bloom treatment in the lakes and reservoirs is shown.Upper approximate membership function of AIndicating the degree of possible negative impact of the case attributes,the severity of the attribute may have an impact on the water bloom remediation decision making process.Lower approximate non-membership function of AIndicating the degree of positive impact of the case attributes,the influence of the high quality degree of the attribute on the water bloom treatment decision process is determined.Upper approximate non-membership function of AIndicating the degree of positive impact possible on the case attributes,it is the impact that a high quality degree of the attributes may have on the water bloom remediation decision making process.

Known, intuitive fuzzy rough set A lower approximate membership functionUpper approximate membership functionLower approximation non-membership functionAnd approximate upper non-membership functionThe following conditions need to be satisfied at the same time.

$\begin{array}{c}{u}_{A^{-}}\left(x_A\right)+\&le;{\mathrm{\&gamma;}}_{A^{-}}\left(x_A\right)\&le;1;u_{A^{+}}\left(x_A\right)+{\mathrm{\&gamma;}}_{A^{+}}\left(x_A\right)\&le;1;\\ u_{A^{-}}\left(x_A\right)\&le;u_{A^{+}}\left(x_A\right);{\mathrm{\&gamma;}}_{A^{-}}\left(x_A\right)\&le;{\mathrm{\&gamma;}}_{A^{+}}\left(x_A\right);\&ForAll;x_A\&Element;X\end{array}---\left(8\right)$

Element x in intuitive fuzzy rough set A_AIs of an intuitive index pi_A(x_A) Is defined as x_AThe measurement of the degree of hesitation of A, but in the case of water bloom treatment, the uncertainty of the influence of the numerical water quality parameters on the water bloom treatment result is large, so that the difference of the similarity is reduced, and therefore, an exponential operator α is introduced_AIndicating a modification to the visual index whenSatisfy 0 ≤ pi_A(x_A) When 1 is established or less, the intuitive index is calculated as follows.

${\mathrm{\&pi;}}_A\left(x_A\right)={\mathrm{\&alpha;}}_A(1-u_{A^{-}}(x_A)-{\mathrm{\&gamma;}}_{A^{-}}(x_A\left)\right)---\left(9\right)$

Wherein, α_AIs an exponential operator, and the value range of the exponential operator is not less than 0.5 and not more than α_A≤1。

For non-null-theory domain X ═ X₁,x₂,…,x_nTwo sets of intuitive blurred roughness, A and B, on A, the intuitive blurred roughness values of A are:

the intuitive blur roughness values for B are:

the similarity calculation for a and B is as follows.

$\begin{array}{c}M(A,B)=M(x_A,x_B)\\ =1-\frac{1}{5}({\mathrm{\&omega;}}_1\left(x_{A,}{x}_B\right)\left|u_{A^{-}}\left(x_A\right)-u_{B^{-}}\left(x_B\right)\right|+{\mathrm{\&omega;}}_2\left(x_{A,}{x}_B\right)\left|u_{A^{+}}\left(x_A\right)-u_{B^{+}}\left(x_B\right)\right|+{\mathrm{\&omega;}}_3\left(x_{A,}{x}_B\right)\left|{\mathrm{\&gamma;}}_{A^{-}}\left(x_A\right)-{\mathrm{\&gamma;}}_{B^{-}}\left(x_B\right)\right|\\ +{\mathrm{\&omega;}}_4\left(x_{A,}{x}_B\right)|{\mathrm{\&gamma;}}_{A^{+}}\left(x_A\right)-{\mathrm{\&gamma;}}_{B^{+}}\left(x_B\right)|+{\mathrm{\&omega;}}_5\left(x_{A,}{x}_B\right)|{\mathrm{\&pi;}}_{A^{-}}\left(x_A\right)-{\mathrm{\&pi;}}_{B^{-}}\left(x_B\right)|)\end{array}---\left(10\right)$

Wherein the weight coefficient omega₁(x_A,x_B),ω₂(x_A,x_B),ω₃(x_A,x_B),ω₄(x_A,x_B),ω₅(x_A,x_B) Can be dynamically determined according to the known attribute information of the specific target case and the source case.

5. Calculating the similarity of the Boolean type and the option type;

for the attributes of the boolean type and the option type in the case base, it is specified that if the options of the boolean type and the options of the option type are the same, the similarity M of the attributes is 1, and if different or matched case attributes are missing, the similarity M is 0.

6. Calculating the comprehensive contribution degree;

according to the calculation methods of the Boolean type, the option type and the numerical type, the similarity of any case attribute can be obtained. And (3) aiming at the definition of similarity, a contribution degree concept in the case matching process is provided, namely in the case matching process, any attribute can contribute to case matching, and if the attribute value is missing, the contribution degree is 0. The calculation formula of the integrated contribution degree is as follows.

$S_{AB}=\underset{i=1}{\overset{m}{\&Sigma;}}{w}_i\&times;\left(\underset{j=1}{\overset{n}{\&Sigma;}}\right(w_{ij}\&times;M_{ij}\left)\right)---\left(11\right)$

Wherein S is_ABRepresents the comprehensive contribution, w, of case A to case B matching_iRepresenting a node v_iWeight of (1), w_ijRepresenting a node v_ijI is 1,2,3 … n, j is 1,2,3 … M, M_ijFor node v in cases A and B_ijThe similarity of (c). And comparing the obtained comprehensive contribution degrees, and selecting the case with the maximum comprehensive contribution degree as a standby case.

Step five, determining, reusing and correcting a water bloom treatment decision case;

comparing the maximum comprehensive contribution degree of the standby case with a matching threshold value, and when the maximum comprehensive contribution degree meets S_ABAnd when the value is more than or equal to 0.5, the standby case can be used as a matching case, and the treatment method is available and can be used in the actual engineering project after being adjusted. If the maximum comprehensive contribution degree does not meet the conditions, the fact that no corresponding case is matched with the case to be managed in the case base is proved, and therefore management decisions should be submitted to field experts so as to guarantee the effectiveness of decision results.

And applying a treatment decision result obtained based on case reasoning to a case to be treated, and recording the reasoning process and the decision effect to a case library to provide reference for a subsequent water bloom treatment decision.

Example 1:

step one, designing a general ontology model for governing decision of water bloom in lakes and reservoirs;

according to the designed decision-making body model for governing the water bloom in the lake and reservoir, taking the water bloom outbreak in the Kunming lake in Yihe garden of Beijing in 2005-8 months as an example, the actual situation of the water bloom outbreak is obtained, as shown in Table 1.

TABLE 1 relevant information of the lake water bloom outbreak case library of Kunming

Meanwhile, on the basis of the decision ontology model for governing the water bloom in the lake and the reservoir, a total network attribute set and a sub-network attribute set are defined as shown below.

The attribute set of the total network is X ═ water bloom outbreak scene, water quality parameter, economic factor, human environment and natural environment }.

The attribute of the water bloom outbreak scene subnetwork is X₁Either the color of the surface or the color of the surface,algae type, eutrophication level, water bloom area, surface character, odor }.

The attribute of the water quality parameter subnetwork is X₂＝{TP，TN，N/P，PH，EC，TW，DO，BOD，COD，chl_a}。

Economic factor sub-network attribute is X₃The effective time, water treatment investment, long-term treatment, ecological safety and secondary pollution.

The attribute of the human context subnetwork is X₄The discharge amount of domestic sewage, the water body utilization rate, the population mobility, the discharge amount of domestic sewage and the discharge amount of industrial sewage.

The property of the natural environment subnetwork is X₅Water area, lake type, ambient environment, air temperature, humidity, illumination intensity, geographical location, wind speed.

Thus X_iI attribute, X, representing the overall network_ijRepresenting the jth attribute of the ith subnet.

Step two, constructing a lake and reservoir water bloom treatment decision case library;

the case library collects water bloom outbreak conditions and treatment conditions of water bodies such as the previous Taihu lake, the Beijing six seas and the Taihu lake, and the case library is constructed according to the form of a general ontology model of water bloom treatment decision, and the information of part of the case library is shown in a table 2.

TABLE 2 Water bloom management case base part case information

Case numbering	Name of lake or reservoir	Burst time of water bloom	Bloom area (Km2)	Eutrophication grade	…
						1	Taihu lake	2015.8.6	183	Moderately rich nutrition	…
2	Taihu lake	2007.5.27	412	Moderately rich nutrition	…
						3	Chaohu lake	2010.7.22	50	Moderately rich nutrition	…
4	Dian pond	2008.7.30	181.09	Severe eutrophication	…
						5	Taihu lake	2007.7.17	80	Moderately rich nutrition	…
6	Dongting lake	2008.7.	10	Moderately rich nutrition	…
						7	Lake on lake mountain	2008.8.26	15	Moderately rich nutrition	…
8	West lake	2010.7	-	Slightly rich in nutrition	…
						9	Basalt lake	2005.9.18	3	Moderately rich nutrition	…
10	Guan bridge lake (Donghu)	2009.8.16	4	Moderately rich nutrition	…
						11	Yanghu	2010.10.19	5.12	Moderately rich nutrition	…
12	South water reservoir	2009.2.6	-	Moderately rich nutrition	…
						…	…	…	…	…	…

Step three, a decision case reasoning engine for water bloom treatment in lakes and reservoirs;

given the situation that the Kunming lake is exposed to water blooms in 2005, the regulation reasoning is firstly carried out on the treatment scheme of the Kunming lake according to the acquired information. Firstly, considering the attribute bloom area of a restriction tank, water treatment investment, long-term treatment, secondary pollution and ecological safety, wherein only the attribute of the secondary pollution is executed, and the secondary pollution is forcibly rejected, so that chemical methods such as an acid-base neutralization method, a chemical algae removal method and the like are eliminated. And secondly, considering attributes of total phosphorus, total nitrogen, a nitrogen-phosphorus ratio, dissolved oxygen, PH and an effect time of the condition tank, integrating actual information of the Kunming lake, and exerting effects of the total nitrogen, the nitrogen-phosphorus ratio and the effect time in the condition tank, so that a chemical sedimentation method, a passivation method, a water hyacinth purification method, a water flushing method, artificial aeration and mechanical algae removal are selected. By integrating the reasoning results of the limiting tank and the condition tank, the final treatment scheme is selected from water diversion and flushing, manual aeration and mechanical algae removal methods. The method comprises 5 cases in total, wherein the cases comprise case 2 Taihu partial watershed water bloom treatment in 5 months in 2007, case 3 Chaohu water bloom treatment in 2010, case 5 Taihu water bloom treatment in 7 months in 2007, case 8 West lake water bloom treatment in 7 months in 2010, and case 11 Yanghu (south khaki wetland) water bloom treatment in 10 months in 2010 in the case library.

Step four, matching the water bloom treatment decision case;

4.1 water bloom governing decision correlation analysis and empowerment calculation based on a complex dynamic network;

1. constructing complex dynamic network models for governance decisions

And establishing a complex dynamic network according to the internal association of the water bloom treatment decision attributes, which are respectively shown in fig. 6 to 10.

2. Computing characteristic parameters of complex dynamic networks

Firstly, the weight of each attribute is calculated for the total network model with complicated water bloom treatment decision. Respectively solving lambda according to the definitions of the node degree, the node distance and the contribution degree distribution parameter_i、d_ijAnd_ijas shown below

λ＝[2 3 1 2 2](12)

$d=\left[\begin{array}{ccccc}0& 1& 3& 2& 1\\ 1& 0& 2& 1& 1\\ 3& 2& 0& 1& 3\\ 2& 1& 1& 0& 2\\ 1& 1& 3& 2& 0\end{array}\right]---\left(13\right)$

$\mathrm{\&delta;}=\left[\begin{array}{ccccc}1& 1& 0& 0& 1\\ 1& 1& 0& 1& 1\\ 0& 0& 1& 1& 0\\ 0& 1& 1& 1& 0\\ 1& 1& 0& 0& 1\end{array}\right]---\left(14\right)$

The node efficiency can be calculated from equation (2) as follows.

$I=\left[\begin{array}{ccccc}\frac{7}{5}& 1& \frac{9}{5}& \frac{6}{5}& \frac{7}{5}\end{array}\right]---\left(15\right)$

3. Construction of importance evaluation matrix

And (3) calculating an importance evaluation matrix of the node according to a formula (5), in the practical application of the complex dynamic network, calculating the optimized node importance of the discriminable directed network according to the corresponding in-degree and out-degree, and equally processing the in-degree and out-degree under the condition that part of the complex dynamic network has no directional influence relation, so that the importance evaluation matrix of the water bloom treatment decision-making complex total network model is as follows.

$H_E=\left[\begin{array}{ccccc}1& 0.75& 0& 0& 0.7\\ 0.7& 1& 0& 0.6& 0.7\\ 0& 0& 1& 0.6& 0\\ 0& 0.75& 0.45& 1& 0\\ 0.75& 0.75& 0& 0& 1\end{array}\right]---\left(16\right)$

Further, the importance C of the node is obtained from equation (6) as [ 2.0321.081.442.03 ], and the weight w of the node is obtained as [ 0.23660.23310.12590.16780.2366 ] after normalization.

The weights of the attributes in the water bloom treatment decision sub-network can be calculated by the same method as follows.

$\begin{array}{c}{w}_1=\left[\begin{array}{cccccc}0.1594& 0.1594& 0.1594& 0.1594& 0.1812& 0.1812\end{array}\right]\\ w_2=\left[\begin{array}{cccccccccc}0.1158& 0.1158& 0.0780& 0.1352& 0.1017& 0.1182& 0.0898& 0.0594& 0.0594& 0.1267\end{array}\right]\\ w_3=\left[\begin{array}{ccccc}0.1259& 0.1678& 0.2331& 0.2366& 0.2366\end{array}\right]\\ w_4=\left[\begin{array}{ccccc}0.1357& 0.2864& 0.1759& 0.2010& 0.2010\end{array}\right]\\ w_5=\left[\begin{array}{cccccccc}0.0284& 0.0933& 0.1532& 0.1683& 0.1901& 0.0710& 0.1373& 0.1583\end{array}\right]\end{array}---\left(17\right)$

4.2 case matching based on the intuitive fuzzy rough set;

1. fuzzy rough set calculation of numerical attributes

Taking total phosphorus as an example, calculating the matching condition of the case to be treated and the screening case. The total phosphorus concentration of Kunming lake to be treated is known to be TP_D63 Total phosphorus concentration in each case of the screening is TP₂＝210,TP₃＝150,TP₅＝130,TP₈＝147,TP₁₁250. According to the positive influence and negative influence of total phosphorus on the water bloom treatment condition given by expert opinions, a total phosphorus positive influence non-membership function gamma (TP) is obtained^-) And negatively affecting the membership function mu (TP)^-) As follows.

$\begin{array}{c}\mathrm{\&gamma;}\left({\mathrm{TP}}^{-}\right)=\left\{\begin{array}{c}0.9,{\mathrm{TP}}_i\&le;25\\ 0.9\exp (-{\left(\left(x_{TP}-25\right)/25\right)}^2),{\mathrm{TP}}_i>25\end{array}\right.\\ \mathrm{\&mu;}\left({\mathrm{TP}}^{-}\right)=\left\{\begin{array}{c}0.9\exp (-{\left(\left(x_{TP}-200\right)/30\right)}^2),{\mathrm{TP}}_i<200\\ 0.9,{\mathrm{TP}}_i\&GreaterEqual;200\end{array}\right.\end{array}---\left(18\right)$

Calculated according to equation (18)Andthe values of the upper approximation are the sum of the lower approximation and the intuitive index by using the formula (9) as the lower approximation of the intuitive fuzzy rough set. Therefore, the intuitionistic fuzzy rough value of total phosphorus TP of the cases to be treated and the screened casesAs shown below.

$\begin{array}{c}{\mathrm{TP}}_D=<\begin{array}{ccccc}0& 0.9107& 0.0893& 2& 0.9107\end{array}>\\ {\mathrm{TP}}_2=<\begin{array}{ccccc}0.9& 1& 0& 0.1& 0.1\end{array}>\\ {\mathrm{TP}}_3=<\begin{array}{ccccc}0.056& 1& 0& 0.944& 0.944\end{array}>\\ {\mathrm{TP}}_5=<\begin{array}{ccccc}0.0039& 1& 0& 0.9961& 0.9961\end{array}>\\ {\mathrm{TP}}_8=<\begin{array}{ccccc}0.0397& 1& 0& 0.9603& 0.9603\end{array}>\\ {\mathrm{TP}}_{11}=<\begin{array}{ccccc}0.9& 1& 0& 0.1& 0.1\end{array}>\end{array}---\left(19\right)$

The similarity between the case to be treated and each case in the case base can be obtained according to the formula (10) as follows.

$\begin{array}{c}M({\mathrm{TP}}_D,{\mathrm{TP}}_2)=0.6043\\ M({\mathrm{TP}}_D,{\mathrm{TP}}_3)=0.9419\\ M({\mathrm{TP}}_D,{\mathrm{TP}}_5)=0.9627\\ M({\mathrm{TP}}_D,{\mathrm{TP}}_8)=0.9484\\ M({\mathrm{TP}}_D,{\mathrm{TP}}_{11})=0.6043\end{array}---\left(20\right)$

The similarity of other numerical attributes is calculated by the same method.

2. Boolean-type and Option-type similarity calculation

And selecting the attribute of the water body type as an example for calculating the similarity between the Boolean type and the option type, wherein M is equal to 1 because the Kunming lake of the case to be treated and the case of the case library are landscape lakes. And if the similarity of the landscape lake and the culture lake is calculated, M is 0.

3. Calculation of the degree of Integrated contribution

According to the calculation method of the formula (11), the comprehensive contribution degrees between the cases to be treated and the screening cases in the case base are sequentially calculated, so the calculation results of the comprehensive contribution degrees are as follows.

S_D2＝0.4384,S_D3＝0.4575,S_D5＝0.4119,S_D8＝0.5498,S_D11＝0.4847 (21)

Comparing the calculation results of the comprehensive contribution degrees, it can be seen that the comprehensive contribution degree of the case 8 is the maximum, so the case 8 is selected as a standby case of the queensland lake to be treated.

Step five, determining, reusing and correcting a water bloom treatment decision case;

the standby case obtained according to the fourth step is the water bloom treatment condition of the West lake in 2010 for 7 months, and the calculation result shows that the maximum contribution degree of the case to the water bloom treatment condition of the Kunming lake is 0.5498, and is greater than the threshold value of 0.5, which indicates that the matching result is acceptable.

According to the scheme for water bloom treatment decision of the case base, the water diversion and flushing in the water bloom is controlled by a physical method in the West lake water bloom treatment, sewage is intercepted through a water regulation project, the exchange of the lake surface layer water body is accelerated, the nutrient substances in the water body are greatly reduced, the increased exchange speed of the surface layer water body can improve the water quality, the content of dissolved oxygen is increased, and the bottom sediment pollutants are released and slowed down, so that the purpose of controlling the water bloom is achieved. The method is simple and quick in effect, but high in cost, and the advantages and the limitations of the method accord with the treatment condition of the Kunming lake, so that the best matching case can be judged to be similar to the treatment condition of the lake reservoir to be treated, and the method can be used for treating the Kunming lake. Meanwhile, a treatment decision scheme for the water bloom outbreak in 7 months of 2010 of the Kunming lake is inquired, a water diversion scouring method in a physical control method is also used, the treatment effect is good, and the result verifies the effectiveness of the water bloom treatment method based on case reasoning.

In addition, the treatment condition of the Kunming lake can be simultaneously stored in the case base, so that the completeness and the coverage of the case base are supplemented, and reference is provided for other treatment conditions in the future.

Claims (4)

1. The case reasoning-based complex dynamic correlation model and decision method for governing lake and reservoir water bloom are characterized in that: the method specifically comprises the following steps of,

step one, designing a general ontology model for governing decision of water bloom in lakes and reservoirs;

defining quintuple according to Ontology construction Rules of a skeleton method and a seven-step method, wherein the quintuple is in the form of P _ Ontology < P _ Consceptions, P _ relationships, P _ Indvidals, P _ Restriction and P _ Rules > and is respectively expressed as Concepts, Relations, examples, constraints and Rules related to lake and reservoir water bloom governing decisions; the method specifically comprises the following steps:

p _ Concepts represents a concept set in the field of water bloom treatment in lakes and reservoirs;

P_Relations＝{r(c₁,c₂) Expressing the relationships among concepts and between the concepts and attributes in the decision-making field of lake and reservoir water bloom treatment; wherein c is₁,c₂Belonging to a set of concepts, r being concept c₁,c₂A direct relationship name;

p _ induviduals ═ { induviduals } represents a set of examples of the lake and reservoir water bloom treatment domain;

p _ recovery ═ { recovery } represents a constraint condition between concepts in the field of lake and reservoir water bloom treatment;

p _ Rules ═ { rule } represents a set of Rules in the field of water bloom treatment in lakes and reservoirs;

step two, constructing a lake and reservoir water bloom treatment decision case library;

recording water bloom treatment decision cases in the case base according to attributes, wherein the cases of the water bloom treatment decision in the lake and reservoir correspond to a water bloom treatment general ontology model, on the basis of the concept and the class of the water bloom treatment general ontology, the attributes of the water bloom treatment decision cases in the case base are divided into case basic information, water quality parameters, natural environment, human environment, water bloom outbreak scenes, economic factors and treatment conditions, the attribute types of the cases are divided into three types of data, namely numerical type, Boolean type and option type, and a Prot gue tool is connected with a database, so that the lake and reservoir water bloom treatment decision case base is formed;

step three, a decision case reasoning engine for water bloom treatment in lakes and reservoirs;

the method of rule reasoning is utilized to realize the preliminary screening of the cases according to whether the attributes in the restriction slot and the condition slot meet specific conditions;

step four, matching water bloom treatment decision cases;

for the attributes in the constraint slot and the condition slot that have played a role in step three, the contribution weight is set to 0; performing dynamic network association analysis and weighting calculation on other attributes expressed in the water bloom treatment decision case ontology, calculating characteristic parameters and an importance evaluation matrix of a complex dynamic network model by adopting a complex dynamic network modeling method in the dynamic weighting process, and representing dynamic weights in the case matching process by utilizing an optimization key degree model of a construction node;

according to the dynamic weight, other attributes of the case to be controlled are respectively matched with the cases preliminarily screened in the case base, the attributes of the cases in the case base are respectively matched according to three types, namely a numerical type, a Boolean type and an option type, the attributes of the Boolean type and the option type are calculated according to the actual attribute values of the cases, and the similarity of the attributes is calculated by introducing an intuitive fuzzy rough set method through the calculation of the numerical contribution; finally, obtaining contribution degree after weight processing; setting the contribution degree of the cases in the case base to be zero according to the condition that the attribute values of the cases are missing, wherein all the contribution degrees in the cases form a comprehensive contribution degree, and obtaining the case with the maximum comprehensive contribution degree as a standby case of the case to be controlled;

step five, determining, reusing and correcting a water bloom treatment decision case;

for the standby case, if the comprehensive contribution degree exceeds a matching threshold value, the standby case is taken as a matching case, and a treatment method is directly applied to the lake and reservoir; if the comprehensive contribution degree does not exceed the threshold value, the treatment method of the standby case is not completely matched with the case to be treated, and the applicability of the decision result is uncertain, the decision result needs to consult domain experts, and the domain experts directly give treatment suggestions to form a final application case; and finally, after the application case is subjected to decision application practice, the application case, the treatment mode and the treatment effect are stored in a case library.

2. The case-based reasoning lake and reservoir water bloom governing complex dynamic correlation model and decision method as claimed in claim 1, wherein: the limiting slot represents a limiting condition of the attribute on the decision result, and when the attribute value meets the requirement, the selection of the treatment method is screened by using an elimination method according to the limiting slot; and the condition slot represents the action condition of the attribute on the decision result, and when the attribute value meets the requirement, the treatment method directly selects according to the condition slot.

3. The case-based reasoning based lake and reservoir water bloom governing complex dynamic correlation model and decision method as claimed in claim 1 or 2, wherein: the attributes of the restriction tank include water bloom area, water treatment investment, long-term treatment, secondary pollution and ecological safety, and the restriction tank rules are respectively:

(1) if the bloom area is more than 30 square kilometers, the physical methods of water diversion scouring, artificial aeration and mechanical algae removal are eliminated;

(2) if the water treatment investment is less than 500 yuan/square meter, the physical methods of water diversion and washing, manual aeration, mechanical algae removal and activated carbon adsorption are eliminated;

(3) if the long-term treatment belongs to the requirement condition, physical methods of water diversion scouring, manual aeration and mechanical algae removal are eliminated;

(4) if the secondary pollution is forcibly denied, chemical methods of acid-base neutralization and chemical algae removal are eliminated;

(5) if the ecological safety is compulsorily required, the biological prevention and control technology and the biological algae inhibiting technology of the source nutritive salt are eliminated;

the attributes of the condition tank comprise total phosphorus, total nitrogen, nitrogen-phosphorus ratio, dissolved oxygen, pH and effective time, and the rule of the condition tank is as follows:

(1) if the total phosphorus concentration is more than 2.00mg/L or the total nitrogen concentration is more than 2.00mg/L or the mass ratio of nitrogen to phosphorus is between 32:1 and 64:1, the biological control technology of source nutritive salt, the chemical sedimentation method, the passivation method and the water hyacinth purification method are selected;

(2) if the concentration of the dissolved oxygen is less than 5mg/L, a water-diversion flushing method and a manual aeration method are selected;

(3) if the pH value is between 7.9 and 8.1, an acid-base neutralization method and a chemical algae removal method are selected;

(4) if the quick effect is required, the water flushing, the manual aeration and the mechanical algae removal are selected.

4. The case-based reasoning lake and reservoir water bloom governing complex dynamic correlation model and decision method as claimed in claim 1, wherein: the concrete implementation process of the step four is as follows:

(1) constructing a complex dynamic network model of the water bloom treatment decision;

the complex dynamic network model is divided into a main network and sub-networks, the nodes of the main network comprise six nodes including water quality parameters, human environment, economic factors, natural environment and water bloom outbreak scene, and each node of the main network can be subdivided into sub-nodes according to the content of the node, so that the sub-networks are formed; the sub-nodes of the water quality parameters comprise total phosphorus, total nitrogen, nitrogen-phosphorus ratio, dissolved oxygen, chemical oxygen demand, biological oxygen demand, pH value, conductivity, chlorophyll a and water temperature, the sub-nodes of the human environment comprise water body utilization rate, population mobility, agricultural sewage discharge, industrial sewage discharge and domestic sewage discharge, the sub-nodes of the economic factors comprise water treatment investment, long-term treatment, quick response time, ecological safety and secondary pollution, the sub-nodes of the natural environment comprise water body area, air temperature, humidity, illumination intensity, geographical position, wind speed, surrounding environment and lake type, and the sub-nodes of the water bloom outbreak scene comprise water bloom area, eutrophication level, algae type, surface color, smell and water body surface shape;

the total network and the sub-networks are both a network point set v (v ═ v { (v))₁,v₂,…,v_i,…,v_n}) and a network edge set e (e ═ e₁,e₂,…e_j,…,e_mH) wherein v is_iFor identifying the ith node, e_jThe method is used for identifying the jth edge, n represents the number of nodes of the network, and m represents the number of edge sets of the network;

(2) calculating characteristic parameters of the complex dynamic network;

lambda for node degree_iIndicating, node degree includes degree of entryDegree of harmonyNode v_iAnd node v_jDistance of nodes between with d_ijExpressed by formula (1), the calculation formula of the network efficiency E is as follows:

$E=\frac{1}{n(n-1)}\underset{i\&NotEqual;j}{\&Sigma;}\frac{1}{d_{ij}}---\left(1\right)$

and calculating the efficiency of each node according to the network efficiency:

$I_k=\frac{1}{n}\underset{i=1,i\&NotEqual;k}{\overset{n}{\&Sigma;}}\frac{1}{d_{ki}}---\left(2\right)$

wherein, I_kRepresenting a node v_kThe efficiency of (c);

(3) constructing an importance evaluation matrix;

the node importance contribution matrix is denoted as H_IC：

$H_{IC}=\left[\begin{array}{cccc}1& {\mathrm{\&delta;}}_{12}{\mathrm{\&lambda;}}_2/<k{>}^2& \mathrm{...}& {\mathrm{\&delta;}}_{1n}{\mathrm{\&lambda;}}_n/<k{>}^2\\ {\mathrm{\&delta;}}_{21}{\mathrm{\&lambda;}}_1/<k{>}^2& 1& \mathrm{...}& {\mathrm{\&delta;}}_{2n}{\mathrm{\&lambda;}}_n/<k{>}^2\\ .& .& & .\\ .& .& \mathrm{...}& .\\ .& .& & .\\ {\mathrm{\&delta;}}_{n1}{\mathrm{\&lambda;}}_1/<k{>}^2& {\mathrm{\&delta;}}_{n2}{\mathrm{\&lambda;}}_2/<k{>}^2& \mathrm{...}& 1\end{array}\right]---\left(3\right)$

Wherein,<k>representing the mean value of the undirected network, λ_i/<k>²Representing a node v_iThe degree of importance of the system itself,_ijthe elements of the complex dynamic network connection matrix represent the connection condition between network nodes, if the node v_iAnd node v_jAre connected with each other, then_ij1, otherwise_ij＝0；

In the undirected network, the node degree does not distinguish the in degree and the out degree, and the node v in the formula (3)_iOf self importance λ_i/<k>²By the formulaOptimizing so that the optimized node importance contribution matrix H'_ICAs follows:

${H^{\&prime;}}_{IC}=\left[\begin{array}{cccc}1& {\mathrm{\&delta;}}_{12}{\left({{\mathrm{\&lambda;}}_2}^{{{\mathrm{\&lambda;}}_2}^{-}/{\mathrm{\&lambda;}}_2}\right)}^2/<k{>}^2& \mathrm{...}& {\mathrm{\&delta;}}_{1n}{\left({{\mathrm{\&lambda;}}_n}^{{{\mathrm{\&lambda;}}_n}^{-}/{\mathrm{\&lambda;}}_n}\right)}^2/<k{>}^2\\ {\mathrm{\&delta;}}_{21}{\left({{\mathrm{\&lambda;}}_1}^{{{\mathrm{\&lambda;}}_1}^{-}/{\mathrm{\&lambda;}}_1}\right)}^2/<k{>}^2& 1& \mathrm{...}& {\mathrm{\&delta;}}_{2n}{\left({{\mathrm{\&lambda;}}_n}^{{{\mathrm{\&lambda;}}_n}^{-}/{\mathrm{\&lambda;}}_n}\right)}^2/<k{>}^2\\ .& .& & .\\ .& .& \mathrm{...}& .\\ .& .& & .\\ {\mathrm{\&delta;}}_{n1}{\left({{\mathrm{\&lambda;}}_1}^{{{\mathrm{\&lambda;}}_1}^{-}/{\mathrm{\&lambda;}}_1}\right)}^2/<k{>}^2& {\mathrm{\&delta;}}_{n2}{\left({{\mathrm{\&lambda;}}_2}^{{{\mathrm{\&lambda;}}_2}^{-}/{\mathrm{\&lambda;}}_2}\right)}^2/<k{>}^2& \mathrm{...}& 1\end{array}\right]---\left(4\right)$

the importance evaluation matrix of the node is expressed as the form of formula (5):

$H_E=\left[\begin{array}{cccc}{I}_1& {\mathrm{\&delta;}}_{12}{\left({{\mathrm{\&lambda;}}_2}^{{{\mathrm{\&lambda;}}_2}^{-}/{\mathrm{\&lambda;}}_2}\right)}^2{I}_2/<k{>}^2& \mathrm{...}& {\mathrm{\&delta;}}_{1n}{\left({{\mathrm{\&lambda;}}_n}^{{{\mathrm{\&lambda;}}_n}^{-}/{\mathrm{\&lambda;}}_n}\right)}^2{I}_n/<k{>}^2\\ {\mathrm{\&delta;}}_{21}{\left({{\mathrm{\&lambda;}}_1}^{{{\mathrm{\&lambda;}}_1}^{-}/{\mathrm{\&lambda;}}_1}\right)}^2{I}_1/<k{>}^2& I_2& \mathrm{...}& {\mathrm{\&delta;}}_{2n}{\left({{\mathrm{\&lambda;}}_n}^{{{\mathrm{\&lambda;}}_n}^{-}/{\mathrm{\&lambda;}}_n}\right)}^2{I}_n/<k{>}^2\\ .& .& & .\\ .& .& \mathrm{...}& .\\ .& .& & .\\ {\mathrm{\&delta;}}_{n1}{\left({{\mathrm{\&lambda;}}_1}^{{{\mathrm{\&lambda;}}_1}^{-}/{\mathrm{\&lambda;}}_1}\right)}^2{I}_1/<k{>}^2& {\mathrm{\&delta;}}_{n2}{\left({{\mathrm{\&lambda;}}_2}^{{{\mathrm{\&lambda;}}_2}^{-}/{\mathrm{\&lambda;}}_2}\right)}^2{I}_2/<k{>}^2& \mathrm{...}& I_n\end{array}\right]---\left(5\right)$

the formula of the importance of the node is shown as (6):

$C_i=I_i\&times;\underset{j=1,j\&NotEqual;i}{\overset{n}{\&Sigma;}}{\mathrm{\&delta;}}_{ij}{\left({{\mathrm{\&lambda;}}_j}^{{{\mathrm{\&lambda;}}_j}^{-}/{\mathrm{\&lambda;}}_j}\right)}^2{I}_j/<k{>}^2---\left(6\right)$

wherein, I_iAnd I_jRespectively represent nodes v_iAnd node v_jThe efficiency of (c);

normalizing the importance of the nodes to obtain normalized attribute weight

(4) Calculating a fuzzy rough set of numerical attributes;

a set of fuzzy rough sets is defined as S,upper and lower approximations in (A) are denoted by X, respectively⁺And X^-Then, one intuitive fuzzy rough set a in the fuzzy rough set S is as follows:

$A=\{<x_A,u_{A^{-}}\left(x_A\right),u_{A^{+}}\left(x_A\right),{\mathrm{\&gamma;}}_{A^{-}}\left(x_A\right),{\mathrm{\&gamma;}}_{A^{+}}\left(x_A\right)>|\&ForAll;x_A\&Element;X\}---\left(7\right)$

wherein,lower approximate membership function of AIndicating the degree of negative impact of the case attributes,the influence of the severity of the attribute on the decision process of water bloom treatment in the lakes and reservoirs is determined;upper approximate membership function of AIndicating the degree of possible negative impact of the case attributes,the severity of the attribute may affect the water bloom treatment decision process;lower approximate non-membership function of AIndicating the degree of positive impact of the case attributes,the influence of the high quality degree of the attribute on the water bloom treatment decision process is determined;upper approximate non-membership function of AIndicating the degree of positive impact possible on the case attributes,the influence of the high quality degree of the attribute on the water bloom treatment decision process is possible;

lower approximate membership function of intuitive fuzzy rough set AUpper approximate membership functionLower approximation non-membership functionAnd approximate upper non-membership functionThe following conditions need to be satisfied simultaneously:

$\begin{array}{c}{u}_{A^{-}}\left(x_A\right)+{\mathrm{\&gamma;}}_{A^{-}}\left(x_A\right)\&le;1;u_{A^{+}}\left(x_A\right)+{\mathrm{\&gamma;}}_{A^{+}}\left(x_A\right)\&le;1;\\ u_{A^{-}}\left(x_A\right)\&le;u_{A^{+}}\left(x_A\right);{\mathrm{\&gamma;}}_{A^{-}}\left(x_A\right)\&le;{\mathrm{\&gamma;}}_{A^{+}}\left(x_A\right);\&ForAll;x_A\&Element;X\end{array}---\left(8\right)$

element x in intuitive fuzzy rough set A_AIs of an intuitive index pi_A(x_A) Is defined as x_AFor a measure of the degree of hesitation of a, the intuitive index is calculated as follows:

${\mathrm{\&pi;}}_A=\left(x_A\right)={\mathrm{\&alpha;}}_A(1-u_{A^{-}}\left(x_A\right)-{\mathrm{\&gamma;}}_{A^{-}}\left(x_A\right))---\left(9\right)$

wherein, α_AIs an exponential operator, and the value range of the exponential operator is not less than 0.5 and not more than α_A≤1；

For non-null-theory domain X ═ X₁,x₂,…,x_nTwo sets of intuitive blurred roughness, A and B, on A, the intuitive blurred roughness values of A are:

the intuitive blur roughness values for B are:

the similarity of A and B is calculated as follows:

$\begin{array}{c}M(A,B)=M(x_A,x_B)\\ =1-\frac{1}{5}({\mathrm{\&omega;}}_1\left(x_{A,}{x}_B\right)|u_{A^{-}}\left(x_A\right)-u_{B^{-}}\left(x_B\right)|+{\mathrm{\&omega;}}_2\left(x_{A,}{x}_B\right)|u_{A^{+}}\left(x_A\right)-u_{B^{+}}\left(x_B\right)|+{\mathrm{\&omega;}}_3\left(x_{A,}{x}_B\right)|{\mathrm{\&gamma;}}_{A^{-}}\left(x_A\right)-{\mathrm{\&gamma;}}_{B^{-}}\left(x_B\right)|\\ +{\mathrm{\&omega;}}_4\left(x_{A,}{x}_B\right)|{\mathrm{\&gamma;}}_{A^{+}}\left(x_A\right)-{\mathrm{\&gamma;}}_{B^{+}}\left(x_B\right)|+{\mathrm{\&omega;}}_5\left(x_{A,}{x}_B\right)|{\mathrm{\&pi;}}_{A^{-}}\left(x_A\right)-{\mathrm{\&pi;}}_{B^{-}}\left(x_B\right)|)\end{array}---\left(10\right)$

wherein the weight coefficient omega₁(x_A,x_B),ω₂(x_A,x_B),ω₃(x_A,x_B),ω₄(x_A,x_B),ω₅(x_A,x_B) Basis toolDynamically determining the known attribute information of the target case of the body and the cases in the case base;

(5) calculating the similarity of the Boolean type and the option type;

for the attributes of the Boolean type and the option type in the case library, the similarity M of the attributes is 1 if the Boolean type option and the option type option are the same, and the similarity M is 0 if different or matched case attributes are missing;

(6) calculating the comprehensive contribution degree;

$S_{AB}=\underset{i=1}{\overset{m}{\&Sigma;}}{w}_i\&times;\left(\underset{j=1}{\overset{n}{\&Sigma;}}\right(w_{ij}\&times;M_{ij}\left)\right)---\left(11\right)$

wherein S is_ABRepresents the comprehensive contribution, w, of case A to case B matching_iRepresenting a node v_iWeight of (1), w_ijRepresenting a node v_ijI is 1,2,3 … n, j is 1,2,3 … M, M_ijFor node v in cases A and B_ijThe similarity of (2); and comparing the obtained comprehensive contribution degrees, and selecting the case with the maximum comprehensive contribution degree as a standby case.