Marine ecological risk digital twin method based on internet of things

Abstract

The application discloses a marine ecological risk digital twin method based on the Internet of Things, and relates to the technical field of digital twin, comprising the following steps: firstly, converting physical and chemical environmental parameters and biological response indexes into characteristic sequences about time; then, using an improved vine Copula joint distribution model to perform one-way matching across time axes on the two types of characteristic sequences, so as to obtain two alignment relationships of environment to biology and biology to environment; further, extracting a forward forced destruction degree and a reverse counter-attack destruction degree, and adaptively generating a weight parameter in combination with the time sequence change; finally, dynamically weighting a physical environment risk sub-index and a biological response risk sub-index, and outputting a marine ecological risk index. The application can simultaneously depict a two-way hysteresis relationship between environmental stress and ecological feedback, and improve the dynamicity, accuracy and early warning reliability of marine ecological risk assessment.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology, and in particular to a digital twin method for marine ecological risk based on the Internet of Things. Background Technology

[0002] With the increasing demand for refined monitoring of marine ranches, nearshore aquaculture areas, port sewage discharge areas, and key ecological protection marine areas, marine ecological monitoring is gradually evolving from manual sampling to continuous sensing via the Internet of Things (IoT). Through buoy sensors, underwater monitoring nodes, remote sensing equipment, and shore-based monitoring platforms, physicochemical environmental parameters such as water temperature, salinity, dissolved oxygen, pH, nutrients, and heavy metals, as well as biological response indicators such as chlorophyll concentration, plankton density, benthic organism activity, and diversity index, can be continuously collected. This multi-source data provides the foundation for constructing a digital twin model of the marine ecosystem, enabling the monitoring system to synchronously map the real-world marine conditions in virtual space.

[0003] However, marine ecosystems are characterized by strong nonlinearity, strong time lag, and multi-factor coupling. Environmental anomalies and biological responses are not simply synchronous changes, and the degradation of biological communities may also have a reverse effect on the local environmental state.

[0004] Existing digital twin risk assessments often focus on data display, threshold warnings, or static comprehensive scoring, making it difficult to reveal the cross-time dependence and two-way feedback relationship between environmental stress and ecological response. This results in risk indices being lagging, one-sided, or lacking in interpretation. Summary of the Invention

[0005] In view of the aforementioned existing problems, the present invention is proposed.

[0006] Therefore, this invention provides a digital twin method for marine ecological risk based on the Internet of Things to solve the problems that existing marine ecological risk assessments cannot characterize the two-way lag effect between the environment and organisms and that risk weights are difficult to determine dynamically.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a digital twin method for marine ecological risk based on the Internet of Things, which includes expressing the physicochemical environmental parameters and biological response indicators in the region through two-dimensional curves, and converting the two-dimensional curves into a feature sequence with respect to time based on the curve characteristics. Using the improved Copula joint distribution model, we perform cross-time axis multidimensional feature unidirectional matching on two feature sequences to obtain two alignment relationships of unidirectional influence between the two sequences. Based on the two alignment relationships, conditional destructive features representing the substantial damage caused to each other between the two feature sequences are generated respectively. Based on an adaptive weight allocation strategy, the temporal changes of the conditional damage characteristics are transformed into weight parameters. Using the weight parameters, the two risk sub-indices generated by the conditional damage characteristics are dynamically weighted and summed, and the summation result is mapped to the final marine ecological risk index.

[0008] As a preferred embodiment of the IoT-based digital twin method for marine ecological risk described in this invention, the curve features include mapping each type of physicochemical environmental parameter data and each type of biological response index data to a two-dimensional coordinate system with time as the horizontal axis and numerical value as the vertical axis, and using a spline interpolation algorithm to fit and generate a continuous two-dimensional curve. Based on the local geometric shape change characteristics of the continuous change curve in adjacent time intervals, shape change detection is performed on each of the continuous change curves to identify the morphological change positions where the local geometric similarity is lower than a preset threshold, and an independent timestamp is inserted on the horizontal axis coordinate corresponding to each morphological change position to realize the dynamic interval division of a single curve. For each physicochemical environmental parameter and each biological response index, the independent timestamps of all curves are combined by taking the union of the sets to obtain the global environmental timestamp set and the global biological timestamp set.

[0009] As a preferred embodiment of the IoT-based digital twin method for marine ecological risk described in this invention, the global environmental timestamp set and the global biological timestamp set are inserted into each type of physicochemical environmental parameter data and each type of biological response index data, respectively, and the physicochemical environmental parameters and biological response indexes are sliced ​​according to the inserted timestamps to obtain the slice sequence of physicochemical environmental parameters and biological response indexes. After encoding the multi-source data in each slice, characteristic sequences of physicochemical environmental parameters and biological response indicators were obtained.

[0010] As a preferred embodiment of the IoT-based digital twin method for marine ecological risk described in this invention, the improved Copula joint distribution model includes constructing a two-dimensional grid matrix by using the time axis of the physicochemical environmental parameter feature sequence as the horizontal axis and the time axis of the biological response index feature sequence as the vertical axis. For each intersecting coordinate node in the two-dimensional grid matrix, the physical environment integrated coding element and the biological response integrated coding element under the corresponding time slice are extracted. The multidimensional feature joint dependency degree between the two is calculated using the vine Copula joint distribution model. The joint dependency degree is then reverse-mapped to generate the basic matching cost of the coordinate node.

[0011] As a preferred embodiment of the IoT-based digital twin method for marine ecological risk described in this invention, the basic matching matrix is ​​copied into a first cost matrix and a second cost matrix. In the first cost matrix, a positive delayed causal penalty is applied: when the physical environment time represented by the horizontal axis lags behind the biological response time represented by the vertical axis, the cross-axis node is given an infinite penalty cost. In the second cost matrix, a reverse delayed causal penalty is applied: when the biological response time represented by the vertical axis lags behind the physical environment time represented by the horizontal axis, an infinite penalty cost is assigned to the cross-axis node. The dynamic programming path optimization rule in the dynamic time warping algorithm is adopted to search for the continuous coordinate point sequence with the minimum global cumulative matching cost from the time start point to the end point in the first cost matrix and the second cost matrix after applying the penalty. The two optimal time warping paths obtained are output as the first alignment relationship and the second alignment relationship representing the unidirectional influence between the two sequences.

[0012] As a preferred embodiment of the IoT-based digital twin method for marine ecological risk described in this invention, the conditional damage characteristics include positive forcing damage and negative backlash damage. Under the optimal time warping path determined by the first alignment relationship, based on the vine Copula joint probability density function, the conditional expected loss of the multidimensional characteristics of biological response indicators under extreme adverse conditions of physicochemical environmental parameters is calculated, and the conditional expected loss is quantified as the positive pressure damage degree. Under the optimal time warping path determined by the second alignment relationship, based on the vine Copula joint probability density function, the expected deterioration of the multidimensional characteristics of the physicochemical environmental parameters of the biological response index under extreme decay conditions is calculated, and the expected deterioration is quantified as the degree of reverse backlash damage.

[0013] As a preferred embodiment of the IoT-based digital twin method for marine ecological risk described in this invention, the adaptive weight allocation strategy includes extracting the time series sequence of the positive forcing damage degree and the time series sequence of the reverse backlash damage degree within a continuous time sliding window. The first derivative of the time series of the positive stress damage degree and the time series of the negative backlash damage degree are respectively obtained with respect to the time variable to obtain the positive damage change rate characterizing the rate of environmental stress intensification and the negative damage change rate characterizing the rate of ecological backlash intensification. A weight ratio allocation equation is constructed, and the ratio of the physical fusion weight parameter corresponding to the physical environment risk sub-index to the biological fusion weight parameter corresponding to the biological response risk sub-index is set to be equal to the ratio of the absolute value of the positive damage change rate to the absolute value of the negative damage change rate. Introduce a weight normalization constraint: set the sum of the physical fusion weight parameter and the biological fusion weight parameter at any time to be 1; Solve the equations for weight ratio allocation and the weight normalization constraints simultaneously to output the physical fusion weight parameters and the biological fusion weight parameters at the current evaluation time.

[0014] As a preferred embodiment of the IoT-based digital twin method for marine ecological risk described in this invention, the real-time physicochemical environmental parameters are quantified using a preset linear function, and the positive stress damage degree is used as a modulation factor for the intensity of damage exerted by the environment on organisms. This is coupled with the quantification results of the physicochemical environmental parameters to obtain the physical environmental risk sub-index. After quantifying the real-time biological response index through a preset encoder, the degree of reverse backlash damage is used as a modulation factor for the intensity of reverse disturbance of the environment caused by biological changes. This factor is coupled with the quantification result of the biological response index to obtain the biological response risk sub-index. Wherein, the coupling calculation is expressed as: R=E×(1+D); R represents the coupling calculation result; E represents the quantitative result of physicochemical environmental parameters or biological response indicators; D represents the degree of positive stress damage or the degree of reverse backlash damage; The marine ecological risk index includes a fixed risk index configured in each segmented interval of a pre-defined weighted summation result; when the weighted summation result falls into the corresponding interval, the corresponding risk index is output.

[0015] In a second aspect, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the Internet of Things-based digital twin method for marine ecological risk as described in the first aspect of the present invention.

[0016] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the Internet of Things-based digital twin method for marine ecological risk as described in the first aspect of the present invention.

[0017] The beneficial effects of this invention are as follows: it enables the unified quantification of environmental stress and ecological feedback in marine ecosystems by comprehensively utilizing physicochemical environmental parameters and biological response indicators within the same assessment framework. By constructing feature matching relationships across time axes, it can more accurately identify the lagged damage caused by environmental changes to biological systems and the reverse perturbation of environmental conditions by biological anomalies, thereby improving the ability of risk assessment to depict real ecological evolution processes. Furthermore, by introducing conditional damage degree to characterize the intensity of bidirectional damage and adaptively generating risk fusion weights based on its temporal changes, it can enhance the responsiveness of assessment results to the switching of risk-dominant factors and complex coupling scenarios. Compared with traditional static weighted or synchronous correlation analysis methods, this invention has better temporal consistency, bidirectional causal explanation capability, and dynamic early warning capability, providing a more reliable quantitative basis for marine pollution prevention and control, ecological restoration, and marine risk management. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of an IoT-based digital twin method for marine ecological risk.

[0020] Figure 2 A diagram of a computer device for a digital twin method of marine ecological risk based on the Internet of Things. Detailed Implementation

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0022] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0023] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0024] Reference Figure 1 and Figure 2 As one embodiment of the present invention, this embodiment provides a digital twin method for marine ecological risk based on the Internet of Things, including the following steps: S1: The physicochemical environmental parameters and biological response indicators within the region are expressed by two-dimensional curves, and the two-dimensional curves are transformed into a time-related feature sequence based on the curve characteristics.

[0025] In a preferred embodiment of the present invention, the physicochemical environmental parameters include, but are not limited to: seawater surface temperature, bottom temperature, salinity, turbidity, hydrodynamic parameters, dissolved oxygen, chemical oxygen demand, inorganic nitrogen concentration, inorganic phosphorus concentration, and typical heavy metal ion concentration. The hydrodynamic parameters include flow velocity and flow direction, and the typical heavy metal ions include one or more of copper, lead, zinc, and cadmium ions. The biological response indicators are used to characterize the state changes of the marine ecosystem at different levels and may include, but are not limited to: chlorophyll a concentration in surface water, growth degradation rate or decay area ratio of macroalgae, seagrass bed coverage decline rate, population density change rate of typical economic fish groups or ecological indicator fish species, abnormal migration path deviation, group mortality rate or deformity rate of fish eggs and larvae, abnormal avoidance behavior or stranding frequency of large marine mammals, biomass change rate of dominant benthic species such as crustaceans, mollusks, and echinoderms, abnormal behavioral frequency, benthic community diversity index, zooplankton mortality rate, and abnormal fluctuations in the activity of toxic stress kinases in indicator organisms. The toxic kinase activity may include superoxide dismutase activity.

[0026] In actual marine environmental monitoring, the aforementioned multidimensional data typically originates from different monitoring methods and exhibits heterogeneous data with discrete sampling and inconsistent sampling frequencies. For example, fish density may come from sonar detection data, seaweed coverage from remote sensing observation data, and water quality parameters from buoy sensor data. Due to differences in sampling period, temporal resolution, and recording methods among different data sources, temporal misalignment can easily occur between subsequent environmental parameters and biological responses. Therefore, in this embodiment, each type of physicochemical environmental parameter data and each type of biological response index data is mapped to a two-dimensional coordinate system with time as the horizontal axis and numerical measurement as the vertical axis, and a spline interpolation algorithm is used for fitting to generate corresponding continuous two-dimensional curves. This achieves a continuous expression of multi-source heterogeneous discrete monitoring data, providing a unified data foundation for subsequent morphological mutation identification, timestamp extraction, and cross-time axis feature matching.

[0027] It is important to understand that in this invention, the Internet of Things (IoT) is not merely used as a means of ordinary data transmission, but rather as the data sensing entry point and the foundation for real-time status updates of the marine ecological digital twin model. Specifically, through IoT sensing devices deployed in the target sea area, such as buoy monitoring nodes, underwater sensing nodes, shore-based monitoring terminals, sonar detection equipment, remote sensing observation terminals, unmanned vessels, or unmanned underwater vehicles, the marine environmental status and the ecological biological response status are continuously collected. Each IoT node, according to its monitoring object and installation location, undertakes tasks such as water quality sensing, hydrodynamic sensing, biological distribution sensing, biological behavior sensing, and remote sensing image acquisition, thereby forming a multi-level sensing network covering the sea surface, water body, seabed, and nearshore areas.

[0028] For physicochemical environmental parameters, the Internet of Things (IoT) is primarily used to achieve continuous online acquisition of environmental conditions. For example, buoy sensors can acquire real-time water quality information such as surface temperature, salinity, turbidity, and dissolved oxygen; underwater fixed sensor nodes can collect bottom temperature, bottom dissolved oxygen, and local pollutant concentrations; acoustic Doppler current profilers can acquire hydrodynamic parameters such as flow velocity and direction; and online water quality analyzers or automatic water sampling and detection equipment can acquire chemical indicators such as inorganic nitrogen, inorganic phosphorus, chemical oxygen demand, and heavy metal ion concentrations. The data collected by these devices is uploaded via IoT communication links, enabling the digital twin system to continuously obtain environmental change information from the real marine environment.

[0029] For biological response indicators, the Internet of Things (IoT) is mainly used to achieve multi-source sensing and event capture of biological states. For example, sonar detection nodes can acquire information on fish density, distribution, and migration path changes; underwater cameras, unmanned surface vessel (USV) inspection equipment, or remote sensing terminals can acquire images of seaweed coverage, seagrass bed degradation area, and benthic organism distribution; biotags, acoustic monitoring devices, or shore-based observation terminals can acquire information on abnormal behavior, risk avoidance behavior, or stranding events of large marine animals; and automatic sampling devices and biological detection terminals can input detection results such as zooplankton mortality rate, benthic biodiversity index, and superoxide dismutase (SOD) activity into the system. Therefore, this invention can transform ecological response information, which traditionally relies on manual surveys or discrete sampling, into IoT data that can be received and processed by a digital twin platform.

[0030] This solution utilizes IoT data in several ways. First, it enables continuous access to multi-source ecological data via IoT, allowing the digital twin model to be updated synchronously with the real-world marine conditions. Second, it unifies asynchronous, multi-frequency, and heterogeneous data collected from different IoT nodes into continuous two-dimensional curves, resolving the issue of inconsistent sampling times between different sensors. Third, it extracts morphological mutation locations and global timestamps based on the continuous curves, enabling comparisons of environmental parameters and biological response indicators on a unified time scale. Finally, it inputs the IoT-collected and uniformly represented data into the Copula joint distribution model and the cross-time axis matching model to further analyze the positive impact of environmental stress on biological responses and the negative feedback of biological degradation on environmental conditions. In other words, IoT plays a fundamental role in this invention: "real-world marine perception—continuous data access—digital twin status update—dynamic risk calculation."

[0031] Furthermore, based on the local geometric shape change characteristics of the continuously changing curves within adjacent time intervals, shape change detection is performed on each of the continuously changing curves to identify abrupt shape changes where the local geometric similarity is below a preset threshold. An independent timestamp is then inserted on the horizontal axis corresponding to each abrupt shape change location, thereby achieving dynamic interval division of a single curve. Specifically: (1) Set a sliding time window on the continuous two-dimensional curve and extract continuous local short-time curve segments (sub-trajectories) that slide along the time axis.

[0032] (2) Use a geometric distance metric algorithm (such as Fréchet distance or waveform cosine similarity) to calculate the short-term geometric similarity between adjacent or time-similar local short-time curve segments.

[0033] (3) Monitor the evolution trajectory of the short-term geometric similarity as the sliding time window evolves.

[0034] (4) When the short-term geometric similarity between adjacent local short-term curve segments drops and the drop amplitude exceeds the preset local steady-state threshold, it is determined that the original waveform structure of the system has disintegrated, and the connection node of the local short-term curve segment that has dropped is established as the morphological change position.

[0035] It completely eliminates the blind spots of traditional segmentation based solely on numerical magnitude or first and second derivatives (extremes / inflection points). Because ecological parameters have specific periodic wave "shapes" (high similarity) under normal conditions, this algorithm is specifically designed to capture such "ruptures in local shape steady state," thus enabling it to most sensitively and accurately segment unknown types of ecological anomalies from background noise.

[0036] The global environmental timestamp set is obtained by taking the union of the independent timestamps of all curves under the category of physicochemical environmental parameters; the global biological timestamp set is obtained by taking the union of the independent timestamps of all curves under the category of biological response indicators.

[0037] Existing fixed segmentation techniques easily interrupt a complete, sudden pollution event (such as a red tide outbreak causing mass fish deaths due to suffocation). To address this, this invention proposes a dynamic slicing mechanism based on local geometric similarity: The system slides a short-time window across each continuous two-dimensional curve to calculate the geometric similarity of the curve waveforms within adjacent time intervals. Under normal conditions, ocean parameters generally exhibit stable fluctuations; however, when crises such as a sudden increase in pollution or a sharp decline in fish populations occur, the curve waveforms undergo drastic abrupt changes. When the system detects a local geometric similarity below a preset steady-state threshold, it determines that a "morphological abrupt change" has occurred at that location and inserts an independent timestamp on the horizontal time coordinate corresponding to each such a morphological abrupt change location.

[0038] Subsequently, the system performs a union operation on the independent timestamps under all physicochemical environmental parameter curves to obtain a "global environmental timestamp set"; similarly, it aggregates the mutation timestamps of all plants and animals to obtain a "global biological timestamp set". Finally, these two global sets are inserted into their respective data sources to synchronously segment all continuous curves.

[0039] After the union of data and the synchronization of slices, each slice interval is no longer a fixed, mechanical passage of time, but rather represents an independent ecological evolutionary stage with a relatively stable internal state and boundaries triggered by specific ecological mutation events (such as a sudden drop in water quality or fish migration). This lays an extremely accurate and high-quality data foundation for the subsequent extraction of high-dimensional feature sequences containing absolute values ​​and geometric shapes, as well as for feeding them into the Copula model for bidirectional causal matching.

[0040] The global environmental timestamp set and the global biological timestamp set are inserted into each type of physicochemical environmental parameter data and each type of biological response index data, respectively. The physicochemical environmental parameters and biological response indexes are sliced ​​according to the inserted timestamps to obtain the slice sequence of physicochemical environmental parameters and biological response indexes.

[0041] After encoding the multi-source data in each slice, feature sequences of physicochemical environmental parameters and biological response indicators are obtained. Each element in the sequence contains corresponding curve slices for multiple parameters. For example, the absolute numerical baseline features of all types of physicochemical environmental parameters within the same time slice are concatenated with geometric morphology vectors using a multidimensional tensor to generate a comprehensive physical environment coding element representing the overall state of that time slice.

[0042] For each time slice, instead of simply extracting the instantaneous values ​​of a single physicochemical environmental parameter or a single biological response index, the curve segment information corresponding to all parameters within that time slice is summarized to construct a comprehensive coding element that can characterize the overall state of that time slice. Specifically, for the physicochemical environmental parameters, firstly, the absolute numerical benchmark features of various physicochemical environmental parameters within the same time slice are extracted. These absolute numerical benchmark features include one or more of the mean, extreme values, final values, change amplitude, or relative benchmark offset of each parameter within that time slice. Then, the geometric shape vectors of the curve segments corresponding to each parameter are extracted. These geometric shape vectors are used to characterize the local structural features of the curve segments within that time slice, and may include one or more of the slope, curvature, arc length, fluctuation frequency, local oscillation intensity, trend direction, and morphological similarity descriptors. Afterward, the absolute numerical benchmark features of various physicochemical environmental parameters and the geometric shape vectors are concatenated along the feature dimension according to the parameter type, and dimension compression, normalization mapping, or tensor fusion is performed through a preset encoder to generate a comprehensive physical environment coding element representing the overall environmental state of that time slice. Correspondingly, for the biological response index side, the absolute numerical baseline features and geometric shape vectors of various biological response indices within the same time slice are extracted, and the same or corresponding encoding methods are used to splice and fuse these features to generate a comprehensive biological response encoding element representing the overall biological state of that time slice. Thus, arranging the comprehensive physical environment encoding elements and comprehensive biological response encoding elements corresponding to each time slice along the time axis forms the physicochemical environmental parameter feature sequence and the biological response index feature sequence, respectively.

[0043] S2: Using the improved Copula joint distribution model, the two feature sequences are matched one-way with each other across the time axis in a multidimensional feature direction, resulting in two alignment relationships of one-way influence between the two sequences.

[0044] Furthermore, the improved Copula joint distribution model includes constructing a two-dimensional grid matrix by using the time axis of the physicochemical environmental parameter feature sequence as the horizontal axis and the time axis of the biological response index feature sequence as the vertical axis.

[0045] For each intersecting coordinate node in the two-dimensional grid matrix, the physical environment integrated coding element and the biological response integrated coding element under the corresponding time slice are extracted. The multidimensional feature joint dependency degree between the two is calculated using the vine Copula joint distribution model. The joint dependency degree is then reverse-mapped to generate the basic matching cost of the coordinate node.

[0046] The "vine-like Copula joint distribution model" refers to a high-dimensional statistical modeling structure used to describe complex joint probabilistic relationships among multiple variables. Its basic idea is not to directly establish a high-dimensional joint distribution for all variables as a whole, but rather to first preserve the marginal distribution characteristics of each variable, and then decompose the high-dimensional joint distribution into a combination of multiple two-dimensional conditional Copula functions through a tree-like hierarchical structure, thereby characterizing the correlation, conditional correlation, and tail dependency among multiple variables layer by layer.

[0047] The "improved vine Copula joint distribution model" refers to the traditional vine Copula model, which is only used to characterize the static joint distribution relationship of multiple variables. It further incorporates the path matching idea of ​​dynamic time warping, and converts the multidimensional joint dependence between the comprehensive coding elements of physicochemical environmental parameters and the comprehensive coding elements of biological response under different time slices into the node matching cost in the dynamic time warping algorithm. This expands the vine Copula model, which was originally only applicable to static high-dimensional correlation analysis, into a dynamic joint distribution matching model applicable to cross-time axis causal alignment and hysteresis relationship identification.

[0048] "Reverse mapping of joint dependency" refers to: first, using the Copula joint distribution model to calculate the joint dependency degree between the physical environment integrated coding element and the biological response integrated coding element at a certain cross-time coordinate node; then, converting this dependency degree into the basic matching cost used in path optimization according to the opposite direction of the correspondence. This means that a higher joint dependency degree indicates a greater likelihood of a real correlation or reasonable pairing between the environmental time slice and the biological time slice, and therefore a lower matching cost; conversely, a lower joint dependency degree indicates a weaker matching rationality, and therefore a higher matching cost. Therefore, "reverse mapping" essentially transforms the joint dependency degree, which is considered "larger, better matching," into a basic matching cost, which is considered "larger, worse mismatch, higher cost," thus satisfying the solution requirement of the dynamic time warping algorithm to find the optimal path with the minimum total cost.

[0049] In this embodiment, a computer-based solution process based on the topological deconstruction of dual-vine (D-Vine) or typical-vine (C-Vine) is provided: Set at any intersection node of the two-dimensional grid matrix At this point, the physical environment comprehensive coding elements extracted by the system are m-dimensional vectors. The biological response comprehensive coding elements are n-dimensional vectors. To calculate the joint correlation between the two, the system concatenates them into a global state vector of total dimension d. , where d = m + n.

[0050] The specific calculation steps are as follows: Step 1: Marginal Distribution Fitting and Probability Integral Transform (PIT) First, the system uses nonparametric kernel density estimation (KDE) or empirical cumulative distribution function (ECDF) to calculate each feature dimension in the global state vector Z. Marginal cumulative distribution function Subsequently, through probability integral transformation, all physical, chemical, and biological absolute values ​​and geometric characteristics with different dimensions are uniformly mapped to a state where... Standard uniform distribution sequence of interval Its mapping formula is: The k-th feature dimension in the global state vector Z can be the absolute numerical or geometrical feature of a certain physicochemical environmental parameter, or the absolute numerical or geometrical feature of a certain biological response index; d represents the total feature dimension of the global state vector Z. Represents the k-th feature dimension The corresponding marginal cumulative distribution function is used to describe the random variable. The cumulative probability when the value is not greater than the current observation value; U represents the standard uniform distribution sequence obtained after probability integral transformation; Indicates the first Each feature dimension The probability value after mapping by the marginal cumulative distribution satisfies .

[0051] This step effectively eliminates the dimensional barriers between heterogeneous parameters such as environmental pollution concentration and biological mortality rate.

[0052] Step S2: High-dimensional joint density deconstruction based on tree topology According to Sklar's theorem, a high-dimensional joint probability density function can be decomposed into the product of a marginal probability density function and a Copula density function. To address the "curse of dimensionality" in high-dimensional Copula computation, a VineStructure Generator is invoked to... dimensional joint Copula density function The structure is deconstructed step by step into a binary Pair-Copula density function.

[0053] Taking the D-Vine structure as an example, it decomposes a high-dimensional dependency network into d-1 trees. The first tree describes the unconditional dependency between adjacent variables, and subsequent trees describe the conditional dependency based on the preceding variable.

[0054] Step S3: Binary Copula Optimization and Conditional Distribution Transformation (h-function) At each node of the vine structure, the system uses the maximum likelihood estimation method (MLE) or the Akaike information criterion (AIC) to automatically find the optimal Copula function from the preset family of Copulas.

[0055] Preferably, in order to accurately capture the asymmetric tail characteristics of "extreme deterioration" of marine ecology, the system selects Gumbel Copula for the upper tail-dependent node representing pollution outbreak and Clayton Copula for the lower tail-dependent node representing ecological collapse.

[0056] For the j-th tree The conditional probability calculation in the text introduces a partial derivative h function for a step-by-step recursive transformation: in, This indicates that, given a variable v and a Copula parameter set... Under the given conditions, the conditional distribution mapping result of variable u; Let represent the bivariate Copula distribution functions corresponding to variables u and v; u and v represent the two input standard uniform variables.

[0057] Step S4: Cascade calculation of joint dependency scalar Finally, the system performs a concatenation and multiplication of the binary probability density values ​​fitted by all nodes in the vine structure to calculate the global joint probability density scalar value. The specific calculation formula is as follows: in, This represents the binary Pair-Copula density value under a given sequence of conditional variables. The system will use the calculated global joint probability density scalar value. This serves as the "multidimensional feature joint dependency degree" under this coordinate node. j represents the tree level index in the vine structure, with a value range from 1 to d-1; i represents the index of the edge in each level of the tree, with a value range from 1 to dj; Indicates the set of condition variables Given conditions, the bivariate pair-copula density function between the i-th variable and the (i+j-th)-th variable; Representing variables In condition variables Conditional distribution values ​​under given conditions; Representing variables The conditional distribution value under the same conditional variable constraints. This joint dependency degree. The numerical value rigorously quantifies the probability strength of a substantial causal fit between the environmental stress feature group and the biological degradation feature group in the probabilistic topological space under the current time slice combination. Subsequently, the system executes... The reverse mapping process generates the pathfinding cost for the Dynamic Time Warping (DTW) algorithm.

[0058] Furthermore, the basic matching matrix is ​​copied into a first cost matrix and a second cost matrix.

[0059] In the first cost matrix, a positive delayed causal penalty is applied: when the physical environment time represented by the horizontal axis lags behind the biological response time represented by the vertical axis, the cross-axis node is given an infinite penalty cost to block the possibility of matching that violates the order of "environmental oppression of organism".

[0060] In the second cost matrix, a reverse delayed causal penalty is applied: when the biological response time represented by the vertical axis lags behind the physical environment time represented by the horizontal axis, an infinite penalty cost is assigned to the cross-axis node to block the possibility of matching that violates the order of "biological decay backlashes on the environment".

[0061] The dynamic programming path optimization rule in the dynamic time warping algorithm is adopted to search for the continuous coordinate point sequence with the minimum global cumulative matching cost from the time start point to the end point in the first cost matrix and the second cost matrix after applying the penalty. The two optimal time warping paths obtained are output as the first alignment relationship and the second alignment relationship representing the unidirectional influence between the two sequences.

[0062] The first alignment relation represents the lagged effect of physicochemical environmental parameters on biological response indicators, where environmental change is the preceding cause and biological change is the lagged result. Under this alignment relation, environmental damage does not immediately trigger biological decline at the same moment, but rather requires processes such as pollution diffusion, material enrichment, toxicity transmission, or ecological accumulation before organisms exhibit abnormalities, inhibited growth, or death. Therefore, in dynamic time warping, the corresponding time matching must satisfy the condition that the physical environmental change time precedes the biological response time to avoid spurious causal matching where "the environment has not yet deteriorated but the organisms have already declined." Taking a heavy metal spill scenario as an example, a sudden environmental change occurs at a certain moment when heavy metals enter the sea. After a period of diffusion and accumulation in the water and food chain, large-scale mortality of benthic bivalves is detected. In this case, the first alignment relation is used to pair the aforementioned environmental change moment with the subsequent biological decline moment across time, thereby extracting the delayed destructive characteristics of the environment causing substantial damage to organisms.

[0063] The second alignment relationship represents the lagged reaction relationship between biological response indicators and physicochemical environmental parameters, where biological change is the preceding cause and environmental change is the lagged result. Under this alignment relationship, the abnormality, decline, or death of biological systems does not merely exist as a victimized outcome; it may also, after a certain period, cause secondary deterioration of the aquatic environment through mechanisms such as decomposition of dead organisms, community imbalance, niche vacancy, or metabolic chain disruption. Therefore, in the dynamic time warping process, the corresponding time matching must satisfy the condition that the biological response time precedes the environmental deterioration time to avoid erroneous alignment where "the biological change is interpreted as a result before secondary environmental deterioration has occurred." Taking the scenario of algal bloom followed by death as an example, after a large number of algae die and settle in a short period of time, aerobic microorganisms decompose them. After several days of continuous oxygen consumption, this eventually leads to a decrease in dissolved oxygen in the bottom water and induces acidification. In this case, the second alignment relationship is used to pair the aforementioned biological anomaly moment with the subsequent environmental deterioration moment across time, thereby extracting the lagged destructive characteristics of the backlash effect of biological change on the environment.

[0064] It is important to understand that there are significant time lags and directional differences between environmental changes and biological responses in marine ecosystems. Environmental degradation often damages organisms through processes such as pollution diffusion, toxicity accumulation, and food chain transmission. Conversely, the reverse disturbances caused by abnormal biological decline typically manifest only gradually after processes such as decomposition, increased oxygen consumption, and community imbalance. Therefore, the two do not satisfy a simple synchronous correspondence. If the current unconstrained matching method is used, it is easy to generate pseudo-causal paths such as "future environment explaining past biological responses" or "future biological anomalies explaining past environmental degradation," reducing the ecological rationality of the alignment results. By replicating the basic matching matrix into a first cost matrix and a second cost matrix, and applying positive and negative delayed causal penalties respectively, the two different unidirectional influence relationships of "environment preceding organisms" and "organisms preceding environment" can be constrained in their respective independent optimization spaces. This allows the dynamic time warping algorithm, when performing global path search, to no longer mechanically pair organisms based solely on geometric distance or minimum numerical cost, but to simultaneously satisfy the temporal order requirements of ecological processes. Therefore, on the one hand, it can effectively block erroneous matching paths that violate the true laws of ecological evolution and improve the causal credibility of cross-time axis alignment results between environmental parameters and biological responses; on the other hand, it can distinguish and extract the retardation and oppression effect of the environment on organisms and the reverse backlash effect of organisms on the environment, thereby providing a more realistic ecological process basis for subsequent condition damage calculation, risk sub-index generation and marine ecological risk index assessment.

[0065] S3: Generate conditional destructive features that characterize the substantial damage caused to each other between the two feature sequences, based on the two alignment relationships respectively.

[0066] Specifically, the conditional destructiveness features include positive pressure destructiveness and negative backlash destructiveness. Under the optimal time warping path determined by the first alignment relationship, based on the vine Copula joint probability density function, the conditional expected loss of the multidimensional features of biological response indicators under extreme adverse conditions of physicochemical environmental parameters is calculated, and the conditional expected loss is quantified as the positive pressure destructiveness.

[0067] Under the optimal time warping path determined by the second alignment relationship, based on the vine Copula joint probability density function, the expected deterioration of the multidimensional characteristics of the physicochemical environmental parameters of the biological response index under extreme decay conditions is calculated, and the expected deterioration is quantified as the degree of reverse backlash damage.

[0068] The Copula joint probability density function is a global probabilistic model describing the nonlinear dependence between multidimensional environmental features and multidimensional biological responses. Since the total dimension d=m+n of environmental indicators (e.g., heavy metals, dissolved oxygen, etc., m-dimensional) and biological indicators (e.g., fish density, algal cover, etc., n-dimensional) is extremely high, directly fitting a high-dimensional distribution would lead to computational failure. The Copula uses a "vine decomposition" technique to decompose this high-dimensional joint probability density function into... A cascaded product of two binary Copula functions.

[0069] It is equivalent to a "probability map of causal relationships". Given a set of environmental and biological data, the larger the value (probability density value) output by this function, the stronger the statistical and ecological match and the stronger the causal relationship between the two.

[0070] Conditional Expected Loss (CEL): Under the premise of "extremely harsh environment (e.g., pollution concentration exceeding the 95th percentile)," CEL is the expected value of the degree of biomarker decline calculated by integrating a Copula function. It represents the expected mortality rate of a organism if the environment deteriorates to this extent. Conditional Expected Deterioration (CED): Under the premise of "biological collapse (e.g., population density below the 5th percentile)," this is the mathematical expectation of further deterioration of environmental indicators. It represents the extent to which the environment will deteriorate if there is widespread biological mortality. S4: Based on an adaptive weight allocation strategy, the temporal changes of the conditional damage characteristics are transformed into weight parameters; using the weight parameters, the two risk sub-indices generated by the transformation of the conditional damage characteristics are dynamically weighted and summed, and the summation result is mapped to the final marine ecological risk index.

[0071] The adaptive weight allocation strategy includes extracting the time series sequences of the positive pressure damage and the negative backlash damage within a continuous time sliding window. The first derivative of each time series sequence is calculated with respect to time to obtain the rate of change of positive damage, which characterizes the rate of increase in environmental stress, and the rate of change of negative damage, which characterizes the rate of increase in ecological backlash.

[0072] A weighting equation is constructed, setting the ratio of the physical fusion weight parameter corresponding to the physical environment risk sub-index to the biological fusion weight parameter corresponding to the biological response risk sub-index to be equal to the ratio of the absolute value of the positive damage rate of change to the absolute value of the negative damage rate of change. A weight normalization constraint is introduced: the sum of the physical fusion weight parameter and the biological fusion weight parameter is set to 1 at any given time.

[0073] Solve the equations for weight ratio allocation and the weight normalization constraints simultaneously to output the physical fusion weight parameters and the biological fusion weight parameters at the current evaluation time.

[0074] The quantification of the positive pressure damage can be understood as follows: Under the optimal time warping path determined by the first alignment relationship, taking the extreme adverse conditions of the physicochemical environmental parameters as a constraint event, first calculate the expected conditional loss of the multidimensional characteristics of the biological response indicators relative to their healthy baseline state, and then map this expected conditional loss into a dimensionless damage intensity value. The quantification of the reverse backlash damage can be understood as follows: Under the optimal time warping path determined by the second alignment relationship, taking the extreme degradation conditions of the biological response indicators as a constraint event, first calculate the expected conditional deterioration of the multidimensional characteristics of the physicochemical environmental parameters relative to their normal environmental baseline state, and then map this expected conditional deterioration into a dimensionless damage intensity value.

[0075] In this embodiment, specifically, the health baseline vector of the biological response index is set as follows: The biological response feature vector under the current paired node is B, and the constraint event under the extremely harsh environmental condition is... Then the conditional expected loss can be expressed as: in, This represents the average loss of biological characteristics relative to a health baseline under environmental stress conditions. This indicates the desired outcome. Subsequently, a preset biological damage reference threshold is used. Alternatively, using historical normal fluctuation amplitude as a normalization benchmark, the conditional expected loss can be quantified as the degree of positive forced destruction: in, Indicates the degree of positive coercion and destruction. To prevent tiny positive numbers with a denominator of zero, this formula indicates that the greater the conditional expected loss, the higher the degree of destructive force; when the loss reaches or exceeds the reference threshold, the degree of destructive force approaches 1.

[0076] After quantifying the real-time physicochemical environmental parameters using a preset linear function, the positive stress damage degree is used as a modulating factor of the intensity of damage exerted by the environment on organisms. This is then coupled with the quantification results of the physicochemical environmental parameters to obtain the physical environment risk sub-index.

[0077] After quantifying the real-time biological response index using a preset encoder, the degree of reverse backlash damage is used as a modulation factor for the intensity of the reverse disturbance of the environment caused by biological changes. This factor is then coupled with the quantification results of the biological response index to obtain the biological response risk sub-index.

[0078] The coupling calculation is expressed as: R=E×(1+D); R represents the coupling calculation result; E represents the quantification result of physicochemical environmental parameters or biological response indicators; D represents the degree of positive pressure damage or the degree of negative backlash damage, and D+1 represents the "modulation factor".

[0079] In this embodiment, the preset encoder is preferably a multilayer perceptron encoder (MLP Encoder). Specifically, the original input vector of the biological response index at the current evaluation time can be input into a pre-trained or pre-structured multilayer perceptron network. Through fully connected layers, nonlinear activation layers, and output mapping layers, biological response indices of different dimensions and types are mapped to a unified dimension, outputting a fixed-dimensional biological response quantification result. The reason for adopting this approach is that the biological response index in this embodiment has usually undergone prior time slicing and feature sequence construction processing, and its input form is closer to a structured feature vector. Using a multilayer perceptron encoder can stably complete feature compression, nonlinear mapping, and comprehensive representation, facilitating subsequent coupling calculation with the reverse backlash destructive degree.

[0080] In other alternative embodiments, the preset encoder may also be one or more of the following: autoencoder, convolutional neural network encoder, one-dimensional convolutional encoder, long short-term memory network encoder, gated recurrent unit network encoder, Transformer encoder, principal component analysis mapping module, or tensor decomposition encoding module.

[0081] The marine ecological risk index includes a pre-defined segmented interval for the weighted summation result, with a fixed risk index configured in each interval; when the weighted summation result falls into the corresponding interval, the corresponding risk index is output as the final digital twin result.

[0082] For example, it can be pre-defined that: when the weighted summation result is in the range of 0 to 0.2, a level 1 risk index is output; when it is in the range of 0.2 to 0.4, a level 2 risk index is output; when it is in the range of 0.4 to 0.6, a level 3 risk index is output; when it is in the range of 0.6 to 0.8, a level 4 risk index is output; and when it is in the range of 0.8 to 1.0, a level 5 risk index is output. In this way, the system ultimately provides a standardized risk result that is not a continuous value that is difficult to interpret directly, but a result with clear warning and management significance.

[0083] It's worth noting that while directly outputting continuous risk values ​​offers a certain level of accuracy, in practical marine environmental monitoring, risk warning, and management decision-making, it often suffers from issues such as unintuitive interpretation, unclear threshold boundaries, and difficulty in making consistent comparisons across different times and regions. By pre-dividing the weighted summation result into segmented intervals and configuring a fixed risk index within each interval, the complex continuous calculation results can be transformed into discrete risk levels with unified judgment criteria, thereby improving the readability, stability, and executability of the output results. Especially in scenarios involving marine pollution prevention, ecological restoration scheduling, or early warning linkage, managers typically require clear level signals rather than simple continuous scores. Therefore, this design makes it easier to directly integrate risk assessment results with early warning rules, control measures, and management processes.

[0084] This embodiment also provides a computer device applicable to the Internet of Things-based digital twin method for marine ecological risk, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the Internet of Things-based digital twin method for marine ecological risk as proposed in the above embodiment.

[0085] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0086] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the method for realizing a digital twin of marine ecological risks based on the Internet of Things as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0087] In summary, this invention achieves a unified quantitative representation of the transmission of environmental stress, biological degradation feedback, and the switching of risk-dominant factors in marine ecosystems by: unifying physicochemical environmental parameters and biological response indicators into time-related feature sequences; using an improved vine Copula joint distribution model to perform cross-time axis unidirectional matching of the bidirectional hysteresis relationship between the environment and organisms; extracting the positive stress and damage degree of the environment on organisms and the negative backlash and damage degree of organisms on the environment; adaptively generating fusion weights based on the temporal changes of the bidirectional damage degree; and dynamically weighting and summing the physical environmental risk sub-index and the biological response risk sub-index. This improves the temporal consistency, bidirectional causal explanation capability, risk identification accuracy, and practicality of early warning results in marine ecological risk assessment.

[0088] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A digital twin method for marine ecological risk based on the Internet of Things, characterized by: This includes expressing the physicochemical environmental parameters and biological response indicators within the region using two-dimensional curves, and transforming the two-dimensional curves into a time-related feature sequence based on the curve characteristics; Using the improved Copula joint distribution model, we perform cross-time axis multidimensional feature unidirectional matching on two feature sequences to obtain two alignment relationships of unidirectional influence between the two sequences. Based on the two alignment relationships, conditional destructive features representing the substantial damage caused to each other between the two feature sequences are generated respectively. Based on an adaptive weight allocation strategy, the temporal changes of the conditional damage characteristics are transformed into weight parameters. Using the weight parameters, the two risk sub-indices generated by the conditional damage characteristics are dynamically weighted and summed, and the summation result is mapped to the final marine ecological risk index.

2. The method for digital twin of marine ecological risk based on the Internet of Things as described in claim 1, characterized in that: The curve features include mapping each type of physicochemical environmental parameter data and each type of biological response index data to a two-dimensional coordinate system with time as the horizontal axis and numerical value as the vertical axis, and using a spline interpolation algorithm to fit and generate a continuous two-dimensional curve. Based on the local geometric shape change characteristics of the continuous change curve in adjacent time intervals, shape change detection is performed on each of the continuous change curves to identify the morphological change positions where the local geometric similarity is lower than a preset threshold, and an independent timestamp is inserted on the horizontal axis coordinate corresponding to each morphological change position to realize the dynamic interval division of a single curve. For each physicochemical environmental parameter and each biological response index, the independent timestamps of all curves are combined by taking the union of the sets to obtain the global environmental timestamp set and the global biological timestamp set.

3. The method for digital twin of marine ecological risk based on the Internet of Things as described in claim 2, characterized in that: The global environmental timestamp set and the global biological timestamp set are inserted into each type of physicochemical environmental parameter data and each type of biological response index data, respectively. The physicochemical environmental parameters and biological response indexes are sliced ​​according to the inserted timestamps to obtain the slice sequence of physicochemical environmental parameters and biological response indexes. After encoding the multi-source data in each slice, characteristic sequences of physicochemical environmental parameters and biological response indicators were obtained.

4. The method for digital twin of marine ecological risk based on the Internet of Things as described in claim 3, characterized in that: The improved Copula joint distribution model includes constructing a two-dimensional grid matrix by using the time axis of the physicochemical environmental parameter feature sequence as the horizontal axis and the time axis of the biological response index feature sequence as the vertical axis. For each intersecting coordinate node in the two-dimensional grid matrix, the physical environment integrated coding element and the biological response integrated coding element under the corresponding time slice are extracted. The multidimensional feature joint dependency degree between the two is calculated using the vine Copula joint distribution model. The joint dependency degree is then reverse-mapped to generate the basic matching cost of the coordinate node.

5. The method for digital twin of marine ecological risk based on the Internet of Things as described in claim 4, characterized in that: The base matching matrix is ​​copied into a first cost matrix and a second cost matrix; In the first cost matrix, a positive delayed causal penalty is applied: when the physical environment time represented by the horizontal axis lags behind the biological response time represented by the vertical axis, the cross-axis node is given an infinite penalty cost. In the second cost matrix, a reverse delayed causal penalty is applied: when the biological response time represented by the vertical axis lags behind the physical environment time represented by the horizontal axis, an infinite penalty cost is assigned to the cross-axis node. The dynamic programming path optimization rule in the dynamic time warping algorithm is adopted to search for the continuous coordinate point sequence with the minimum global cumulative matching cost from the time start point to the end point in the first cost matrix and the second cost matrix after applying the penalty. The two optimal time warping paths obtained are output as the first alignment relationship and the second alignment relationship representing the unidirectional influence between the two sequences.

6. The method for digital twin of marine ecological risk based on the Internet of Things as described in claim 5, characterized in that: The conditional destructiveness characteristics include positive pressure destructiveness and negative backlash destructiveness; Under the optimal time warping path determined by the first alignment relationship, based on the vine Copula joint probability density function, the conditional expected loss of the multidimensional characteristics of biological response indicators under extreme adverse conditions of physicochemical environmental parameters is calculated, and the conditional expected loss is quantified as the positive pressure damage degree. Under the optimal time warping path determined by the second alignment relationship, based on the vine Copula joint probability density function, the expected deterioration of the multidimensional characteristics of the physicochemical environmental parameters of the biological response index under extreme decay conditions is calculated, and the expected deterioration is quantified as the degree of reverse backlash damage.

7. The method for digital twin of marine ecological risk based on the Internet of Things as described in claim 6, characterized in that: The adaptive weight allocation strategy includes extracting the time series sequence of the positive forcing destruction degree and the time series sequence of the negative backlash destruction degree within a continuous time sliding window; The first derivative of the time series of the positive stress damage degree and the time series of the negative backlash damage degree are respectively obtained with respect to the time variable to obtain the positive damage change rate characterizing the rate of environmental stress intensification and the negative damage change rate characterizing the rate of ecological backlash intensification. A weight ratio allocation equation is constructed, and the ratio of the physical fusion weight parameter corresponding to the physical environment risk sub-index to the biological fusion weight parameter corresponding to the biological response risk sub-index is set to be equal to the ratio of the absolute value of the positive damage change rate to the absolute value of the negative damage change rate. Introduce a weight normalization constraint: set the sum of the physical fusion weight parameter and the biological fusion weight parameter at any time to be 1; Solve the equations for weight ratio allocation and the weight normalization constraints simultaneously to output the physical fusion weight parameters and the biological fusion weight parameters at the current evaluation time.

8. The method for digital twin of marine ecological risk based on the Internet of Things as described in claim 7, characterized in that: After quantifying the real-time physicochemical environmental parameters using a preset linear function, the positive stress damage degree is used as a modulating factor of the intensity of damage exerted by the environment on organisms. This is coupled with the quantification results of the physicochemical environmental parameters to obtain the physical environment risk sub-index. After quantifying the real-time biological response index through a preset encoder, the degree of reverse backlash damage is used as a modulation factor for the intensity of reverse disturbance of the environment caused by biological changes. This factor is coupled with the quantification result of the biological response index to obtain the biological response risk sub-index. Wherein, the coupling calculation is expressed as: R=E×(1+D); R represents the coupling calculation result; E represents the quantitative result of physicochemical environmental parameters or biological response indicators; D represents the degree of positive stress damage or the degree of reverse backlash damage; The marine ecological risk index includes a fixed risk index configured in each segmented interval of a pre-defined weighted summation result; when the weighted summation result falls into the corresponding interval, the corresponding risk index is output.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the Internet of Things-based digital twin method for marine ecological risk as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the Internet of Things-based digital twin method for marine ecological risk as described in any one of claims 1 to 8.

Images