Intelligent decision support system driven by environment big data

Abstract

The invention discloses a smart decision support system driven by environmental big data, which integrates multi-source heterogeneous data to construct a knowledge base through an environmental data fusion module, realizes accurate prediction by combining a mechanism model and an AI algorithm through a scene deduction module, and generates an optimal scheme set by adopting a multi-objective optimization algorithm through a smart decision generation module. And the decision suggestion output module provides dynamic interactive decision support. The system realizes longitudinal transition from a data layer to an intelligent layer, has the advantages of multi-source data intelligent fusion, comprehensive and accurate scene prediction, man-machine cooperation dynamic adjustment and the like, and remarkably improves the scientificity and response speed of complex environment decision making.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of environmental management and big data technology, and particularly relates to a wisdom decision support system for macro-environmental decision support by using artificial intelligence and big data. BACKGROUND

[0002] In the field of macro-environmental management such as climate change, energy planning and urban governance, the decision-making process usually relies on the experience of experts. In the face of massive multi-source heterogeneous environmental data (such as sensor monitoring data, satellite remote sensing images, economic indicators, social network public opinion, etc.), there is a lack of effective intelligent systems to convert these data into high-level wisdom for decision-making. Although some environmental simulation and data analysis tools have appeared in the prior art, they are often limited to a single field or can only provide local analysis results, and cannot automatically extract strategic decision-making wisdom from the whole to cope with complex environmental problems. The traditional method lacks data support and has slow response speed, and it is difficult to timely and scientifically formulate long-term strategies to cope with climate change and major environmental events. The conversion from data to wisdom (i.e. extracting feasible decision-making schemes from massive facts) is still a blank in the current patent layout, and it is also a highly challenging and valuable technical direction. SUMMARY

[0003] The present application aims to overcome the above-mentioned deficiencies and provide an environmental big data driven wisdom decision support system, which can automatically extract knowledge from complex multi-source environmental data and generate optimized strategy schemes, helping governments or organizations to efficiently and scientifically cope with complex environmental problems such as climate, energy, disaster, etc. Through data fusion, scenario prediction and multi-objective optimization, the system realizes the vertical transition from the data layer to the wisdom layer, providing intelligent and semi-automatic support for macro-environmental management decision-making.

[0004] To achieve the above-mentioned purpose, the present application provides the following technical scheme: an environmental big data driven wisdom decision support system, comprising:

[0005] An environmental data fusion module is used to collect various sources of environmental related data and perform spatio-temporal alignment and semantic integration to form a unified environmental information view. This module fuses multi-modal data such as sensor monitoring data, satellite images, economic and social data, etc. and constructs a structured environmental state knowledge base (such as a multi-level related environmental knowledge network of "city-energy-weather" based on a knowledge graph).

[0006] A scenario deduction and prediction module is configured to simulate and predict the results of different intervention measures or scenario schemes based on the environmental knowledge base. The module combines mechanism models in the environmental field and data-driven machine learning models to simulate the evolution trend of the environmental system under various hypothetical conditions, and outputs the prediction results of the short-term effects and long-term impacts (spillover effects) of each scheme.

[0007] A smart decision generation module is configured to integrate the prediction results and combine the value preferences and strategic objectives provided by the decision maker to generate an optimized set of decision schemes. The module uses multi-objective optimization algorithms to analyze the scenario prediction results and search for a set of Pareto optimal schemes that balance different objectives under the constraint conditions. By introducing the value orientation of human experts (such as economic priority or environmental priority) as a decision parameter, the system can iteratively optimize and filter out several alternative strategic schemes based on knowledge deduction.

[0008] A decision suggestion output module is configured to convert the alternative strategic schemes into intuitive decision support reports for user reference. The module summarizes and presents the expected effects, advantages and disadvantages, risk assessment, etc. of each candidate scheme, and provides a human-computer interaction interface for the decision maker to adjust the decision preferences or constraint conditions. After the decision maker adjusts the target parameters, the system can update and output new optimized schemes in real time to support dynamic adjustment of the decision-making process.

[0009] Preferably, the scenario deduction and prediction module described above is built-in with mechanism simulation models in the environmental field (such as climate models, air pollution propagation models, etc.) and combined with machine learning algorithms to improve the prediction accuracy and adaptability to unconventional scenarios. Preferably, the smart decision generation module described above uses multi-objective optimization methods such as genetic algorithms and particle swarm algorithms to perform global search on candidate schemes, ensuring that a set of decision schemes that balance multiple performance indicators is obtained. Preferably, the environmental knowledge base uses knowledge graph technology to link different types of data entities (such as locations, pollution sources, energy facilities, populations, etc.) and their relationships, thereby supporting cross-domain causal relationship analysis and scenario simulation.

[0010] Compared with the prior art, the present application has the following advantages:

[0011] Intelligent fusion of multi-source data: the system can integrate multi-source heterogeneous data such as Internet of Things sensors, remote sensing images, historical statistics, and social networks, automatically complete data cleaning, alignment, and semantic fusion, and construct a unified environmental knowledge base. The massive and dispersed data are thus organically combined, laying an objective foundation for comprehensive decision-making.

[0012] Comprehensive and accurate scenario prediction: By combining domain mechanism models and data-driven models, the system can simulate the evolution of complex environmental systems under different intervention measures, providing both short-term direct impact predictions and assessments of long-term cascading effects and external influences, allowing decision-makers to foresee the long-term consequences of their plans.

[0013] Automatic generation of optimized decisions: Using multi-objective optimization and artificial intelligence algorithms, the system automatically selects the optimal combination of scenarios that meet multiple objectives from a vast space of scenario results. Compared to human experience, the "intelligent decision-making generation module" can explore more options and quantitatively trade off indicators, producing more scientific and reasonable strategic recommendations.

[0014] Human-machine collaboration and dynamic adjustment: The invention incorporates the value preferences of human decision-makers as inputs into the decision-making generation process, ensuring that the final plan aligns with the decision-makers' strategic goals and values. The decision-making suggestion output module provides a friendly interface that allows decision-makers to adjust constraints and obtain updated plans in real-time, enabling human-machine collaborative decision optimization and improving decision flexibility.

[0015] Improving the scientific nature and response speed of decisions: Through the above automatic process from data to knowledge to wisdom, the decision-making process for complex environmental problems is shortened from weeks to days or even real-time interaction, significantly improving the response speed and scientific basis of major decisions. The decision-making process is supported by data and models, reducing subjective randomness and enhancing the reliability and explainability of decisions.

[0016] In summary, the system of the invention covers the complete path from raw data to decision-making wisdom, realizing a new architecture that integrates data layer (D) → information layer (I) → knowledge layer (K) → wisdom layer (W) → decision-making intent / preference layer (P) in a step-by-step refinement process. This not only fills the technical gap in the field of macro-environmental wisdom decision support, but also verifies the effectiveness of the "DIKWP" layered model combined with TRIZ innovation methods in complex system decision-making, with significant practical application value. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 is the overall architecture diagram of the environment big data driven wisdom decision support system of the invention. DETAILED DESCRIPTION

[0018] The technical solutions of the invention will be described in detail below with reference to the accompanying drawings. It should be emphasized that these embodiments are used to explain the invention, but do not limit the scope of the invention. Those skilled in the art can make various improvements or changes without departing from the principles of the invention, and these improvements or changes fall within the scope of the invention.

[0019] As Figure 1As shown, the environmental big data-driven intelligent decision support system provided by this invention includes an environmental data fusion module, a scenario simulation and prediction module, an intelligent decision generation module, and a decision suggestion output module. These modules are connected through internal data interfaces and work collaboratively to transform massive amounts of environmental data into feasible intelligent decision-making solutions. The specific functions and implementation methods of each module are described below.

[0020] The environmental data fusion module is used to collect and integrate multi-source environmental data to form a unified knowledge representation. In one embodiment, the module connects to various data source interfaces, including: an IoT sensor network interface (to acquire real-time monitored environmental data, such as temperature, rainfall, and air quality sensor data), a remote sensing satellite imagery interface (to acquire macro-environmental information such as land use, vegetation cover, and pollution diffusion), a government open database interface (to acquire statistical data such as economic indicators, population distribution, and energy consumption), and a social media and public participation platform interface (to acquire public opinion information related to environmental events). The module performs spatiotemporal alignment (e.g., mapping data from different times and spatial resolutions to the same time axis and geographic grid) and semantic standardization processing on the acquired data (e.g., using a unified environmental ontology to standardize the indicator names and units of each data point).

[0021] The preprocessed multi-source data is further transformed into a structured environmental knowledge base in this module using knowledge graph construction technology. Specifically, the module extracts key entities and their relationships from information from different data sources, such as entities like cities, factories, and power plants, and the relationships between them (emissions, energy supply, impact relationships, etc.). Knowledge graphs are used to link data at the "city-energy-weather-ecology" level, forming a networked environmental state model. Through this knowledge base, the system can represent knowledge such as "the correlation between industrial emissions and local PM2.5 concentration in a certain city" or "the impact of changes in forest cover on regional climate." This provides a unified semantic environment and initial conditions for subsequent scenario extrapolation.

[0022] Furthermore, in a preferred embodiment, the environmental data fusion module also includes a big data storage and computing unit, supporting distributed storage and rapid retrieval of historical environmental data, and incorporating data cleaning and anomaly detection mechanisms to ensure the objectivity and reliability of data entering the knowledge base. Through the above processing, the system of this invention establishes a connection between the I (information layer) and K (knowledge layer), realizing the transformation from raw data to environmental knowledge, laying a solid data foundation for intelligent decision-making.

[0023] The scenario simulation and prediction module, based on the aforementioned environmental knowledge base, simulates and calculates given decision scenarios or intervention measures to predict their potential effects. This module consists of a scenario configuration unit, a simulation engine, and a predictive analysis unit.

[0024] First, the scenario configuration unit receives scenario parameters set by the decision-maker or a set of scenarios automatically generated by the system. Scenario parameters include various possible decision-making measures or changes in external conditions, such as setting greenhouse gas emission reduction targets, adjusting the proportion of renewable energy adoption, enacting environmental policies and regulations, or the occurrence of natural events (such as extreme weather or geopolitical events). The configuration unit uses the current environmental state stored in the environmental knowledge base as a baseline, applies scenario parameters on this basis, and generates different initial scenario conditions for the simulation engine to use.

[0025] The simulation engine includes mechanistic and data-driven models across multiple environmental domains to simulate the evolution of environmental systems under various scenario conditions. In practical implementation, the simulation engine can integrate, but is not limited to, the following models:

[0026] Climate and atmospheric models: such as global climate models (GCM), regional climate models, air quality models, etc., are used to simulate environmental impacts such as changes in greenhouse gas concentrations, rising temperature trends, and the spread of air pollution.

[0027] Energy and economic models: such as energy supply and demand balance models and macroeconomic models, are used to assess the impact of emission reduction measures on energy structure and economic development, such as changes in GDP growth rate and employment impact.

[0028] Disaster and ecological models: such as hydrological models, forest ecological models, sea-level rise models, etc., are used to predict the impact of different measures on natural disaster risks and ecosystem services.

[0029] Machine learning prediction models: Predictive models trained on historical big data are used to provide auxiliary predictions in fields where mechanistic models are lacking or difficult to accurately simulate. For example, time-series prediction models can be used to predict energy consumption trends, or deep learning models can be used to predict the frequency of extreme weather events.

[0030] The simulation engine runs the aforementioned models (which can be run in parallel) under each scenario, obtaining predictions for key indicators in the short term (e.g., within 1-3 years) and long term (e.g., 10 years or several decades later). Short-term results include directly observable effects, such as the reduction in total carbon emissions and the extent of air quality improvement; long-term results include indirect impacts, such as the spillover effects of technological innovation brought about by policies on the future economy and the long-term benefits of climate improvement to public health.

[0031] The predictive analysis unit summarizes simulation results and performs aggregated calculations and uncertainty analysis on key indicators. Considering the inherent uncertainty in model predictions, the predictive analysis unit can assess the reliability of the results or provide confidence intervals through methods such as Monte Carlo simulation. This information will serve as the foundational data input for subsequent decision optimization.

[0032] Through the scenario deduction and prediction module, an initial leap from the knowledge layer (K) to the wisdom layer (W) is achieved, that is, generating dynamic scenario-based intelligent information from a static knowledge network. The prediction results of multiple alternative scenarios constitute a vast decision search space, covering all possible future states.

[0033] The intelligent decision generation module is the core of the system of this invention, used to automatically find the optimal combination of decision schemes based on the scenario prediction results. This module mainly includes a target setting unit, an optimization solution unit, and a scheme evaluation unit.

[0034] The goal-setting unit receives strategic objectives and value preference parameters provided by decision-makers. These parameters define the problem framework for decision optimization, such as the objective functions that need to be considered simultaneously (e.g., "minimize carbon emissions," "minimize negative impacts on GDP," "increase the proportion of renewable energy," "ensure employment," etc.) and the importance or weight of each objective. If decision-makers do not provide explicit preferences, the system can also provide default multi-objective settings or guide users to set them. Furthermore, the goal-setting unit allows setting necessary constraints, such as "carbon emissions must decrease by more than X% by 2030" or "annual GDP growth rate must not be lower than Y%", to ensure that the generated solution meets basic requirements.

[0035] The optimization unit uses a multi-objective optimization algorithm to search for the optimal policy combination in all scenario prediction result spaces. Since scenario prediction provides a mapping relationship between decision measures and outcome indicators, this problem can be viewed as a process of finding the Pareto optimal solution in a multi-dimensional objective space. Algorithms that the optimization unit can employ include, but are not limited to: genetic algorithms, multi-objective evolutionary algorithms (such as NSGA-II), particle swarm optimization, simulated annealing, or exact algorithms for specific linear / nonlinear programming problems. The algorithm encodes scenario combinations (i.e., a set of specific decision measures) as solutions and uses the performance calculated from each prediction indicator as the objective vector of the solution, iteratively optimizing the population to approximate the Pareto front.

[0036] During the optimization process, the system guides the decision-maker's choices by weighing different objectives. For example, when the decision-maker prioritizes environmental protection, the weight of environmental indicators in the evaluation function can be increased, causing the algorithm to favor the solution with better environmental performance; conversely, the opposite applies when prioritizing economic benefits. Through this integration of human and machine values, the optimization process generates a set of balanced solutions that align with the decision-maker's value orientation, rather than a single solution. This set of solutions represents the Pareto optimal solution set, where no further improvement is possible across all objectives.

[0037] The scheme evaluation unit further analyzes and filters the candidate scheme set generated by optimization. First, this unit calculates key indicator values ​​for each scheme based on the prediction model results, such as cumulative emission reduction, economic cost, and risk level. It can then use decision analysis methods (such as TOPSIS, Analytic Hierarchy Process, AHP, etc.) to comprehensively score or rank the schemes. Second, the scheme evaluation unit marks the advantages and disadvantages of each scheme: for example, scheme A has the best emission reduction effect but a higher economic cost, scheme B has stable economic performance but moderate environmental benefits, and scheme C represents a compromise between the two, allowing decision-makers to understand the trade-offs. Finally, the evaluation unit selects several (e.g., 3-5) of the most representative preferred schemes and submits them to the decision suggestion output module. It should be noted that if the candidate schemes are concentrated in a narrow range, the system can also use a guided algorithm to diversify solutions, ensuring that the schemes provided to decision-makers have diversity in objective trade-offs and cover different strategic options with different focuses.

[0038] Through the intelligent decision generation module, this invention achieves a crucial leap from knowledge and contextual intelligence to high-level decision-making intelligence, enabling the system to replace manual labor in identifying optimal solutions from massive amounts of data and complex scenarios. This process demonstrates how the W (intelligence layer) is further enhanced under the guidance of decision-makers' preferences and fed back to the knowledge layer (through constraints and screening of alternative solutions applied to the knowledge / context model correction), thus forming a closed-loop intelligent decision support system.

[0039] The decision suggestion output module is responsible for transforming the generated decision solutions into an easily understandable and operable format for the final decision-maker. This module consists of two parts: a report generation unit and an interactive feedback unit.

[0040] The report generation unit automatically generates a decision support report based on the preferred solution set provided by the solution evaluation unit. The report is presented in language and visual format easily understood by human decision-makers, and includes: a description of the combination of measures for each proposed solution, a summary of the corresponding predicted effects, comparative tables or charts of various target indicators (e.g., emission reductions, economic impact, risk levels), a comparison of the advantages and disadvantages between solutions, and a highlighting of potential implementation risks and uncertainties. The report can also provide long-term impact assessments, such as using concise text and diagrams to illustrate the environmental and economic status of a solution 5 or 10 years from now. Through this information, decision-makers can comprehensively understand the consequences and merits of each candidate solution.

[0041] The interactive feedback unit provides decision-makers with a human-computer interface, allowing them to adjust input preferences and view real-time updates. Specifically, the interface allows users to modify strategic goal weights, add or remove constraints, and even manually specify which measures must or must not be adopted, triggering the system to recalculate optimization. The interactive interface lowers the technical barrier through intuitive control panels (such as sliders for adjusting weights and checkboxes for selecting constraints), enabling non-technical decision-makers to participate in customizing solutions. Simultaneously, when users adjust parameters, the system quickly invokes the aforementioned modules (especially the optimization solution unit) to incrementally calculate new optimal solutions and provides real-time feedback on the interface. This makes the decision-making process transparent and controllable: decision-makers can repeatedly weigh different objectives, see corresponding changes in solutions, and ultimately choose the strategy that best meets their expectations.

[0042] In a practical application scenario, this module also supports collaborative decision-making: multiple decision-makers or departments can jointly view solution reports, discuss and evaluate them through a networked collaborative platform, and after reaching a consensus on preferred parameters, the system provides a comprehensive optimization solution. This helps balance the interests of all parties in complex decision-making and form a more consensus-based final decision.

[0043] In summary, the decision suggestion output module presents the intelligent decisions generated by the system in a user-friendly format and integrates them into the human decision-making process. This achieves the transformation from the wisdom layer (W) back to the information layer (I), that is, converting high-level wisdom into specific and actionable information and solutions, enabling the final decision-maker to efficiently obtain insights and put them into action.

[0044] The following application example illustrates the workflow of the system of the present invention. Example: Optimization of Urban Climate Action Plans. Assume a city government wants to develop a climate change response strategy for the next ten years, including reducing carbon emissions and improving air quality, while maintaining moderate economic growth. Decision-makers face various possible measures (such as increasing the proportion of renewable energy, restricting energy-intensive industries, improving public transportation, and afforestation) as well as uncertain external factors (such as macroeconomic fluctuations and changes in the frequency of extreme weather events).

[0045] Data Fusion: The system's environmental data fusion module collects relevant historical data of the city, including energy consumption and carbon emission statistics, air pollution monitoring data, economic and population data from the past few years, as well as information such as urban heat island intensity and green coverage obtained from satellite remote sensing. It constructs an urban environmental knowledge graph, revealing the relationship between energy structure and emissions, meteorological conditions, and socio-economic factors.

[0046] Scenario Configuration: Decision-makers input several proposed solutions into the interface, such as: "Solution 1: Strict Environmental Protection" includes measures such as increasing the proportion of renewable energy power generation and implementing industrial emission caps; "Solution 2: Balanced and Stable" adopts moderate environmental protection measures and encourages green industries; "Solution 3: Economic Priority" focuses on economic development while maintaining relatively relaxed environmental protection measures. The system transforms these solutions into scenario parameter combinations that the model can recognize and automatically generates intermediate solutions to enrich the optimization options.

[0047] Simulation Prediction: The scenario simulation and prediction module utilizes the integrated urban energy-climate model and related machine learning models to predict indicators such as total carbon emissions, annual PM2.5 concentration, and GDP growth rate for each scenario in 2025, 2030, and 2035. For example, the system predicts that Scenario 1 will reduce carbon emissions by 30% and PM2.5 by 40% by 2030, but with a cumulative GDP loss of 5%; Scenario 3 will reduce carbon emissions by 10%, PM2.5 by 15%, and have the highest relative GDP growth; Scenario 2's indicators are in the middle. The system also provides a prediction reliability analysis, indicating that the economic impact estimate for Scenario 1 has higher uncertainty.

[0048] Optimization Decision: The intelligent decision generation module performs multi-objective optimization based on the city's proposed goals ("reduce carbon emissions by at least 20% by 2030, and achieve an average annual GDP growth of no less than 2%) and preferences (balancing environmental protection and economic development), seeking trade-offs among Schemes 1-3 and automatically generated variant schemes. The final output includes several optimized schemes, such as: Scheme A (Environmentally Enhanced): 25% carbon emission reduction, slight impact on GDP; Scheme B (Balanced Optimization): 20% carbon emission reduction, no significant loss in GDP and moderate improvement in air quality; Scheme C (Economically Conservative): 15% carbon emission reduction, slight increase in GDP. All these schemes satisfy the minimum constraints and each has its own emphasis, forming a Pareto solution set.

[0049] Report Output: The decision-making recommendation output module generates a report that presents a list of measures and a comparison of the effects of options A, B, and C. For example, it displays a bar chart comparing carbon emissions, air quality, and GDP in 2030, with accompanying text descriptions such as "Option A is the most environmentally friendly but has the highest economic cost, Option C promotes the economy the most but has a relatively weaker environmental effect, and Option B achieves a balance between the two." The report also includes implementation risk analysis (e.g., the risks of Option A lie in business backlash and employment pressure) and provides interactive controls that allow government policymakers to fine-tune the weights and view changes in the options.

[0050] Decision-making: After multiple rounds of discussions, city decision-makers can use an interactive interface to adjust their preferences (such as slightly increasing the economic weight), see the slightly optimized indicators for Option B, and ultimately choose Option B as the implementation route. At the same time, they also save Option A as a more aggressive option for subsequent evaluation. The entire process is completed with the support of data and models, taking significantly less time than traditional manual decision-making discussions, and the results are well-documented.

[0051] The above embodiments clearly demonstrate the working method and advantages of the system of the present invention in decision-making in complex environments. It significantly reduces the burden on decision-makers in manually sifting and synthesizing massive amounts of information, providing an end-to-end intelligent decision support platform from data acquisition and scenario analysis to decision recommendation. The system of the present invention has broad application prospects in many fields, including environmental governance, climate change response, energy strategy formulation, urban planning, and the design of major public policies.

[0052] This invention proposes an environmental big data-driven intelligent decision support system that realizes a complete process of fusing and extracting knowledge from multi-source environmental data, simulating complex scenarios, generating intelligent optimization decision solutions, and providing human-computer interactive decision output. This solves the technical problem of the lack of data-driven intelligent decision support in existing macro-environmental management. Through the application of this system, massive environmental "big data" is transformed into valuable decision-making "wisdom," significantly improving the scientific rigor, efficiency, and foresight of decisions made in response to complex challenges such as climate change. The system has good scalability and adaptability, allowing for the introduction of new models and data sources according to different decision-making scenarios, meeting the intelligent decision-making needs of various fields. In summary, this invention has outstanding innovation and practicality, and possesses a high probability of being granted a patent.

Claims

1. An environment big data driven intelligent decision support system, characterized in that, Comprise: An environmental data fusion module for collecting sensor monitoring data, satellite remote sensing images, economic indicators and social network public opinion data, constructing an environmental knowledge base through spatio-temporal alignment and semantic integration; A scenario deduction and prediction module for simulating and deducing different intervention measures based on the environmental knowledge base, predicting short-term effects and long-term impacts by combining field mechanism models and machine learning algorithms; A smart decision generation module for generating a Pareto optimal solution set by combining the prediction results and decision-maker value preferences, using a multi-objective optimization algorithm; A decision suggestion output module for converting candidate solutions into visual decision reports, supporting parameter adjustment and real-time solution updating through a human-computer interaction interface.

2. The system of claim 1, wherein, The environmental data fusion module correlates multi-dimensional data entities such as "city-energy-weather-ecology" through knowledge graph technology to form a networked environmental state model.

3. The system of claim 2, wherein, The scenario deduction module integrates climate models, energy economy models, disaster ecology models and machine learning prediction models to generate multi-dimensional prediction indicators through parallel simulation calculation.

4. The system of claim 3, wherein, The smart decision generation module uses NSGA-II genetic algorithm or particle swarm optimization algorithm to search for Pareto frontier solutions that balance economic, environmental and social benefits under constraint conditions.

5. The system of claim 4, wherein, The decision suggestion output module includes a TOPSIS comprehensive evaluation unit that weights and scores candidate solutions based on multi-dimensional indicators through an analytic hierarchy process, and marks the advantages and disadvantages of each solution.

Images