A multi-dimensional collaborative intelligent energy management system

Abstract

This invention relates to a multi-dimensional collaborative intelligent energy management and control system. The system includes: a data acquisition unit, a data processing unit, a digital twin and risk assessment unit, and a multi-objective collaborative optimization and visualization unit. It acquires multi-dimensional raw data and integrates it into a multi-source dataset; processes the multi-source dataset to generate a standardized dataset and a multi-dimensional feature set; reconstructs the physical topology and equipment parameters of the energy system, constructs a real-time synchronous digital twin, and combines the multi-dimensional feature set to obtain the operating status, analyze risk factors, and output early warning results, integrating and generating an operating status assessment report; obtains optimization benchmarks based on the assessment report, constructs a multi-agent reinforcement learning system using the multi-dimensional feature set to obtain energy dispatching strategies, decomposes them into executable control commands and obtains execution results, and simultaneously builds a multi-zone monitoring interface to visualize the entire energy management and control process data. This system achieves energy management and control using a multi-dimensional collaborative mechanism.

Description

Technical Field

[0001] This invention belongs to the field of smart energy management technology, specifically relating to a multi-dimensional collaborative smart energy management and control system. Background Technology

[0002] The following areas still need improvement in energy management: Driven by the global energy transition and dual-carbon goals, integrated energy systems are gradually developing towards diversification, intelligence, and decarbonization, encompassing multiple aspects such as energy storage, power grids, and new energy power generation, and involving multi-dimensional management needs related to safety, economy, energy efficiency, and carbon emissions. However, the technical architecture and functional design of existing energy management systems are ill-suited to this complex scenario, presenting numerous problems that urgently need to be addressed, as follows: I. Fragmented control dimensions and lack of multi-objective coordination Existing energy management systems often employ a single-dimensional control model, with core functional modules such as safety monitoring, economic dispatch, energy efficiency analysis, and carbon emission management operating independently, lacking a unified intelligent hub for overall coordination. For example, the safety management module focuses only on threshold monitoring of equipment operating parameters, while the economic dispatch module only considers cost control under electricity price fluctuations. The lack of data sharing and goal coordination between the two often leads to situations where excessive restrictions on equipment operating load are imposed to ensure safety, resulting in reduced energy efficiency and increased operation and maintenance costs; or carbon emission constraints are ignored in pursuit of economic benefits, contradicting low-carbon development goals. This fragmented control approach fails to balance the priorities of multi-dimensional objectives, making it difficult to achieve the comprehensive optimal operation of the energy system.

[0003] Second, insufficient data processing capabilities limit the effectiveness of supporting decision-making. Integrated energy systems involve multi-source heterogeneous data, including electrical safety parameters (temperature, insulation, arc), equipment operating status (auxiliary power consumption of energy storage systems, liquid cooling control mode), market electricity prices, carbon emission factors, and energy flow. This data is scattered, varied in format, and massive in volume. Existing systems lack efficient data integration mechanisms. On the one hand, they fail to effectively remove redundant data collected repeatedly, use simplistic criteria for identifying abnormal data, and lack targeted methods for completing missing data, resulting in inconsistent data quality. On the other hand, they fail to deeply explore the correlation characteristics between multi-dimensional data, remaining only at the level of simple data display and preliminary statistics, unable to transform data resources into effective information to support decision-making, thus lacking precise data basis for scheduling strategy formulation.

[0004] III. Delayed risk warnings and passive safety management Traditional energy management systems rely heavily on manual inspections and post-fault analysis for risk monitoring, lacking the ability to proactively predict potential risks. While some systems possess basic parameter monitoring capabilities, they fail to integrate this with the overall operational status of the energy system for coordinated analysis, and they do not incorporate advanced simulation technologies. For example, regarding risks such as thermal runaway in energy storage devices and grid overload, alarms are only issued after parameters exceed safety thresholds. The systems cannot predict risk evolution trends through simulated operational conditions, resulting in a perpetually reactive approach to safety management. This makes it difficult to take early intervention measures to prevent faults, posing a significant threat to the stable operation of the energy system.

[0005] IV. Low level of intelligent decision-making and poor scenario adaptability Existing energy management systems often rely on fixed rules or simple algorithms for scheduling strategies, lacking the ability to learn autonomously and adjust dynamically. Faced with complex scenarios such as fluctuating electricity prices in the energy market, changes in equipment operating status, and adjustments to carbon emission policies, fixed rules struggle to quickly adapt to the control requirements under different operating conditions. Furthermore, the systems fail to consider the weight differences of multi-dimensional objectives across various application scenarios. For example, the safety priorities, economic demands, and carbon emission constraints of industrial park energy storage systems differ from those of urban power grid energy storage systems. Existing systems cannot flexibly adjust their decision-making logic according to the characteristics of each scenario, resulting in insufficient practicality and optimality of the scheduling strategies, and hindering the full realization of the operational potential of energy equipment.

[0006] Existing systems generally suffer from a disconnect between decision-making and execution, and a fragmentation between data and visualization. On the one hand, optimization strategies are mostly macro-level guidance schemes, failing to be broken down into control commands that equipment can directly recognize and execute. They also lack compatibility with energy equipment control protocols, leading to significant manual intervention during strategy implementation, resulting in low efficiency and a high risk of operational errors. On the other hand, visualization displays focus only on single-dimensional data or localized operational status, lacking a multi-partition, end-to-end integrated display interface. This makes it difficult for maintenance personnel to intuitively grasp the overall system operational status, risk warning information, and the effectiveness of optimization strategy execution, greatly hindering remote monitoring and rapid decision-making, and significantly increasing maintenance costs and management complexity. Summary of the Invention

[0007] To address the aforementioned problems in the existing technology, this invention provides a multi-dimensional collaborative intelligent energy management and control system; The objective of this invention can be achieved through the following technical solutions: a data acquisition unit, a data processing unit, a digital twin and risk assessment unit, and a multi-objective collaborative optimization and visualization unit; The data acquisition unit is used to acquire raw data from multiple dimensions; The data processing unit processes the multi-dimensional raw data according to preset classification rules to obtain a multi-source raw dataset; it removes outliers and fills in missing values ​​in the multi-source raw dataset, and extracts multi-dimensional features to obtain a standardized dataset and a multi-dimensional feature set. The digital twin and risk assessment unit reconstructs the physical topology of the energy system and the parameters of energy equipment based on the standardized dataset, constructs a digital twin that is synchronized with the physical system in real time, calls the multi-dimensional feature set to drive the digital twin, and obtains the operating status of the energy conditions; it analyzes risk factors and obtains risk warning results by combining the multi-dimensional feature set; and it integrates the operating status and the risk warning results to generate an operating status assessment report. The multi-objective collaborative optimization and visualization unit sets safety constraint boundaries based on the risk warning results and obtains energy optimization benchmarks in the corresponding dimensions by combining the operation status assessment report; it constructs multi-agent reinforcement learning using the multi-dimensional feature set and obtains energy scheduling strategies through collaborative game among the agents; it decomposes the energy scheduling strategies into executable control instructions, verifies and drives the execution of the control instructions, and obtains the instruction execution results; it constructs a multi-zone monitoring interface to visualize the entire process data of energy management and control.

[0008] As a preferred technical solution of the present invention, the specific process of obtaining multi-dimensional raw data includes: obtaining the target data type based on the energy system equipment deployment list, and establishing communication connections with the energy storage system, the power grid platform, and the carbon emission database; based on the preset sampling frequency of the communication connection channel, collecting multi-dimensional raw data in real time and synchronously recording the data collection timestamp and energy equipment identifier, thereby obtaining a multi-dimensional raw data set with traceability information.

[0009] Specifically, the process of obtaining the multi-source raw dataset includes: classifying the multi-dimensional raw data with traceability information, removing redundant data collected repeatedly, and integrating the classification results to obtain a structured multi-source raw dataset.

[0010] Specifically, the outlier removal process involves: establishing an evaluation criterion for outlier data based on the structured multi-source original dataset, detecting the corresponding dataset, obtaining the detection results, and removing the outlier data.

[0011] Specifically, the missing value completion operation includes: based on the dataset after outlier removal, statistically analyzing the missing locations and proportions of statistical data fields; based on the missing proportions and data types, selecting an appropriate completion algorithm, and calculating the completion value using the statistical characteristics of the missing data; and using the calculated completion value to fill in the corresponding dataset to obtain a standard dataset.

[0012] Specifically, the process of constructing a digital twin that is synchronized with the physical system in real time includes: extracting the physical topology and equipment parameters of the energy system based on the standard state dataset; using digital twin modeling technology to build a virtual model framework consistent with the physical system structure, establishing a data mapping relationship, and associating the real-time data in the standard state dataset with the virtual model; and synchronizing the operating data of the physical system in real time based on the data mapping relationship to construct a digital twin that is synchronized with the physical system in real time.

[0013] Specifically, the process of obtaining the operating status of the working condition includes: extracting a multi-dimensional feature set based on the standard state dataset and filtering the feature data of the corresponding working condition; inputting the feature data of the working condition into the operation simulation module of the digital twin based on the filtering results to drive the virtual model to reproduce the operation process of the physical system; and extracting the working condition indicators and integrating them to obtain the operating status report based on the simulation operation results of the digital twin.

[0014] Specifically, the process of obtaining risk warning results includes: screening safety feature data based on the multi-dimensional feature set to obtain risk assessment factors; classifying risk levels by combining safety standards and equipment failure thresholds; monitoring the changing trends of the risk assessment factors in real time based on the operating status fed back by the digital twin, and comparing the results of the risk level classification; identifying potential risks based on the comparison results, and generating risk warning results.

[0015] Specifically, the process of obtaining the corresponding energy optimization benchmark includes: setting corresponding safety constraint boundaries based on the risk warning results; extracting the level values ​​of the corresponding dimensions based on the operation status assessment report, obtaining the economic optimization benchmark, and integrating them to obtain the energy optimization benchmark report.

[0016] Specifically, the process of obtaining the energy scheduling strategy includes: constructing a multi-agent reinforcement learning framework based on the multi-dimensional feature set; using the energy optimization benchmark report to preset the reward function of the agents; and coordinating the agents to play games and cooperate through a central coordinator to generate an energy scheduling strategy.

[0017] Specifically, the executable control instructions are as follows: the strategy content is split based on the energy scheduling strategy and the energy equipment type to obtain the operation requirements of the corresponding energy equipment; based on the operation requirements of the corresponding energy equipment and the energy equipment control protocol, the strategy content is transformed to obtain an instruction format that the energy equipment can recognize; based on the instruction format, the parameters for instruction execution are supplemented and sorted using the preset instruction execution logic to obtain a set of control instructions for the energy equipment.

[0018] Specifically, the process of obtaining the instruction execution result includes: transmitting the instructions through a communication module based on the control instruction set; using the feedback mechanism of the energy equipment to receive the operation data of the energy equipment executing the instructions in real time; comparing the target value required by the instruction, analyzing the process data of the instruction execution, and obtaining the instruction execution result.

[0019] The beneficial effects of this invention are as follows: multi-dimensional collaborative management and control, breaking down data and objective barriers: through a unified data collection and processing mechanism, it integrates multi-source data such as safety, economy, energy efficiency, and carbon emissions, and combines multi-agent collaborative optimization to achieve global optimization of multiple objectives, solving the problem of dimensional fragmentation in traditional systems; the digital twin is synchronized with the physical system in real time, which can not only reproduce the operating status, but also combine AI technology to achieve early warning of risks, significantly improving the initiative and accuracy of safety management and control compared with traditional post-event monitoring; the multi-agent reinforcement learning framework has the ability to learn autonomously and adjust dynamically, which can quickly adapt to complex scene changes, generate optimal scheduling strategies, and standardize the instruction conversion and execution process to ensure efficient implementation of decisions; the full-process data visualization display from data collection to instruction execution, combined with the data traceability mechanism, improves the transparency of management and control and the convenience of operation and maintenance, and reduces the cost of manual intervention. Attached Figure Description

[0020] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0021] Figure 1 This is a flowchart illustrating a multi-dimensional collaborative intelligent energy management system according to the present invention. Figure 2 This is a flowchart of the intelligent decision-making process in this invention. Detailed Implementation

[0022] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0023] Please see Figure 1-2 A multi-dimensional collaborative smart energy management and control system includes: a data acquisition unit, a data processing unit, a digital twin and risk assessment unit, and a multi-objective collaborative optimization and visualization unit; The data acquisition unit is used to acquire raw data from multiple dimensions; The data processing unit processes the multi-dimensional raw data according to preset classification rules to obtain a multi-source raw dataset; it removes outliers and fills in missing values ​​in the multi-source raw dataset, and extracts multi-dimensional features to obtain a standardized dataset and a multi-dimensional feature set. The digital twin and risk assessment unit reconstructs the physical topology of the energy system and the parameters of energy equipment based on the standardized dataset, constructs a digital twin that is synchronized with the physical system in real time, calls the multi-dimensional feature set to drive the digital twin, and obtains the operating status of the energy conditions; it analyzes risk factors and obtains risk warning results by combining the multi-dimensional feature set; and it integrates the operating status and the risk warning results to generate an operating status assessment report. The multi-objective collaborative optimization and visualization unit sets safety constraint boundaries based on the risk warning results and obtains energy optimization benchmarks in the corresponding dimensions by combining the operation status assessment report; it constructs multi-agent reinforcement learning using the multi-dimensional feature set and obtains energy scheduling strategies through collaborative game among the agents; it decomposes the energy scheduling strategies into executable control instructions, verifies and drives the execution of the control instructions, and obtains the instruction execution results; it constructs a multi-zone monitoring interface to visualize the entire process data of energy management and control.

[0024] As a preferred technical solution of the present invention, the specific process of obtaining multi-dimensional raw data includes: obtaining the target data type based on the energy system equipment deployment list, and establishing communication connections with the energy storage system, the power grid platform, and the carbon emission database; based on the preset sampling frequency of the communication connection channel, collecting multi-dimensional raw data in real time and synchronously recording the data collection timestamp and energy equipment identifier, thereby obtaining a multi-dimensional raw data set with traceability information.

[0025] In this embodiment, the deployment list of energy system equipment in the industrial park is compiled, and the target data types are clearly defined: covering electrical safety parameters (temperature, insulation performance, and arc-related monitoring data), equipment operation status data (auxiliary power consumption, liquid cooling control mode, and charging and discharging status of energy storage systems, output status and operation mode of new energy power generation devices, energy consumption intensity and working hours of electricity load, etc.), market data (electricity price fluctuation data of the regional power grid), carbon emission data (industry carbon emission factors and actual carbon emission data of the industrial park), and energy flow data (flow and loss data of energy in each link of power generation, storage, transmission, and consumption).

[0026] Establish stable communication connections between the system and energy storage systems within the industrial park, regional power grid platforms, and authoritative industry carbon emission databases. Adaptive communication protocols should be adopted based on the interface specifications of each device and platform to ensure the stability and compatibility of data transmission channels.

[0027] Based on the characteristics and control requirements of different data types, a differentiated sampling frequency is preset: a higher sampling frequency is set for electrical safety parameters and energy flow data with high change frequency, and a relatively lower sampling frequency is set for data with low change frequency such as electricity price and carbon emission factor; during the collection process, the collection timestamp (accurate to the collection time) and the corresponding energy equipment identifier (such as the unit number of the energy storage system, the model and installation location of the power generation device, the region and equipment name of the power load, etc.) are recorded synchronously for each piece of raw data, and finally integrated to form a multi-dimensional raw data set with complete traceability information.

[0028] Specifically, the process of obtaining the multi-source raw dataset includes: classifying the multi-dimensional raw data with traceability information, removing redundant data collected repeatedly, and integrating the classification results to obtain a structured multi-source raw dataset.

[0029] In this embodiment, multi-dimensional raw data with traceability information is categorized according to preset classification rules: by data type (safety parameters, operating status, market, carbon emissions, and energy flow); by equipment affiliation (energy storage system, power generation device, and electricity load); and by time dimension (real-time data and time period statistics). During the classification process, redundant data (such as data repeatedly collected from the same device at the same time by different acquisition channels) is eliminated by comparing the data collection timestamps, equipment identifiers, and data content. The categorized data is then structured and integrated in a unified format to obtain a multi-source raw dataset.

[0030] Specifically, the outlier removal process involves: establishing an evaluation criterion for outlier data based on the structured multi-source original dataset, detecting the corresponding dataset, obtaining the detection results, and removing the outlier data.

[0031] In this embodiment, based on the data attributes of the structured multi-source raw dataset, targeted abnormal data evaluation criteria are established. For numerical data (such as temperature and energy consumption values), reasonable value ranges and fluctuation thresholds are set in conjunction with the physical characteristics of equipment operation and industry-standard practices. For logical data (such as equipment operating modes and charging / discharging states), valid state combination rules are set based on the equipment's working principles and operating logic. The dataset is then examined line by line according to these criteria, identifying data that exceeds reasonable value ranges, exhibits abnormal fluctuations, or does not conform to logical state combinations. This data is then judged as abnormal data and removed.

[0032] Specifically, the missing value completion operation includes: based on the dataset after outlier removal, statistically analyzing the missing locations and proportions of statistical data fields; based on the missing proportions and data types, selecting an appropriate completion algorithm, and calculating the completion value using the statistical characteristics of the missing data; and using the calculated completion value to fill in the corresponding dataset to obtain a standard dataset.

[0033] In this embodiment, a comprehensive statistical analysis is performed on the dataset after outlier removal to identify the missing locations (which specific record or field is missing) and the missing percentage (the proportion of missing data in a particular field relative to the total data in that field). Based on the missing percentage and data type, a suitable completion algorithm is selected: if the missing percentage is low (e.g., below a preset threshold) and the data is numerical, completion is performed using the statistical characteristics of the field (e.g., mean, median, distribution pattern) or its correlation with other relevant fields (e.g., deriving missing energy intensity data through the correlation between energy consumption data and runtime); if the missing percentage is high or the data is categorized (e.g., device operating modes), a completion method based on analogy with similar scenario data is used (e.g., referencing data from similar historical operating conditions of the same device, or data from similar devices under the same operating conditions) to calculate completion values, accurately filling in the missing locations in the dataset and obtaining a complete and high-quality standard dataset.

[0034] Based on a standard dataset, feature engineering methods are used to mine multi-dimensional features related to control objectives such as safety, economy, energy efficiency, and carbon emissions. Features reflecting equipment safety status (e.g., temperature change rate, insulation performance degradation trend, frequency of electric arcs) are extracted from safety parameter data; features reflecting energy utilization efficiency (e.g., charge / discharge conversion efficiency of energy storage systems, power generation utilization rate, overall energy loss rate) are extracted from operating status and energy flow data; features affecting economic costs (e.g., electricity costs at different times, and operation and maintenance costs of energy storage systems) are extracted from market and operating status data; and features reflecting carbon emission reduction effects (e.g., carbon emissions per unit of energy consumption, correlation between the proportion of new energy power generation and carbon emission reduction) are extracted from carbon emission and energy consumption data, thus obtaining a multi-dimensional feature set.

[0035] Specifically, the process of constructing a digital twin that is synchronized with the physical system in real time includes: extracting the physical topology and equipment parameters of the energy system based on the standard state dataset; using digital twin modeling technology to build a virtual model framework consistent with the physical system structure, establishing a data mapping relationship, and associating the real-time data in the standard state dataset with the virtual model; and synchronizing the operating data of the physical system in real time based on the data mapping relationship to construct a digital twin that is synchronized with the physical system in real time.

[0036] In this embodiment, the physical topology of the industrial park's energy system (including the installation locations of energy storage systems, power generation devices, electrical loads, connections between equipment, and the layout of energy transmission lines) and equipment parameters (including the rated operating parameters, performance parameters, and structural parameters of each device, such as the rated capacity and charging / discharging power limits of the energy storage system, and the rated output and efficiency parameters of the power generation device) are extracted from the standard dataset. Digital twin modeling technology is used to build a virtual model framework that is completely consistent with the physical system structure, establishing a precise data mapping relationship between physical quantities and virtual variables—linking and binding real-time data (such as equipment operating parameters and energy flow data) from the standard dataset with the corresponding modules of the virtual model (such as virtual energy storage units and virtual power generation devices).

[0037] Formula description: Let the real-time operating data set of the physical system be P={p1,p2,...,p...} m (m is the number of physical data types), the set of variables corresponding to the virtual model is V={v1,v2,...,v...} m}; Construct a linear mapping relationship v k =f(p k )= +k0, where k1 is the scaling factor and k0 is the offset factor, obtained by fitting calibration data of the physical system and the virtual model; during real-time synchronization, it is achieved through v k (t)= +k0 implements the virtual variable v k With physical data p k Time synchronization (t is the timestamp).

[0038] Based on this data mapping relationship, the operating data of the physical system is synchronized to the virtual model in real time, enabling the virtual model to dynamically reproduce the operating status of the physical system, and ultimately constructing a digital twin that is synchronized in real time with the energy physical system of the industrial park.

[0039] Specifically, the process of obtaining the operating status of the working condition includes: extracting a multi-dimensional feature set based on the standard state dataset and filtering the feature data of the corresponding working condition; inputting the feature data of the working condition into the operation simulation module of the digital twin based on the filtering results to drive the virtual model to reproduce the operation process of the physical system; and extracting the working condition indicators and integrating them to obtain the operating status report based on the simulation operation results of the digital twin.

[0040] In this embodiment, feature data related to the current energy operation of the industrial park (such as current electricity load characteristics, new energy power generation output characteristics, and initial state characteristics of the energy storage system) are selected from a multi-dimensional feature set and input into the operation simulation module of the digital twin. The virtual model is driven by AI algorithms to accurately reproduce the real-time operation of the physical system under the current conditions, including changes in the working status of each device, energy transmission paths and flow distribution, and energy losses at each stage. Based on the simulation results of the digital twin, key operating condition indicators (such as load rate, energy conversion rate, and operational stability indicators of each device, the overall energy supply and demand balance of the system, and energy consumption distribution characteristics) are extracted and integrated to form a comprehensive operating condition report reflecting the current energy system operation status.

[0041] Specifically, the process of obtaining risk warning results includes: screening safety feature data based on the multi-dimensional feature set to obtain risk assessment factors; classifying risk levels by combining safety standards and equipment failure thresholds; monitoring the changing trends of the risk assessment factors in real time based on the operating status fed back by the digital twin, and comparing the results of the risk level classification; identifying potential risks based on the comparison results, and generating risk warning results.

[0042] In this embodiment, safety feature data directly related to safety management are screened from a multi-dimensional feature set to determine risk assessment factors (such as equipment temperature-related factors, insulation performance factors, overload duration factors, and thermal runaway-related feature factors). Combining industry safety standards and fault thresholds provided by equipment manufacturers (such as safety operation parameter thresholds for energy storage systems), risks are classified into three levels: minor, medium, and severe, with the threshold range for risk assessment factors corresponding to each level clearly defined. Based on the operating status feedback from the digital twin, the changing trends of each risk assessment factor (such as whether the temperature continues to rise or whether the insulation performance continues to decline) are monitored in real time and dynamically compared with preset risk level thresholds. When a risk assessment factor is detected to be close to or exceed the corresponding level threshold, or when the changing trend indicates a potential safety hazard, the potential risk type (such as thermal runaway risk of energy storage systems, equipment overload risk, insulation failure risk, etc.) is promptly identified, generating a risk warning result that includes the risk type, risk level, scope of impact (involved equipment and areas), potential hazards, and response suggestions (such as adjusting equipment operating parameters or suspending the operation of some equipment).

[0043] Specifically, the process of obtaining the corresponding energy optimization benchmark includes: setting corresponding safety constraint boundaries based on the risk warning results; extracting the level values ​​of the corresponding dimensions based on the operation status assessment report, obtaining the economic optimization benchmark, and integrating them to obtain the energy optimization benchmark report.

[0044] In this embodiment, risk warning results are used as the core basis to set safety constraint boundaries for the operation of the energy system—clearly defining the prohibited conditions for equipment operation (such as prohibiting equipment from operating beyond safety thresholds) and restrictive ranges (such as the safe range of equipment load rate, the reasonable range of temperature control, etc.). Based on the simulated operation data of the digital twin, operating condition indicators are extracted, and the comprehensive operating status value is calculated by weighted summation.

[0045] Formula description: Let the set of operating condition indicators output by the digital twin be M = {m1, m2, ..., m l (l represents the number of operating condition indicators): Define the indicator weights w k ( (Set based on the importance of control objectives); comprehensive operating status value. , where m k,min m k,max These represent the minimum and maximum reasonable values ​​of indicator k, used for indicator normalization. Current level values ​​for each dimension are extracted from the operational status assessment report, including safe operation level (e.g., frequency of risk occurrence, equipment safety parameter compliance rate), economic operation level (e.g., current operation and maintenance costs, total electricity costs), energy efficiency level (e.g., energy utilization efficiency, equipment operating efficiency), and carbon emission level (e.g., current carbon emissions, carbon emission intensity per unit of energy consumption). Combining benchmark levels with those of similar industrial parks in the industry and the industrial park's pre-set energy management goals (e.g., energy saving goals, emission reduction goals, cost control goals), optimization benchmarks for each dimension are determined: economic optimization benchmarks (e.g., operation and maintenance cost reduction targets, electricity cost optimization range), energy efficiency optimization benchmarks (e.g., energy utilization efficiency improvement targets, equipment operating efficiency compliance standards), and carbon emission optimization benchmarks (e.g., carbon emission reduction ratio, carbon emission intensity control targets per unit of energy consumption), integrating these to form a multi-dimensional, quantifiable energy optimization benchmark.

[0046] Specifically, the process of obtaining the energy scheduling strategy includes: constructing a multi-agent reinforcement learning framework based on the multi-dimensional feature set; using the energy optimization benchmark report to preset the reward function of the agents; and coordinating the agents to play games and cooperate through a central coordinator to generate an energy scheduling strategy.

[0047] In this embodiment, a multi-agent reinforcement learning framework is constructed based on a multi-dimensional feature set, enabling each agent to operate independently and work collaboratively. A single management objective is assigned to each agent: the safety and reliability agent focuses on ensuring the system meets safety constraints and avoids risks; the economic operation agent focuses on reducing operation and maintenance costs and electricity costs, aligning with economic optimization benchmarks; the energy efficiency improvement agent is dedicated to improving energy utilization efficiency and equipment operating efficiency; and the carbon emission optimization agent takes reducing carbon emissions and carbon emission intensity as its core objective. Guided by the energy optimization benchmark system, a personalized reward function is set for each agent: if the agent's decision-making behavior enables the corresponding dimension indicator to reach or exceed the optimization benchmark, a positive reward is given; if the safety constraints are exceeded or the optimization benchmark requirements are not met, a negative penalty is given; for collaborative behaviors between different agents (such as the collaborative achievement of economic operation and carbon emission optimization objectives), additional collaborative rewards are given. A central coordinator coordinates the various agents in a game-like and collaborative process. When the goals of different agents conflict (e.g., pursuing economic gains may lead to increased carbon emissions), the central coordinator balances the overall management priorities of the industrial park (which can be dynamically adjusted according to actual needs) to obtain a globally optimal energy scheduling strategy with multiple objectives. The process is as follows: Let the policy spaces of each agent be ΠS, ΠE, Π E f,ΠC The central coordinator's objective is to maximize the total global reward R. total =∑ k∈{S,E,Ef,C} λ k R k (λ) k (Assign target weights to each agent, based on global control priority) Each agent generates a candidate policy π based on its own reward function and policy space. S ∈Π S ,π E ∈Π E, π E f∈Π E f,π C ∈Π C The central coordinator calculates the total global reward R for each candidate strategy combination. total ( , , , Find the strategy combination corresponding to the equilibrium point in the game. , , , ), satisfying R total ( , , , ,)=max , , , R total ( , , , The scheduling scheme corresponding to this strategy combination is the globally optimal energy scheduling strategy.

[0048] At the same time, ensure that the strategy is reconfigurable and can be adapted to future scenarios such as equipment expansion and load changes in the industrial park.

[0049] Specifically, the executable control instructions are as follows: the strategy content is split based on the energy scheduling strategy and the energy equipment type to obtain the operation requirements of the corresponding energy equipment; based on the operation requirements of the corresponding energy equipment and the energy equipment control protocol, the strategy content is transformed to obtain an instruction format that the energy equipment can recognize; based on the instruction format, the parameters for instruction execution are supplemented and sorted using the preset instruction execution logic to obtain a set of control instructions for the energy equipment.

[0050] Specifically, the process of obtaining the instruction execution result includes: transmitting the instructions through a communication module based on the control instruction set; using the feedback mechanism of the energy equipment to receive the operation data of the energy equipment executing the instructions in real time; comparing the target value required by the instruction, analyzing the process data of the instruction execution, and obtaining the instruction execution result.

[0051] In this embodiment, based on the energy dispatch strategy and the types of energy equipment in the industrial park (energy storage systems, new energy power generation devices, and various types of electricity loads), the macro-dispatch strategy is broken down into specific operational requirements for the corresponding equipment. This includes clarifying the direction of charging and discharging power adjustment, charging and discharging time planning, and liquid cooling control mode switching requirements for the energy storage system; the output adjustment range and operating mode switching instructions for the power generation device; and the staggered operation arrangements and energy consumption control targets for the electricity load. Based on the equipment operation requirements and the control protocols of each energy device, the operation requirements are converted into instruction formats that the equipment can recognize (such as instruction codes conforming to the energy storage system control interface specification, and control signal formats that the power generation device can receive). Key parameters for instruction execution are supplemented, including the execution start time (such as a specific time period or start when specific conditions are met), operation duration, operating parameter adjustment gradient (such as the step size of power adjustment), and instruction priority (safety-related instructions have the highest priority, followed by instructions related to core production loads). These instructions are then sorted according to a preset instruction execution logic (safety-assurance instructions are executed first, followed by optimization instructions; core equipment and critical loads are prioritized before general equipment and ordinary loads), forming a complete set of control instructions. The control command set is precisely transmitted to the corresponding energy equipment through the communication module. Utilizing the equipment's feedback mechanism, operational data (such as actual equipment operating parameters and status change information) is received in real time during command execution. The actual equipment operating data is compared with the target values ​​required by the commands to analyze the completion rate of command execution (whether it was executed as required) and the achievement of the desired effect (whether the expected optimization goal was achieved), generating a detailed command execution result report.

[0052] A multi-zone monitoring interface is constructed, adhering to a modular design concept. Interface zones correspond one-to-one with system functional units, including data acquisition, data processing, twin simulation, risk warning, optimization strategy, and command execution zones. Simultaneously, multi-dimensional indicator dashboards for safety, economy, energy efficiency, and carbon emissions are set up, displaying the current status, historical trends, and optimization effects of each indicator in real time (e.g., the number of risk warnings in safety indicators, cost changes in economic indicators, efficiency improvement rates in energy efficiency indicators, and emission reductions in carbon emissions indicators). The interface allows maintenance personnel to filter and query data by equipment type (e.g., viewing energy storage systems individually), management dimension (e.g., focusing only on carbon emissions), and time period (e.g., real-time data, daily data, monthly data). They can also view the source information of each data point, the detailed process of command execution, and result feedback. Through the intelligent maintenance dashboard, maintenance personnel can intuitively grasp key information throughout the entire energy management process, enabling remote monitoring, rapid response to risk warnings, and timely adjustment of management strategies, significantly improving maintenance efficiency and decision-making accuracy while reducing manual intervention costs.

[0053] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A multi-dimensional collaborative intelligent energy management and control system, characterized in that, include: Data acquisition unit, data processing unit, digital twin and risk assessment unit, multi-objective collaborative optimization and visualization unit; The data acquisition unit is used to acquire raw data from multiple dimensions; The data processing unit processes the multi-dimensional raw data according to preset classification rules to obtain a multi-source raw dataset; it removes outliers and fills in missing values ​​in the multi-source raw dataset, and extracts multi-dimensional features to obtain a standardized dataset and a multi-dimensional feature set. The digital twin and risk assessment unit reconstructs the physical topology of the energy system and the parameters of energy equipment based on the standardized dataset, constructs a digital twin that is synchronized with the physical system in real time, and calls the multi-dimensional feature set to drive the digital twin to obtain the operating status of the energy conditions. The risk factors are analyzed by combining the aforementioned multi-dimensional feature set to obtain risk warning results; Integrate the operational status with the risk warning results to generate an operational status assessment report; The multi-objective collaborative optimization and visualization unit sets safety constraint boundaries based on the risk warning results, and obtains energy optimization benchmarks in the corresponding dimensions by combining the operation status assessment report; it constructs multi-agent reinforcement learning using the multi-dimensional feature set, and obtains energy scheduling strategies through the collaborative game of the agents; The energy dispatch strategy is broken down into executable control instructions, the control instructions are verified and driven to execute, and the execution results are obtained; a multi-zone monitoring interface is constructed to visualize the entire process data of energy management and control.

2. The system according to claim 1, characterized in that, The specific process of acquiring multi-dimensional raw data includes: obtaining the target data type based on the energy system equipment deployment list, and establishing communication connections with the energy storage system, power grid platform, and carbon emission database; based on the preset sampling frequency of the communication connection channel, collecting multi-dimensional raw data in real time and synchronously recording the data collection timestamp and energy equipment identifier to obtain a multi-dimensional raw data set with traceability information.

3. The system according to claim 1, characterized in that, The specific process of obtaining the multi-source raw dataset includes: classifying the multi-dimensional raw data with traceability information, removing redundant data collected repeatedly, and integrating the classification results to obtain a structured multi-source raw dataset.

4. The system according to claim 1, characterized in that, The outlier removal process involves: establishing an evaluation criterion for outlier data based on the structured multi-source original dataset, detecting the corresponding dataset, obtaining the detection results, and removing the outlier data.

5. The system according to claim 1, characterized in that, The specific process of the missing value completion operation includes: based on the dataset after outlier removal, statistically analyzing the missing locations and proportions of statistical data fields; based on the missing proportions and data types, selecting an appropriate completion algorithm, and calculating the completion value using the statistical characteristics of the missing data; and using the calculated completion value to fill in the corresponding dataset to obtain a standard dataset.

6. The system according to claim 1, characterized in that, The specific process of constructing a digital twin that is synchronized with the physical system in real time includes: extracting the physical topology and equipment parameters of the energy system based on the standard state dataset; using digital twin modeling technology to build a virtual model framework consistent with the physical system structure, establishing a data mapping relationship, and associating the real-time data in the standard state dataset with the virtual model; and synchronizing the operating data of the physical system in real time based on the data mapping relationship to construct a digital twin that is synchronized with the physical system in real time.

7. The system according to claim 1, characterized in that, The specific process of obtaining the operating status of the working condition includes: extracting a multi-dimensional feature set based on the standard state dataset and filtering the feature data of the corresponding working condition; inputting the feature data of the working condition into the operation simulation module of the digital twin based on the filtering results to drive the virtual model to reproduce the operation process of the physical system; and extracting the working condition indicators and integrating them to obtain the operating status report based on the simulation operation results of the digital twin.

8. The system according to claim 1, characterized in that, The specific process for obtaining risk warning results includes: screening safety feature data based on the multi-dimensional feature set to obtain risk assessment factors; classifying risk levels by combining safety standards and equipment failure thresholds; monitoring the changing trends of the risk assessment factors in real time based on the operating status fed back by the digital twin, and comparing the results of the risk level classification; identifying potential risks based on the comparison results, and generating risk warning results.

9. The system according to claim 1, characterized in that, The specific process of obtaining the corresponding energy optimization benchmark includes: setting corresponding safety constraint boundaries based on the risk warning results; extracting the level values ​​of the corresponding dimensions based on the operation status assessment report, obtaining the economic optimization benchmark, and integrating them to obtain the energy optimization benchmark report.

10. The system according to claim 1, characterized in that, The specific process of obtaining the energy scheduling strategy includes: constructing a multi-agent reinforcement learning framework based on the multi-dimensional feature set, using the energy optimization benchmark report to preset the reward function of the agents, and coordinating the agents to play games and cooperate through a central coordinator to generate an energy scheduling strategy.

11. The system according to claim 1, characterized in that, The executable control instructions are as follows: based on the energy scheduling strategy and energy equipment type, the strategy content is split to obtain the operation requirements of the corresponding energy equipment; based on the operation requirements of the corresponding energy equipment and the energy equipment control protocol, the strategy content is transformed to obtain an instruction format that the energy equipment can recognize; based on the instruction format, the parameters for instruction execution are supplemented and sorted using the preset instruction execution logic to obtain a set of control instructions for the energy equipment.

12. The system according to claim 1, characterized in that, The specific process of obtaining the instruction execution result includes: transmitting the instructions through a communication module based on the control instruction set; using the feedback mechanism of the energy equipment to receive the operation data of the energy equipment executing the instructions in real time; comparing the target value required by the instruction, analyzing the process data of the instruction execution, and obtaining the instruction execution result.

Images