Intelligent multi-source energy collaborative optimization system

Abstract

The invention discloses an intelligent multi-source energy collaborative optimization system. According to the method, the efficiency and the economical efficiency of energy management are remarkably improved through the dynamic decision-making capability of the multi-objective optimization engine. The engine is based on real-time energy data and environment variables, deep prediction and reinforcement learning technologies are combined, accurate distribution and dynamic balance of multi-source energy are achieved, for example, when the electricity price fluctuates or the equipment load suddenly changes, the system can rapidly adjust the proportion of electric energy, heat energy and waste heat recovery, energy waste is reduced, and the operation cost is optimized. And meanwhile, the energy recovery module converts traditional waste energy into available resources, the dependence on external energy is further reduced in cooperation with global scheduling of the cloud service platform, and the comprehensive energy efficiency of the enterprise can be improved by more than 20%. The system supports a multi-scene adaptive strategy, takes into account short-term cost control and long-term equipment health management on the premise of ensuring production continuity, and provides an energy solution with high cost performance for industrial and commercial users.

Description

Technical Field

[0001] This invention belongs to the field of energy management technology, specifically an intelligent multi-source energy collaborative optimization system. Background Technology

[0002] Energy management systems are increasingly widely used in industrial and commercial sectors. However, with the diversification of energy structures and increasing environmental protection requirements, traditional energy management systems are gradually revealing problems such as low efficiency and insufficient flexibility. In recent years, the rapid development of IoT technology and artificial intelligence algorithms has provided new technical means for energy management. However, how to effectively integrate these technologies to achieve dynamic optimization of multi-source energy remains a technical challenge. In addition, enterprises' needs to reduce operating costs and achieve green and sustainable development are also driving innovation in energy management technologies.

[0003] However, existing technologies have the following drawbacks: - Single energy management system: It only manages a single energy source such as electricity or heat, and lacks the ability to optimize multiple sources, resulting in low energy utilization.

[0004] - Static energy allocation strategy: Energy allocation is based on fixed rules or historical data, which cannot adapt to dynamically changing energy demands and environmental conditions.

[0005] - Simple waste heat recovery systems: These systems recover only a portion of the waste heat through a basic heat exchanger, without optimization in conjunction with other energy systems, resulting in limited recovery efficiency.

[0006] Traditional carbon emission monitoring relies on manual recording or single sensor monitoring, resulting in delayed data updates and insufficient accuracy. Summary of the Invention

[0007] The purpose of this invention is to provide an intelligent multi-source energy collaborative optimization system in order to solve the problems mentioned above.

[0008] The technical solution adopted in this invention is as follows: an intelligent multi-source energy collaborative optimization system, the system including a data acquisition module, a multi-objective optimization engine module, an energy recovery module, a carbon emission monitoring module, a carbon trading interface module, and a cloud service platform module, wherein the multi-objective optimization engine module is internally equipped with a feature engineering module, a deep prediction model module, a reinforcement learning strategy generation module, and a blockchain verification and execution module; The real-time multi-source energy data of the data acquisition module is transmitted to the input of the multi-objective optimization engine module through the output end, driving the latter to generate a dynamic optimization strategy. The output of the multi-objective optimization engine module simultaneously sends energy recovery instructions to the input of the energy recovery module and pushes optimization parameters to the input of the carbon emission monitoring module to support carbon emission monitoring. The output of the carbon emission monitoring module is further connected to the input of the carbon trading interface module, so as to input real-time carbon emission data into the carbon trading interface module. The outputs of the data acquisition module, multi-objective optimization engine module, energy recovery module, carbon emission monitoring module, and carbon trading interface module all aggregate data to the input of the cloud service platform module, where global integration and analysis are completed by the cloud. The output of the cloud service platform module is connected in reverse to the input of the multi-objective optimization engine module via a feedback link.

[0009] In a preferred embodiment, the data acquisition module internally incorporates a low-power, high-precision intelligent sensor network and a self-organizing network communication unit. Sensor nodes are deployed at various energy-consuming terminals, such as those using electricity, heat, and natural gas, and employ TI CC2650 series chips to achieve millisecond-level data acquisition, covering multi-dimensional parameters such as voltage, current, temperature, and flow rate. The master node aggregates data from various sensors via Zigbee or LoRa wireless self-organizing network technology, and its built-in edge computing unit performs preliminary filtering and noise reduction on the raw data, reducing the cloud transmission load. The communication protocol supports dynamic routing adjustments to ensure signal stability in complex industrial scenarios, and the data acquisition frequency can be configured to be adjustable from 1 second to 10 minutes as needed.

[0010] In a preferred embodiment, the feature engineering module integrates real-time data and environmental parameters from multiple energy sources, including electricity, heat, and natural gas. Through data cleaning, noise reduction, and normalization, it eliminates noise interference and unifies dimensions, ensuring data consistency and usability. The module utilizes sliding window sampling technology to extract time-series features, capturing the volatility of energy demand and the periodicity of waste heat emissions. It also integrates historical energy consumption patterns with external variables such as carbon price fluctuations, real-time electricity prices, and weather forecast data to construct a dynamically labeled dataset to support model training and inference. The module uses a time-series database to store high-concurrency real-time data. Based on a distributed computing framework, it performs frequency domain feature extraction and dimensionality reduction on the original data and introduces data augmentation strategies to synthesize extreme scenario data, improving the model's adaptability to complex operating conditions. Furthermore, the module uses multimodal fusion technology to uniformly encode structured sensor data and unstructured equipment logs into multidimensional feature vectors. Combined with a dynamic weight fusion mechanism, it balances the influence of real-time data, historical patterns, and external market variables, providing high-quality, highly generalizable input features for subsequent deep prediction models, significantly improving the accuracy and robustness of multi-energy collaborative optimization.

[0011] In a preferred embodiment, the deep prediction model module employs a Transformer-based deep neural network to accurately predict future supply and demand trends for various energy sources, including electricity, heat, and natural gas. This module integrates real-time collected energy consumption data, equipment operating status indicators, historical energy consumption patterns, and external environmental variables such as temperature, light intensity, and carbon price fluctuations. It dynamically captures the correlation of energy demand at different time scales through a hierarchical self-attention mechanism. Input features include energy consumption rates, equipment efficiency curves, energy supply stability, and market dynamic parameters. The model uses time series analysis techniques to generate short-term, medium-term, and long-term energy supply and demand balance predictions. An adaptive weight allocation mechanism is introduced during the prediction process, adjusting feature weights based on the importance of energy types and real-time environmental impact factors to ensure the model's robustness in complex scenarios. Prediction errors are further optimized using adversarial training techniques, enabling the model to effectively cope with extreme weather or sudden equipment failures. The prediction results are transmitted in real-time to a reinforcement learning policy generator, providing forward-looking input for dynamic energy allocation and supporting the low-carbon and efficient operation of the system. The formula for dynamic feature weighted prediction is: Where: w i The baseline weight of the i-th energy type is calculated from historical consumption efficiency and equipment performance indicators, reflecting the inherent importance of energy.

[0012] λ: Time decay coefficient, which controls the decay rate of the impact of historical data on the current prediction, and is related to the update frequency of real-time data.

[0013] Δt: The time interval between the current time and historical data, used to quantify the timeliness of the data.

[0014] c i Environmental impact factors are dynamically adjusted based on real-time carbon emission intensity and carbon price to enhance the forecasting priority of low-carbon energy.

[0015] f i : The input feature vector of the i-th energy type, which includes real-time consumption rate, supply stability and external correlation variables.

[0016] In a preferred embodiment, the reinforcement learning strategy generation module, based on the output of a deep prediction model and real-time energy data, employs a near-end policy optimization algorithm to construct a dynamic decision-making framework. This framework aims to minimize operating costs and maximize energy efficiency, driving multi-source energy collaborative optimization. The module defines its state space as the energy supply-demand difference, equipment health, carbon emission exceedance risk level, and real-time carbon price fluctuations. Its action space encompasses multi-dimensional control variables such as energy allocation ratio adjustments, waste heat recovery device start-up and shutdown commands, and energy storage device charging and discharging strategies. Through Markov decision process modeling, the module dynamically generates energy allocation strategies by combining immediate reward functions and long-term value functions. The reward function integrates operating cost savings, energy utilization improvement, and carbon emission reduction, and dynamically adjusts the weights of each objective based on external carbon emission quotas and market carbon prices, achieving adaptive switching of strategy priorities. After strategy execution, actual energy consumption and carbon emission data are fed back to the module, forming a closed-loop learning mechanism that continuously optimizes the strategy generation logic, ensuring the system's stability and low-carbon orientation in complex scenarios.

[0017] The formula for calculating the dynamic multi-objective reward function is as follows: R = α·C saved +β·U improved +γ·E reduced ; in: α: Cost-saving weighting factor, which is positively correlated with real-time electricity prices and energy procurement cost fluctuations, dynamically amplifying cost-saving benefits in cost-sensitive scenarios.

[0018] C saved It represents the amount of operating cost savings per unit of time, calculated from the difference between the actual cost and the benchmark cost.

[0019] β represents the energy efficiency weighting factor, which is related to the health of the equipment and the historical energy efficiency baseline. When the equipment condition deteriorates, this weight is automatically increased to maintain system efficiency.

[0020] U improved It indicates the improvement in energy utilization rate, which is quantified by the ratio of real-time energy conversion rate to the ideal value.

[0021] γ represents the carbon emission reduction weighting factor, which increases linearly with rising carbon prices and decreasing carbon emission allowances, thus forcing policies to tilt towards low-carbon directions.

[0022] E reduced This represents the reduction in carbon emissions, calculated based on the difference between real-time emission data provided by the monitoring module and the target threshold.

[0023] In a preferred embodiment, the blockchain verification and execution module builds a private blockchain network based on the Hyperledger Fabric framework, deploying verification nodes on edge computing devices and cloud servers to achieve decentralized consensus across physical locations. The smart contract pre-configures a multi-source energy coordination rule base, covering core strategies such as carbon emission thresholds, energy allocation ratios, and equipment security parameters. When the optimization engine generates instructions to be executed, the verification engine simultaneously triggers contract logic verification, using asymmetric encryption algorithms to verify data integrity and ensure that the strategies comply with industry standards and internal enterprise standards. The execution terminal interface uses the OPC UA protocol to interconnect with industrial equipment, sending verified instructions in real time to execution units such as the energy recovery subsystem and energy storage devices, and collecting equipment feedback data to write into the blockchain ledger, forming an immutable audit and traceability chain.

[0024] In a preferred embodiment, the energy recovery module internally houses a high-temperature resistant stainless steel heat exchanger and a micro-thermoelectric energy conversion device. The heat exchanger, installed at the industrial waste heat discharge outlet, maximizes waste heat capture efficiency through a multi-layer finned structure, achieving a thermal conductivity of 400 W / (m·K). The energy conversion device integrates a thermoelectric generator, utilizing the Seebeck effect to directly convert the waste heat temperature difference into electrical energy, achieving a conversion efficiency of 12%–15%. The output electrical energy is then fed into the enterprise's microgrid or energy storage unit via a voltage regulator circuit. The system incorporates a waste heat priority algorithm, dynamically adjusting the recovery ratio based on real-time energy demand to ensure a coordinated balance between waste heat reuse and primary energy supply.

[0025] In a preferred embodiment, the carbon emission monitoring module employs a multispectral infrared sensor array and a scene-adaptive analysis algorithm. Sensor nodes are deployed at key locations such as flues and exhaust outlets, supporting ppm-level concentration monitoring of six greenhouse gases, including CO2 and CH4, with a sampling accuracy of ±1.5%. The scene analysis algorithm constructs a dynamic carbon emission model based on equipment operating status, ambient temperature and humidity, and production rhythm data, enabling early warning and source tracing analysis of exceedance risks. Monitoring data is uploaded to edge computing nodes at 5-second intervals, and a carbon emission heatmap is generated in real time through a lightweight convolutional network, providing spatial decision-making support for enterprises' low-carbon control.

[0026] In a preferred embodiment, the carbon trading interface module internally integrates a carbon market API adapter and a smart contract engine. The module has a built-in multi-regional carbon trading rule library and obtains real-time carbon allowance prices, policy changes, and market supply and demand data via a RESTful API. The smart contract engine automatically generates allowance purchase / sale recommendations based on the carbon emission margin and compliance period provided by the monitoring module and performs compliance verification. Trading instructions are encrypted with AES-256 and pushed to an external carbon exchange, simultaneously updating the enterprise's ERP system and supporting real-time visualization analysis of carbon asset gains and losses.

[0027] In a preferred embodiment, the cloud service platform module is built upon a microservice architecture and digital twin technology to construct a comprehensive management hub. The platform deploys a Kubernetes cluster, supporting the elastic scaling of microservices such as energy data storage, multi-objective optimization model training, and equipment health diagnostics. The digital twin engine maps physical devices to virtual entities, driving simulation operations with real-time data to predict equipment failures and optimize maintenance strategies. The platform provides a visual dashboard function, integrating key indicators such as energy efficiency, carbon emission trends, and transaction records, supporting multi-terminal access on PCs and mobile devices, and offering an SDK for third-party developers to customize industry-specific analysis tools.

[0028] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. This invention significantly improves the efficiency and economy of energy management through the dynamic decision-making capabilities of a multi-objective optimization engine. Based on real-time energy data and environmental variables, and combined with deep prediction and reinforcement learning techniques, the engine achieves precise allocation and dynamic balance of multi-source energy. For example, when electricity prices fluctuate or equipment loads change abruptly, the system can quickly adjust the proportions of electrical energy, thermal energy, and waste heat recovery, reducing energy waste and optimizing operating costs. Simultaneously, the energy recovery module converts traditional waste energy into usable resources. Combined with the global scheduling of the cloud service platform, this further reduces dependence on external energy sources, resulting in an overall energy efficiency improvement of over 20% for enterprises. The system supports multi-scenario adaptive strategies, ensuring production continuity while balancing short-term cost control and long-term equipment health management, providing cost-effective energy solutions for industrial and commercial users.

[0029] 2. This invention integrates carbon emission monitoring and carbon trading mechanisms to construct a closed-loop carbon reduction path of "monitoring-optimization-trading". The multi-objective optimization engine uses carbon emission intensity as the core constraint, dynamically limiting the proportion of high-polluting energy use, and ensures the compliance and transparency of strategy execution through blockchain technology. The carbon trading interface connects to market data in real time, automatically generating quota trading suggestions to help enterprises transform emission reduction results into economic benefits and alleviate carbon tax pressure. In addition, the system's predictive model and reinforcement learning algorithm can avoid the risk of exceeding carbon emission limits in advance, and combined with waste heat recovery and priority scheduling of clean energy, it promotes the green and low-carbon transformation of the industrial sector. Attached Figure Description

[0030] Figure 1 This is an overall system block diagram of the present invention; Figure 2 This is a system block diagram of the multi-objective optimization engine module in this invention.

[0031] The diagram is labeled as follows: 1 - Data acquisition module, 2 - Multi-objective optimization engine module, 3 - Energy recovery module, 4 - Carbon emission monitoring module, 5 - Carbon trading interface module, 6 - Cloud service platform module, 7 - Feature engineering module, 8 - Deep prediction model module, 9 - Reinforcement learning strategy generation module, 10 - Blockchain verification and execution module. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0033] Example: Reference Figure 1-2 , A smart multi-source energy collaborative optimization system includes a data acquisition module 1, a multi-objective optimization engine module 2, an energy recovery module 3, a carbon emission monitoring module 4, a carbon trading interface module 5, and a cloud service platform module 6. The multi-objective optimization engine module 2 is internally equipped with a feature engineering module 7, a deep prediction model module 8, a reinforcement learning strategy generation module 9, and a blockchain verification and execution module 10. The real-time multi-source energy data from the data acquisition module 1 is transmitted to the input of the multi-objective optimization engine module 2 through the output end, driving the latter to generate a dynamic optimization strategy. The output of the multi-objective optimization engine module 2 simultaneously sends energy recovery instructions to the input of the energy recovery module 3 and pushes optimization parameters to the input of the carbon emission monitoring module 4 to support carbon emission monitoring. The output of the carbon emission monitoring module 4 is further connected to the input of the carbon trading interface module 5, so as to input real-time carbon emission data into the carbon trading interface module. The outputs of data acquisition module 1, multi-objective optimization engine module 2, energy recovery module 3, carbon emission monitoring module 4, and carbon trading interface module 5 all aggregate data to the input of cloud service platform module 6, where global integration and analysis are completed by the cloud. The output of cloud service platform module 6 is connected in reverse to the input of multi-objective optimization engine module 2 via a feedback link.

[0034] The data acquisition module 1 internally incorporates a low-power, high-precision intelligent sensor network and a self-organizing network communication unit. Sensor nodes are deployed at various energy-consuming terminals, including those using electricity, heat, and natural gas. They utilize TI CC2650 series chips to achieve millisecond-level data acquisition, covering multi-dimensional parameters such as voltage, current, temperature, and flow rate. The master node aggregates data from various sensors via Zigbee or LoRa wireless self-organizing network technology. An integrated edge computing unit performs preliminary filtering and noise reduction on the raw data, reducing the cloud transmission load. The communication protocol supports dynamic routing adjustments to ensure signal stability in complex industrial scenarios. The data acquisition frequency can be configured to be adjustable from 1 second to 10 minutes as needed.

[0035] Feature Engineering Module 7 integrates real-time data and environmental parameters from multiple energy sources, including electricity, heat, and natural gas. Through data cleaning, noise reduction, and normalization, it eliminates noise interference and unifies dimensions, ensuring data consistency and usability. The module utilizes sliding window sampling technology to extract time-series features, capturing the volatility of energy demand and the periodicity of waste heat emissions. It also integrates historical energy consumption patterns with external variables such as carbon price fluctuations, real-time electricity prices, and weather forecasts to construct a dynamically labeled dataset to support model training and inference. The module uses a time-series database to store high-concurrency real-time data. Based on a distributed computing framework, it performs frequency domain feature extraction and dimensionality reduction on the raw data and introduces data augmentation strategies to synthesize extreme scenario data, improving the model's adaptability to complex operating conditions. Furthermore, the module uses multimodal fusion technology to uniformly encode structured sensor data and unstructured equipment logs into multidimensional feature vectors. Combined with a dynamic weight fusion mechanism, it balances the influence of real-time data, historical patterns, and external market variables, providing high-quality, highly generalizable input features for subsequent deep prediction models, significantly improving the accuracy and robustness of multi-energy collaborative optimization.

[0036] The deep prediction model module 8 employs a Transformer-based deep neural network to accurately predict future supply and demand trends for various energy sources, including electricity, heat, and natural gas. This module integrates real-time collected energy consumption data, equipment operating status indicators, historical energy consumption patterns, and external environmental variables such as temperature, light intensity, and carbon price fluctuations. It dynamically captures the correlation between energy demand at different time scales through a hierarchical self-attention mechanism. Input features include energy consumption rates, equipment efficiency curves, energy supply stability, and market dynamic parameters. The model utilizes time series analysis techniques to generate short-term, medium-term, and long-term energy supply and demand balance predictions. An adaptive weight allocation mechanism is introduced during the prediction process, adjusting feature weights based on the importance of energy types and real-time environmental impact factors to ensure the model's robustness in complex scenarios. Prediction errors are further optimized through adversarial training techniques, enabling the model to effectively handle extreme weather or sudden equipment failures. The prediction results are transmitted in real-time to a reinforcement learning policy generator, providing forward-looking input for dynamic energy allocation and supporting the system's low-carbon and efficient operation. The dynamic feature weighted prediction formula is as follows: Where: w i The baseline weight of the i-th energy type is calculated from historical consumption efficiency and equipment performance indicators, reflecting the inherent importance of energy.

[0037] λ: Time decay coefficient, which controls the decay rate of the impact of historical data on the current prediction, and is related to the update frequency of real-time data.

[0038] Δt: The time interval between the current time and historical data, used to quantify the timeliness of the data.

[0039] c i Environmental impact factors are dynamically adjusted based on real-time carbon emission intensity and carbon price to enhance the forecasting priority of low-carbon energy.

[0040] f i : The input feature vector of the i-th energy type, which includes real-time consumption rate, supply stability and external correlation variables.

[0041] The reinforcement learning strategy generation module 9, based on the output of a deep prediction model and real-time energy data, employs a near-end policy optimization algorithm to construct a dynamic decision-making framework. With the dual objectives of minimizing operating costs and maximizing energy efficiency, it drives multi-source energy collaborative optimization. This module defines its state space as the energy supply-demand difference, equipment health, carbon emission exceedance risk level, and real-time carbon price fluctuations. Its action space encompasses multi-dimensional control variables such as energy allocation ratio adjustments, waste heat recovery device start-up and shutdown commands, and energy storage device charging and discharging strategies. Through Markov decision process modeling, the module dynamically generates energy allocation strategies by combining immediate reward functions and long-term value functions. The reward function integrates operating cost savings, energy utilization improvement, and carbon emission reduction, and dynamically adjusts the weights of each objective based on external carbon emission quotas and market carbon prices, achieving adaptive switching of strategy priorities. After strategy execution, actual energy consumption and carbon emission data are fed back to the module, forming a closed-loop learning mechanism that continuously optimizes the strategy generation logic, ensuring the system's stability and low-carbon orientation in complex scenarios.

[0042] The formula for calculating the dynamic multi-objective reward function is as follows: R = α·C saved +β·U improved +γ·E reduced ; in: α: Cost-saving weighting factor, which is positively correlated with real-time electricity prices and energy procurement cost fluctuations, dynamically amplifying cost-saving benefits in cost-sensitive scenarios.

[0043] C saved It represents the amount of operating cost savings per unit of time, calculated from the difference between the actual cost and the benchmark cost.

[0044] β represents the energy efficiency weighting factor, which is related to the health of the equipment and the historical energy efficiency baseline. When the equipment condition deteriorates, this weight is automatically increased to maintain system efficiency.

[0045] U improved It indicates the improvement in energy utilization rate, which is quantified by the ratio of real-time energy conversion rate to the ideal value.

[0046] γ represents the carbon emission reduction weighting factor, which increases linearly with rising carbon prices and decreasing carbon emission allowances, thus forcing policies to tilt towards low-carbon directions.

[0047] E reduced This represents the reduction in carbon emissions, calculated based on the difference between real-time emission data provided by the monitoring module and the target threshold.

[0048] The blockchain verification and execution module 10 is built on the Hyperledger Fabric framework to construct a private blockchain network, deploying verification nodes on edge computing devices and cloud servers to achieve decentralized consensus across physical locations. The smart contract pre-configures a multi-source energy coordination rule base, covering core strategies such as carbon emission thresholds, energy allocation ratios, and equipment security parameters. When the optimization engine generates instructions to be executed, the verification engine simultaneously triggers contract logic verification, using asymmetric encryption algorithms to verify data integrity and ensure that the strategies comply with industry standards and internal enterprise standards. The execution terminal interface uses the OPC UA protocol to interconnect with industrial equipment, sending verified instructions in real time to execution units such as the energy recovery subsystem and energy storage devices, and collecting equipment feedback data to write into the blockchain ledger, forming an immutable audit and traceability chain.

[0049] The energy recovery module 3 internally houses a high-temperature resistant stainless steel heat exchanger and a micro thermoelectric energy conversion device. The heat exchanger, installed at the industrial waste heat discharge outlet, maximizes waste heat capture efficiency through a multi-layer finned structure, achieving a thermal conductivity of 400 W / (m·K). The energy conversion device integrates a thermoelectric generator (TEG), utilizing the Seebeck effect to directly convert waste heat temperature differences into electrical energy, achieving a conversion efficiency of 12%–15%. The output electricity is then fed into the enterprise's microgrid or energy storage unit via a voltage regulator circuit. The system incorporates a waste heat priority algorithm, dynamically adjusting the recovery ratio based on real-time energy demand to ensure a coordinated balance between waste heat reuse and primary energy supply.

[0050] The carbon emission monitoring module 4 employs a multispectral infrared sensor array and a scene-adaptive analysis algorithm. Sensor nodes are deployed at key locations such as flues and exhaust outlets, supporting ppm-level concentration monitoring of six greenhouse gases, including CO2 and CH4, with a sampling accuracy of ±1.5%. The scene analysis algorithm constructs a dynamic carbon emission model based on equipment operating status, ambient temperature and humidity, and production rhythm data, enabling early warning and source tracing analysis of exceedance risks. Monitoring data is uploaded to edge computing nodes at 5-second intervals, generating real-time carbon emission heatmaps through lightweight convolutional networks, providing spatial decision-making support for enterprises' low-carbon control.

[0051] The carbon trading interface module 5 is internally equipped with an integrated carbon market API adapter and smart contract engine. The module has a built-in multi-regional carbon trading rule library and obtains real-time carbon allowance prices, policy changes, and market supply and demand data via a RESTful API. Based on the carbon emission margin and compliance period provided by the monitoring module, the smart contract engine automatically generates allowance purchase / sale recommendations and performs compliance verification. Trading instructions are encrypted with AES-256 and pushed to external carbon exchanges, simultaneously updating the enterprise's ERP system and supporting real-time visualization analysis of carbon asset gains and losses.

[0052] Module 6 of the cloud service platform is built upon a microservice architecture and digital twin technology to construct a comprehensive management hub. The platform deploys a Kubernetes cluster, supporting the elastic scaling of microservices such as energy data storage, multi-objective optimization model training, and equipment health diagnostics. The digital twin engine maps physical devices to virtual entities, driving simulation operations with real-time data to predict equipment failures and optimize maintenance strategies. The platform provides a visual dashboard function, integrating key indicators such as energy efficiency, carbon emission trends, and transaction records. It supports multi-terminal access via PC and mobile devices and offers an open SDK for third-party developers to customize industry-specific analysis tools.

[0053] From the above, we can conclude that: This invention significantly improves the efficiency and economy of energy management through the dynamic decision-making capabilities of a multi-objective optimization engine. Based on real-time energy data and environmental variables, and combined with deep prediction and reinforcement learning techniques, the engine achieves precise allocation and dynamic balance of multi-source energy. For example, when electricity prices fluctuate or equipment loads change abruptly, the system can quickly adjust the proportions of electrical energy, thermal energy, and waste heat recovery, reducing energy waste and optimizing operating costs. Simultaneously, the energy recovery module converts traditional waste energy into usable resources. Combined with the global scheduling of the cloud service platform, this further reduces dependence on external energy sources, resulting in an overall energy efficiency improvement of over 20% for enterprises. The system supports multi-scenario adaptive strategies, ensuring production continuity while balancing short-term cost control and long-term equipment health management, providing cost-effective energy solutions for industrial and commercial users.

[0054] This invention integrates carbon emission monitoring and carbon trading mechanisms to construct a closed-loop carbon reduction path of "monitoring-optimization-trading". The multi-objective optimization engine uses carbon emission intensity as a core constraint, dynamically limiting the proportion of high-polluting energy use, and ensures the compliance and transparency of strategy execution through blockchain technology. The carbon trading interface connects to market data in real time, automatically generating quota trading suggestions to help enterprises convert emission reduction results into economic benefits and alleviate carbon tax pressure. Furthermore, the system's predictive model and reinforcement learning algorithm can proactively avoid the risk of exceeding carbon emission limits. Combined with waste heat recovery and priority scheduling of clean energy, it helps enterprises achieve carbon neutrality goals and promotes the green and low-carbon transformation of the industrial sector.

[0055] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the term "comprising" or any other variations thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0056] 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 the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent multi-source energy collaborative optimization system, characterized in that: The system includes a data acquisition module (1), a multi-objective optimization engine module (2), an energy recovery module (3), a carbon emission monitoring module (4), a carbon trading interface module (5), and a cloud service platform module (6). The multi-objective optimization engine module (2) is internally equipped with a feature engineering module (7), a deep prediction model module (8), a reinforcement learning strategy generation module (9), and a blockchain verification and execution module (10). The real-time multi-source energy data of the data acquisition module (1) is transmitted to the input of the multi-objective optimization engine module (2) through the output end, driving the latter to generate a dynamic optimization strategy; The output of the multi-objective optimization engine module (2) simultaneously sends an energy recovery command to the input of the energy recovery module (3) and pushes optimization parameters to the input of the carbon emission monitoring module (4) to support carbon emission monitoring. The output of the carbon emission monitoring module (4) is further connected to the input of the carbon trading interface module (5) to input real-time carbon emission data into the carbon trading interface module; The outputs of the data acquisition module (1), multi-objective optimization engine module (2), energy recovery module (3), carbon emission monitoring module (4), and carbon trading interface module (5) all aggregate data to the input of the cloud service platform module (6), and the cloud completes global integration and analysis. The output of the cloud service platform module (6) is connected in reverse to the input of the multi-objective optimization engine module (2) via a feedback link.

2. The intelligent multi-source energy collaborative optimization system as described in claim 1, characterized in that: The data acquisition module (1) is equipped with a low-power, high-precision intelligent sensor network and a self-organizing network communication unit. The sensor nodes are deployed at the terminals that use various energy sources such as electricity, heat, and natural gas. The TI CC2650 series chip is used to achieve millisecond-level data acquisition, covering multiple parameters such as voltage, current, temperature, and flow.

3. The intelligent multi-source energy collaborative optimization system as described in claim 1, characterized in that: The feature engineering module (7) integrates real-time data and environmental parameters from multiple energy sources, including electricity, heat, and natural gas. It eliminates noise interference and unifies the dimensions through data cleaning, noise reduction, and normalization to ensure data consistency and availability. It uses sliding window sampling technology to extract time-series features, capture the fluctuations in energy demand and the periodic patterns of waste heat emissions, and integrates historical energy consumption patterns with external variables such as carbon price fluctuations, real-time electricity prices, and weather forecast data to construct a dynamically labeled dataset to support model training and inference.

4. The intelligent multi-source energy collaborative optimization system as described in claim 1, characterized in that: The deep prediction model module (8) uses a deep neural network with a Transformer architecture to accurately predict the future supply and demand trends of various energy sources, including electricity, heat, and natural gas. This module integrates real-time collected energy consumption data, equipment operating status indicators, historical energy consumption patterns, and external environmental variables such as temperature, light intensity, and carbon price fluctuations, and dynamically captures the correlation of energy demand at different time scales through a hierarchical self-attention mechanism. Input features include energy consumption rate, equipment efficiency curve, energy supply stability, and market dynamics parameters. The model uses time series analysis techniques to generate short-term, medium-term, and long-term energy supply and demand balance forecasts. The weighted forecast formula for dynamic features is as follows: Where: w i The baseline weight of the i-th energy type is calculated from historical consumption efficiency and equipment performance indicators, reflecting the inherent importance of energy. λ: Time decay coefficient, which controls the rate at which the influence of historical data on the current prediction decays, and is related to the update frequency of real-time data; Δt: The time interval between the current time and historical data, used to quantify the timeliness of the data; c i Environmental impact factors, dynamically adjusted based on real-time carbon emission intensity and carbon price, are used to strengthen the forecasting priority of low-carbon energy; f i : The input feature vector of the i-th energy type, which includes real-time consumption rate, supply stability and external correlation variables.

5. The intelligent multi-source energy collaborative optimization system as described in claim 1, characterized in that: The reinforcement learning strategy generation module (9) is based on the output of the deep prediction model and real-time energy data. It uses a near-end strategy optimization algorithm to build a dynamic decision framework with the dual objectives of minimizing operating costs and maximizing energy efficiency, driving multi-source energy collaborative optimization. The module defines the state space as the energy supply and demand difference, equipment health, carbon emission exceedance risk level and real-time carbon price fluctuation. The action space covers the energy allocation ratio adjustment, waste heat recovery device start and stop instructions, and energy storage equipment charging and discharging strategy multi-dimensional control variables. Through Markov decision process modeling, the module combines the immediate reward function and the long-term value function to dynamically generate energy allocation strategies. The reward function integrates the amount of operating cost savings, the increase in energy utilization rate and the amount of carbon emission reduction, and dynamically adjusts the weight of each objective according to the external carbon emission quota and market carbon price to achieve adaptive switching of strategy priority. After the strategy is executed, the actual energy consumption and carbon emission data are fed back to the module to form a closed-loop learning mechanism, continuously optimize the strategy generation logic, and ensure the stability and low-carbon orientation of the system in complex scenarios. The formula for calculating the dynamic multi-objective reward function is as follows: R=α·C saved +β·U improved +γ·E reduced ; in: α: Cost-saving weighting factor, which is positively correlated with real-time electricity prices and energy procurement cost fluctuations, and dynamically amplifies cost-saving benefits in cost-sensitive scenarios; C saved It represents the amount of operating cost savings per unit of time, calculated from the difference between the actual cost and the benchmark cost; β represents the energy efficiency weighting factor, which is related to the health of the equipment and the historical energy efficiency baseline. When the equipment condition deteriorates, this weight is automatically increased to maintain system efficiency. U improved It indicates the improvement in energy efficiency, quantified by the ratio of real-time energy conversion rate to the ideal value; γ represents the carbon emission reduction weighting factor, which increases linearly with rising carbon prices and decreasing carbon emission allowances, thus forcing the strategy to tilt towards low-carbon practices. E reduced This represents the reduction in carbon emissions, calculated based on the difference between real-time emission data provided by the monitoring module and the target threshold.

6. The intelligent multi-source energy collaborative optimization system as described in claim 1, characterized in that: The blockchain verification and execution module (10) is built on the Hyperledger Fabric framework to construct a private chain network, deploy verification nodes on edge computing devices and cloud servers to achieve decentralized consensus across physical locations; the smart contract has a pre-set multi-source energy coordination rule library, covering carbon emission thresholds, energy allocation ratios, and core strategies for equipment safety parameters. When the optimization engine generates instructions to be executed, the verification engine synchronously triggers contract logic verification, verifies data integrity through asymmetric encryption algorithms, and ensures that the strategy complies with industry norms and enterprise internal standards; the execution terminal interface uses the OPC UA protocol to interconnect with industrial equipment, sends verified instructions to the energy recovery subsystem and energy storage device execution unit in real time, and collects equipment feedback data to write into the blockchain ledger to form an immutable audit traceability chain.

7. The intelligent multi-source energy collaborative optimization system as described in claim 1, characterized in that: The energy recovery module (3) is equipped with a high-temperature resistant stainless steel heat exchanger and a micro thermoelectric energy conversion device. The heat exchanger is installed at the industrial waste heat discharge port and maximizes the waste heat capture efficiency through a multi-layer fin structure, with a thermal conductivity of 400W / (m·K).

8. The intelligent multi-source energy collaborative optimization system as described in claim 1, characterized in that: The carbon emission monitoring module (4) adopts a multispectral infrared sensor array and a scene adaptive analysis algorithm. The sensor nodes are deployed at key locations in the flue and exhaust outlet. The scene analysis algorithm constructs a dynamic carbon emission model based on equipment operating status, ambient temperature and humidity and production rhythm data to realize risk warning and source tracing analysis of exceeding standards.

9. The intelligent multi-source energy collaborative optimization system as described in claim 1, characterized in that: The carbon trading interface module (5) is internally equipped with an integrated carbon market API adapter and a smart contract engine; the module has a built-in multi-region carbon trading rule library, and obtains carbon quota prices, policy changes and market supply and demand data in real time through RESTful API; the smart contract engine automatically generates quota purchase / sale suggestions based on the carbon emission surplus and compliance cycle provided by the monitoring module, and performs compliance verification.

10. The intelligent multi-source energy collaborative optimization system as described in claim 1, characterized in that: The cloud service platform module (6) is built on a microservice architecture and digital twin technology to create a full-domain management hub; the platform deploys a Kubernetes cluster to support the elastic expansion of energy data storage, multi-objective optimization model training, and equipment health diagnosis microservices. The digital twin engine maps physical devices to virtual entities, drives simulation operations through real-time data, predicts device failures, and optimizes maintenance strategies.

Images