FSRU re-gasification heat source intelligent switching control system based on multi-source coupling optimization

Abstract

The application relates to the technical field of FSRU regasification control systems, and particularly discloses an FSRU regasification heat source intelligent switching control system based on multi-source coupling optimization, which comprises the following: a multi-source hierarchical acquisition and edge preprocessing module, which is used for acquiring five-dimensional basic data of a heat source-load-environment-heat storage-exception and dynamic operation constraint data, and performing hierarchical transmission and normalization processing on the data; and a dual-mode adaptive multi-source-heat storage coupling model module, which is used for pre-establishing a lightweight coupling model and a high-precision coupling model, and dynamically switching between the lightweight coupling model and the high-precision coupling model according to a working condition trigger signal. By introducing a transient asynchronous compensation mechanism, the application effectively solves the problem of a control empty window period caused by time-consuming high-precision model calculation, ensures that the system can still maintain smooth and continuous instruction output in an extreme load fluctuation instant, and eliminates the pressure shock risk caused by model switching.

Description

Technical Field

[0001] This invention relates to the field of FSRU regasification control system technology, and in particular to an intelligent switching control system for FSRU regasification heat source based on multi-source coupling optimization. Background Technology

[0002] As a core node in the liquefied natural gas (LNG) industry chain, the stability of the regasification process of the floating storage and regasification unit (FSRU) directly determines the safety of downstream gas consumption. FSRU systems often need to couple multiple heat sources such as waste heat recovery, electric heating, and seawater heat exchange, and integrate thermal storage devices to balance system energy efficiency.

[0003] However, heat source scheduling in a multi-source coupled environment faces extremely high nonlinear challenges. Traditional fixed logic control struggles to find the global optimal solution amidst complex energy price constraints, equipment loss limitations, and transient load fluctuations, making it difficult for the overall system energy efficiency to reach an ideal level.

[0004] Currently, although the industry is attempting to introduce high-precision physical models or multi-objective optimization algorithms to improve control quality, significant conflicting technical challenges remain in actual operation:

[0005] On the one hand, in order to cope with sudden load changes, the system must have an extremely high response speed. However, high-precision coupled models that can guarantee control accuracy often have a huge amount of computation, resulting in a serious physical delay conflict between computation time and real-time requirements when the model is triggered to switch.

[0006] On the other hand, in the process of multi-objective optimization, there is an inherent mutual exclusion between energy cost, carbon emission reduction target and equipment life guarantee. Weight drift in dynamic environment can easily lead to the divergence of algorithm search trajectory and the frequent jitter of control commands. This mismatch between response accuracy and computation time, as well as the contradiction between multi-objective trade-offs and convergence stability, has become a technical bottleneck restricting the realization of intelligent and refined switching control of FSRU regasification system. Summary of the Invention

[0007] This invention aims to at least partially solve one of the technical problems in related technologies. Therefore, the purpose of this invention is to propose an intelligent switching control system for the regasification heat source of an FSRU based on multi-source coupling optimization, so as to achieve accurate and stable switching of the regasification heat source.

[0008] To achieve the above objectives, a first aspect of the present invention proposes an intelligent switching control system for the regasification heat source of an FSRU based on multi-source coupling optimization, comprising:

[0009] The multi-source hierarchical acquisition and edge preprocessing module is used to acquire five-dimensional basic data of heat source, load, environment, heat storage and anomaly, as well as dynamic operation constraint data, and to perform hierarchical transmission and normalization processing on the data.

[0010] The dual-mode adaptive multi-source-thermal storage coupling model module is used to pre-build a lightweight coupling model and a high-precision coupling model, and dynamically switch between the lightweight coupling model and the high-precision coupling model according to the working condition trigger signal;

[0011] The hybrid optimization and pre-calculation module is used to construct a multi-objective optimization function that includes heat source utilization efficiency, energy cost and equipment life, and outputs heat source scheduling strategy through hybrid optimization algorithm or typical operating condition pre-calculation library;

[0012] The predictive dynamic decision-making and switching module is used to predict load and energy prices based on dual time windows, and dynamically adjust the priority of heat sources according to the prediction results to achieve coordinated switching of multi-loop heat sources.

[0013] The hardware and software co-acceleration module is used to perform matrix operations of the coupled model through a dedicated computing unit and to execute the hybrid optimization algorithm through multi-core parallel computing.

[0014] The graded anomaly tolerance processing module is used to identify different levels of operational anomalies and adapt the corresponding fault tolerance compensation strategy in combination with the dynamic operational constraint data.

[0015] To achieve the above objectives, a second aspect of the present invention proposes a method for intelligent switching control of FSRU regasification heat source based on multi-source coupling optimization, comprising the following steps:

[0016] Multi-source hierarchical acquisition and edge preprocessing steps: Acquire five-dimensional basic data and dynamic operation constraint data of heat source-load-environment-heat storage-anomaly, and divide the five-dimensional basic data into first-level core data, second-level auxiliary data and third-level redundant data for hierarchical transmission. The data is then subjected to anomaly identification and normalization processing through the edge computing unit.

[0017] Dual-mode adaptive modeling and transient compensation steps: Based on the monitored working condition trigger signal, dynamically switch between the pre-built lightweight coupling model and the high-precision coupling model; during the transient period when the lightweight coupling model is switched to the high-precision coupling model, start the asynchronous parallel computing path, use the dynamic prediction compensation operator generated by the transient asynchronous compensation module to correct the lightweight output command, and correct the compensation parameters by aligning the residuals after the high-precision calculation is completed;

[0018] Hybrid optimization scheduling steps: Construct a multi-objective optimization function that includes heat source utilization efficiency, energy cost and equipment life, and adaptively adjust the weight of each sub-objective according to the operation scenario label; when the gradient direction cosine value of the quantification of the conflict sensitivity between sub-objectives exceeds the preset conflict threshold, the multi-objective optimization function is converted into a constraint-guided optimization mode, and the optimization algorithm is guided to converge in the feasible region by introducing a penalty operator, and the heat source scheduling strategy is output.

[0019] Predictive dynamic decision-making and multi-loop switching steps: Based on dual time windows, load and energy prices are predicted, heat source priorities are dynamically allocated according to the prediction results, and the actuators are controlled to perform coordinated switching of heat sources between independent loop, series loop, or parallel loop modes.

[0020] Hardware and software collaborative acceleration and fault tolerance processing steps: Accelerate the operation of the coupled model and the iteration of the optimization algorithm through dedicated computing units and multi-core parallel computing architecture, and identify the level of abnormal operation in real time to adapt to the corresponding fault tolerance compensation strategy.

[0021] To achieve the above objectives, a third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory. When the computer program is executed by the processor, it implements the above-described intelligent switching control method for FSRU regasification heat source based on multi-source coupling optimization.

[0022] The intelligent switching control system for FSRU regasification heat source based on multi-source coupling optimization in this invention achieves precise and stable switching of the regasification heat source under complex and variable operating conditions through multi-module collaboration. Specifically:

[0023] By introducing a transient asynchronous compensation mechanism, this system effectively solves the control window problem caused by the time consumption of high-precision model calculations, ensuring that the system can maintain smooth and continuous command output even during extreme load fluctuations, and eliminating the risk of pressure oscillation caused by model switching. At the same time, the system overcomes the non-convergence problem of multi-objective optimization in the weight drift environment by using conflict sensitivity assessment and constraint-guided dynamic transformation of optimization mode, significantly improving the determinism and consistency of the control strategy, and effectively extending the service life of core heat exchange equipment while ensuring the efficiency of heat source utilization.

[0024] Furthermore, the integration of the hardware and software collaborative acceleration architecture and the hierarchical fault-tolerant mechanism gives the system extremely high robustness, enabling it to adaptively balance energy costs and carbon emission reduction targets, and providing a reliable technical means for the safe and green operation of FSRU. Attached Figure Description

[0025] Figure 1This is a schematic diagram illustrating the implementation of the intelligent switching control system for FSRU regasification heat source based on multi-source coupling optimization provided by the present invention.

[0026] Figure 2 This is a flowchart illustrating the intelligent switching control method for FSRU regasification heat source based on multi-source coupling optimization provided by the present invention.

[0027] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0028] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0029] The following description, with reference to the accompanying drawings, describes an embodiment of the present invention: an intelligent switching control system, method, and electronic device for FSRU regasification heat source based on multi-source coupling optimization.

[0030] Example 1:

[0031] This embodiment provides an intelligent switching control system for FSRU regasification heat sources based on multi-source coupling optimization. This system is applied in floating storage regasification units. By integrating multiple heat sources and utilizing heat storage devices for energy balance, it ensures the stability and economy of the LNG regasification process.

[0032] It should be noted that the FSRU regasification heat source intelligent switching control system based on multi-source coupling optimization consists of six core functional modules, namely, multi-source hierarchical acquisition and edge preprocessing module, dual-mode adaptive multi-source-thermal storage coupling model module, hybrid optimization and pre-calculation module, predictive dynamic decision-making and switching module, software and hardware collaborative acceleration module, and hierarchical anomaly fault tolerance processing module.

[0033] Specifically, the multi-source hierarchical acquisition and edge preprocessing module undertakes the fundamental functions of the system's perception layer. This module is responsible for real-time monitoring and acquisition of five basic dimensions of data: heat source, load, environment, heat storage, and anomalies, while also acquiring dynamic operational constraint data. This dynamic operational constraint data includes, but is not limited to, real-time energy prices, equipment wear and tear data, and carbon emission reduction constraint data. The acquired data does not directly enter the core algorithm area but first undergoes hierarchical transmission and edge preprocessing within this module to ensure the quality and timeliness of the massive amount of sensor data.

[0034] Optionally, the hierarchical acquisition mechanism of the multi-source hierarchical acquisition and edge preprocessing module reflects the refined management of data priorities. The system divides the five-dimensional basic data into primary core data, secondary auxiliary data, and tertiary redundant data. Specifically, the primary core data includes key operating parameters of the heat source, regasification load mutation signals, and thermal storage tank status parameters. To address the instantaneous nature of FSRU load fluctuations, the sampling frequency of the primary core data is set to 100 milliseconds per sampling, with the regasification load mutation signal being judged by a change rate of not less than 20% per minute. The secondary auxiliary data includes ambient temperature and humidity, as well as non-critical equipment operating parameters, with a sampling frequency of 1 second per sampling, used to provide environmental field background information. The tertiary redundant data includes historical operating condition backtracking data and equipment maintenance logs, with a sampling frequency of 5 seconds per sampling, mainly used for long-term status analysis and diagnosis. The dynamic operational constraint data is acquired in real time through an external data interface, with energy prices updated every 5 minutes and carbon emission reduction constraint data updated every 30 minutes.

[0035] The edge preprocessing process in the multi-source hierarchical acquisition and edge preprocessing module has extremely high real-time computing capabilities. The system deploys edge computing units at physical locations such as heat source equipment clusters and thermal storage tank monitoring points. These units first perform abnormal data identification, employing a dual judgment logic of threshold and trend. That is, when the parameter change rate of three consecutive sampling points exceeds a set threshold, it is determined to be a physical layer anomaly. Subsequently, the system normalizes the core parameters, mapping data of different dimensions to a preset range of 0 to 1. The calculation formula for the normalization process is expressed as: ;

[0036] In the formula, The target value after normalization; These are the real-time data values ​​collected by the original sensors; This represents the minimum value of this category of data within the current data collection window. The only accurate definition is the maximum value of that category of data within the current acquisition time window. This normalization process eliminates the impact of differences in the magnitude of data from different sensors such as temperature, pressure, and flow rate on the accuracy of the subsequent coupled model and the convergence speed of the optimization algorithm.

[0037] Specifically, the dual-mode adaptive multi-source-thermal storage coupling model module is the core modeling layer of this system. This module is used to pre-build a lightweight coupling model and a high-precision coupling model, aiming to balance computational speed and physical description accuracy. The system dynamically switches between the lightweight coupling model and the high-precision coupling model according to the operating condition trigger signal to adapt to different requirements of steady-state operation and severe transient conditions.

[0038] It is also important to note that the switching logic of the dual-mode adaptive multi-source-thermal storage coupling model module fully considers the complexity of the FSRU operating environment. When the system detects a load fluctuation rate of no more than 10% per minute and the ambient temperature is within a preset stable range, the system is in normal operating condition, and the aforementioned lightweight coupling model is invoked. This model ignores minor thermodynamic coupling effects through piecewise linearization, reducing the computational load by 60% compared to the complete physical model, ensuring an extremely high computational response frequency.

[0039] However, when any of the following conditions are detected: load fluctuation rate is not less than 20% per minute, ambient temperature change rate is not less than 5 degrees Celsius per hour, operational constraint change occurs, or an abnormal signal is received from the graded anomaly fault-tolerant processing module, the system will immediately switch to the high-precision coupled model. Specifically, operational constraint change refers to energy price fluctuations of not less than 20% or carbon quota surplus not exceeding 20%. During the switchover process, the system employs a weighted smooth transition mechanism. In the first three calculation cycles of the initial switchover, the weight of the calculation results of the lightweight coupled model is gradually reduced from 70% to 50% and then to 30%, while the calculation weight of the high-precision coupled model is correspondingly increased. This effectively avoids step oscillations in the control strategy at the moment of model switching.

[0040] Specifically, the high-precision coupling model encompasses deep physical interaction logic, including an energy price-heat source coupling model, an equipment loss-operating parameter coupling model, and a carbon reduction-green heat source coupling model. The energy price-heat source coupling model maps the unit cost efficiency characteristics of each heat source under different energy price ranges, ensuring that low-cost heat sources are prioritized during peak price periods. The equipment loss-operating parameter coupling model determines the real-time loss level of equipment based on the heat source's output power, operating time, and number of start-stop cycles through loss coefficient calculation logic. The formula for calculating the loss coefficient is expressed as: ;

[0041] In the formula, This is the current real-time loss coefficient value of the equipment; This serves as the initial wear and tear baseline coefficient for the equipment after it leaves the factory or has undergone maintenance. This represents the cumulative operating time of the equipment since the last maintenance, expressed in hours. This represents the cumulative number of start-ups and shutdowns of the equipment within the statistical period. The carbon reduction-green heat source coupling model is used to map the correlation between the usage ratio of green heat sources, such as waste heat recovery and solar-assisted heat pumps, and their contribution to carbon reduction in real time, providing data support for low-carbon operation strategies.

[0042] Specifically, the hybrid optimization and budget calculation module is responsible for outputting the final heat source scheduling command. This module constructs a multi-objective optimization function and adaptively adjusts the weight coefficients of each sub-objective according to the operational scenario label. When balancing system performance, the multi-objective optimization function simultaneously considers heat source utilization efficiency, energy cost, system response time, thermal storage efficiency, fault tolerance rate, equipment lifespan guarantee indicators, and carbon emission reduction indicators.

[0043] It should also be noted that the calculation formula for the multi-objective optimization function can be expressed as: ;

[0044] In the formula, The only accurate definition of it is the comprehensive evaluation value of the multi-objective optimization function; Heat source utilization efficiency is the ratio of effective heat energy output to total energy input. The operating energy cost per unit of time; The time period from the change in operating conditions to the system returning to steady state; This refers to the actual energy charging and discharging efficiency of the thermal storage device. This represents the probability of system operational stability under abnormal triggering conditions. This is a performance indicator for ensuring equipment lifespan, specifically equal to 1- ; Carbon emission reduction targets are determined based on the ratio of green heat source use, the remaining carbon quota, and the ratio of total carbon quota. to These are the corresponding dynamic weighting coefficients, and their sum is always equal to one.

[0045] Specifically, the hybrid optimization and pre-computation module employs a two-stage hybrid optimization algorithm. The first stage utilizes a particle swarm optimization algorithm for rapid global optimization, approximating the suboptimal solution space with fewer iterations. The second stage employs a genetic algorithm for refined searching of the local area surrounding the suboptimal solution, using crossover and mutation operations to escape local optima and ultimately lock in the optimal heat source scheduling strategy.

[0046] In addition, this module has established a pre-computation strategy library, covering a variety of typical scenarios including technical operating conditions and operational constraints. When the operating condition features uploaded by the edge preprocessing module have a similarity of no less than 90% with the pre-stored operating conditions in the library, the system directly calls the pre-stored strategy, thereby compressing the response time to an extremely low level.

[0047] Specifically, the predictive dynamic decision-making and switching module is used to execute the final decision output. This module predicts the external environment and internal load based on a dual time window, and dynamically allocates heat source priorities in conjunction with real-time operating status. The predictive mechanism includes short-term and medium-term predictions. The short-term prediction model is used to predict load and price fluctuations within the next 30 seconds, and its main purpose is to provide pre-trigger signals for the hardware execution mechanism; the medium-term prediction model predicts the system state trend within the next 5 minutes, and is used to adjust the charging and discharging heat margin of the heat storage tank in advance to ensure that the system has sufficient adjustment depth when facing long-term fluctuations.

[0048] It is also important to note that the predictive dynamic decision-making and switching module controls the actuator to switch between multiple loop modes. The multi-loop heat source coordinated switching includes independent loop mode, series loop mode, and parallel loop mode. In independent loop mode, the system activates a single high-efficiency heat source based on the principles of low cost and high environmental friendliness. In series loop mode, the system dynamically adjusts the coupling sequence of the heat source and the heat storage tank according to the load fluctuation amplitude. For example, during load ramp-up, the low-temperature reflux liquid is preheated through the heat storage tank before entering the main heat source for heating. In parallel loop mode, by adjusting the flow ratio of the control valves in each branch, multiple heat sources and the heat storage tank can be simultaneously supplied with energy to cope with extreme peak load scenarios.

[0049] Specifically, the hardware-software co-acceleration module provides computing power to ensure the real-time operation of all the aforementioned complex logic. This module employs a heterogeneous computing architecture, comprising a dedicated computing module and a multi-core parallel computing module. The dedicated computing module is responsible for handling large-scale matrix operations in the dual-mode adaptive coupling model and fitness calculation tasks in the optimization algorithm, significantly reducing computational latency through hardware logic pipelines. The multi-core parallel computing module is responsible for breaking down the hybrid optimization algorithm into multiple threads, such as fitness calculation, selection, crossover, mutation, and optimal solution selection, which are deployed and run on different physical cores of the multi-core processor. Through decision instruction pre-generation technology, the system constructs an on-chip cache containing a high-frequency candidate instruction set, resulting in extremely short instruction matching time and ensuring the overall high performance of the system.

[0050] The graded anomaly fault-tolerant processing module forms the last line of defense for the safe operation of the system. This module classifies anomalies into three levels: mild, moderate, and severe, based on anomaly identification signals. For mild anomalies, such as minor sensor fluctuations, the system smooths the situation by fine-tuning the heat source output power and combining it with energy compensation from the heat storage tank. For moderate anomalies, such as a single heat source branch failure, the system immediately isolates the faulty branch and activates the backup heat source based on the current cost or loss priority principle. For severe anomalies, such as core equipment failure, the system quickly enters emergency protection mode, prioritizing full-load heat release from the heat storage tank to ensure the basic load supply to the downstream regasification system, while simultaneously initiating the emergency heat source switching process.

[0051] In summary, the FSRU regasification heat source intelligent switching control system provided in this embodiment, based on multi-source coupling optimization, effectively solves the problems of response lag and insufficient accuracy of traditional control systems when facing complex operational constraints and extreme operating condition fluctuations by means of precise hierarchical data acquisition, dual-mode coupling modeling, multi-objective hybrid optimization and predictive decision-making logic, combined with high-performance software and hardware acceleration methods and hierarchical fault tolerance mechanism.

[0052] It is also important to note that the modules exchange data via a standardized data bus and dedicated high-speed interfaces, ensuring tight alignment between command and data streams. In actual deployment, the system can adaptively and dynamically optimize heat source configuration based on real-time changes in the external energy market and the environmental characteristics of the FSRU site. This not only improves the technical and economic indicators of the regasification process but also provides solid technical support for the intelligent transformation of offshore LNG unloading terminals.

[0053] Specifically, during system execution, the multi-source hierarchical acquisition and edge preprocessing module continuously updates the real-time database, providing the latest boundary conditions for the dual-mode adaptive multi-source-thermal storage coupling model module. The coupling model module selects an appropriate physical description depth based on the stability of the current operating condition and transmits its output physical field information to the hybrid optimization and pre-calculation module. Under the principle of multi-objective balance, the optimization module uses a hybrid algorithm to find the optimal combination of operational variables and sends this strategy to the predictive dynamic decision-making and switching module. Finally, the decision-making module, in conjunction with the instruction cache of the hardware and software co-acceleration module, issues instructions to the control valves and pump actuators of each heat source, forming a highly autonomous and predictive intelligent closed loop throughout the entire control loop.

[0054] Example 2:

[0055] This embodiment focuses on describing the core compensation mechanism of the present invention in dealing with extreme transient conditions. Specifically, this embodiment provides a transient control method for an intelligent switching control system for FSRU regasification heat source based on multi-source coupling optimization. Its core lies in solving the calculation delay problem of high-precision physical models under sudden operating conditions through a transient asynchronous compensation module.

[0056] Specifically, the transient control scenario involved in this embodiment mainly addresses the control stability requirements of floating storage regasification units when encountering drastic fluctuations in end-user gas load or sudden drops in ambient heat source temperature. In the actual operation of FSRUs, sudden changes in regasification load often require the control system to respond within a very short time to maintain the pressure and temperature stability of the ethylene glycol aqueous solution circulation loop. However, as mentioned earlier, coupled models capable of providing extremely high accuracy typically involve solving numerous partial differential equations and performing multidimensional matrix operations, resulting in a single complete calculation time typically on the order of one hundred milliseconds. This exceeds the stringent requirement of a fifty-millisecond response time limit for the system under sudden operating conditions.

[0057] It is also worth noting that, in order to address the transient control window issue where the computation time of the high-precision coupled model exceeds the response time of sudden operating condition switching, this system specifically introduces and integrates a transient asynchronous compensation module. This transient asynchronous compensation module serves as a logical bridge between the hardware / software co-acceleration module and the dual-mode adaptive multi-source-heat storage coupled model module. Its core function is to maintain the continuity of control commands through predictive mathematical compensation during the transition time between the lightweight coupled model and the high-precision coupled model.

[0058] Specifically, during the transient period when the lightweight coupled model switches to the high-precision coupled model, the system does not stop outputting instructions to wait for the calculation results of the high-precision model. Instead, it immediately and synchronously starts the asynchronous parallel computing path. The asynchronous parallel computing path is logically divided into two independent and parallel processing flows: the first path maintains the output through the lightweight coupled model, aiming to utilize the high-frequency characteristics of the lightweight model to ensure that the system does not go out of control within the first fifty-millisecond period; the second path starts the asynchronous calculation task of the high-precision coupled model in the background to prepare a steady-state solution for subsequent precise control.

[0059] For example, the transient asynchronous compensation module, upon sensing a switching action, immediately extracts model deviation features from the system's historical operating condition database. These model deviation features are the residual distribution laws of the lightweight and high-precision models under specific temperature and pressure curves, accumulated over long-term system operation. Based on these features, the transient asynchronous compensation module calculates and generates dynamic predictive compensation operators in real time.

[0060] Specifically, the generation process of the dynamic prediction compensation operator fully considers the physical causes that lead to sudden changes in operating conditions. The calculation formula of the dynamic prediction compensation operator can be expressed as: ;

[0061] In the formula, The number of dynamically predictive compensation operators generated is used to correct the control increment; , , These are the load impact factor, environmental impact factor, and historical experience correction factor. This represents the real-time regasification load change rate; This represents the real-time rate of change of ambient temperature. This is the model bias gradient extracted from the historical feature database. Using this operator, the system can predict the possible correction direction and magnitude of the high-precision model relative to the lightweight model.

[0062] It is also important to note that the transient asynchronous compensation module superimposes the calculated dynamic prediction compensation operator onto the output command of the first path. Specifically, the system fuses the basic scheduling command generated by the lightweight coupled model with the dynamic prediction compensation operator to generate a transient predictive control command for issuance. This process ensures that, in the initial stage of switching, although high-precision calculation results have not yet been generated, the commands issued to the actuators already include deviation pre-corrections based on historical experience and real-time trends, thereby significantly reducing the performance deviation of the lightweight model under extreme conditions.

[0063] For example, the fusion logic for generating transient predictive control commands is defined as follows: The basic control quantity output by the lightweight coupled model at the current sampling time is summed with the dynamic predictive compensation operator, expressed by the following formula: ;

[0064] In the formula, This refers to the transient estimated control command value issued to the heat source actuator; This represents the basic control command value output by the lightweight coupling model in the first path. This approach ensures the smoothness of the command flow and avoids the control lag that may occur in the traditional direct switching mode.

[0065] Specifically, after the high-precision coupled model in the second path completes its long-cycle asynchronous calculation and produces results, the system does not directly overwrite the current predicted instruction. Instead, the transient asynchronous compensation module evaluates the accuracy of the predicted instruction and corrects subsequent compensation parameters by calculating the alignment residual. The formula for calculating the alignment residual is as follows: ;

[0066] In the formula, To align residual values; This is the first precise control solution produced by the high-precision coupling model in the second path.

[0067] It is also important to note that the alignment residual not only reflects the accuracy of the current compensation but is also used to correct subsequent compensation parameters. The system uses the alignment residual as a feedback signal to adjust the weighting factors in the aforementioned dynamic prediction compensation operator formula through a self-learning algorithm. This allows the system to generate prediction commands that more closely approximate the true solution of the high-precision model when facing similar transient conditions in the future. This self-evolutionary mechanism based on alignment residuals enables the transient asynchronous compensation module to continuously optimize over time.

[0068] Specifically, the implementation of this transient asynchronous compensation module has proven highly effective in the actual operation of the FSRU regasification heat source. For example, in extreme winter weather, seawater temperature may drop rapidly due to ocean current fluctuations, causing a sharp decline in the efficiency of the seawater heat exchanger. In such cases, high-precision coupled models require calculating complex icing risk boundaries and heat exchange efficiency curves, increasing computation time. With the transient asynchronous compensation module, the system can predict and add compensation based on load demand the instant the temperature drop is detected, proactively activating electric heaters or auxiliary boiler branches. This avoids the technical risk of excessively low propane cycle pressure due to computational delays.

[0069] For example, the transient asynchronous compensation module is implemented at the hardware level using programmable logic resources within a hardware-software co-acceleration module. The system pre-stores the deviation feature library in the on-chip static random access memory of the dedicated computing unit to ensure deterministic low latency in the data extraction process. When the asynchronous parallel computing path is initiated, the FPGA dedicated computing unit is responsible for high-speed execution of the arithmetic logic of the dynamic prediction compensation operator, while the multi-core CPU synchronously monitors the process status of the high-precision model. This heterogeneous parallel execution mode is the key physical basis for ensuring that transient prediction control instructions can be accurately generated within a fifty-millisecond response window.

[0070] It is also important to note that the switching process of the asynchronous parallel computing path involves not only the issuance of instructions but also the seamless handover of control loop authority. After the alignment residual has been corrected over multiple cycles and converged to a preset range, the system automatically transfers control from the transient asynchronous compensation module to the high-precision coupled model, thus entering the steady-state fine-tuning optimization stage. During this transition, the system gradually weakens the influence of the compensation operator through a weighted attenuation coefficient, ensuring that the opening rate of the heat source regulating valve remains within the mechanically permissible safety threshold.

[0071] Specifically, the transient asynchronous compensation module employs clustering analysis to extract historical deviation features. The system divides common FSRU operating conditions into dozens of typical clusters, each cluster corresponding to a set of offline-optimized initial deviation features. When the system identifies in real time that the current operating condition belongs to a specific cluster, the transient asynchronous compensation module prioritizes calculating the deviation gradient under that cluster. This feature extraction method based on operating condition matching significantly improves the accuracy of the dynamic prediction compensation operator in the initial stages of sudden operating conditions.

[0072] The control logic based on the transient asynchronous compensation module embodies the innovative thinking of this invention in addressing the contradiction between the performance limitations of physical model computation and the real-time requirements of complex industrial control. Through asynchronous parallel architecture design, this invention cleverly bypasses the technical bottleneck of excessively long computation time for a single high-precision model. By utilizing prediction and compensation methods, it maximizes the control accuracy of transient processes without sacrificing response speed, which has significant engineering value for ensuring the safe and stable operation of offshore energy facilities.

[0073] The transient asynchronous compensation module defined in this embodiment is not only applicable to the heat source switching process, but its design concept can also be extended to other high-inertia control components of the FSRU regasification system. In subsequent system iterations, this module can also be combined with machine learning algorithms to achieve predictive compensation for more complex coupled failure scenarios. Through this multi-dimensional transient protection, the intelligent switching control system provided by this invention can maintain robust control of the FSRU regasification process under various extreme operating conditions.

[0074] Example 3:

[0075] This embodiment focuses on the core algorithm logic of the hybrid optimization and pre-computation module in dealing with extreme dynamic environments. Specifically, this embodiment provides a global optimization scheme in the intelligent switching control method of FSRU regasification heat source based on multi-source coupling optimization, which aims to solve the technical problems of non-convergence of algorithm search trajectory and oscillation of control strategy caused by the dynamic adjustment of multi-objective weights with the operation scenario.

[0076] Specifically, the hybrid optimization and pre-computation module in this embodiment not only outputs the optimal heat source scheduling strategy, but also needs to ensure the stability of the search process under complex and ever-changing unsteady conditions. In the actual operation of floating storage regasification units, fluctuations in regasification load, diurnal changes in seawater temperature, and real-time price changes in the international energy market all cause dynamic drift in the weight coefficients of each sub-objective in the aforementioned multi-objective optimization function. Mathematically, this drift is manifested as the topographic deformation of the hypersurface of the objective function. If the traditional fixed-weighted optimization method is still used, the search particles or population generated by the algorithm are prone to repeatedly jumping between multiple mutually exclusive objectives, making it difficult to converge to the optimal solution within the required response time, thus leading to unnecessary frequent adjustment actions by the actuator.

[0077] It is also important to note that, to overcome the non-convergence problem caused by dynamic weight drift, the hybrid optimization and pre-computation module introduces a conflict-sensitive quantification mechanism. Before each iteration of the hybrid optimization algorithm, the system first calculates the gradient direction cosine of each sub-objective function in the current solution space in real time. These sub-objective functions cover the aforementioned heat source utilization efficiency, energy cost, response time, thermal storage efficiency, fault tolerance rate, equipment lifespan assurance indicators, and carbon emission reduction indicators.

[0078] For example, the calculation of the gradient direction cosine value can intuitively reflect the synergy or competition of different optimization objectives at the current control point. When the gradient direction cosine values ​​of two sub-objectives approach one, it indicates that their optimization directions are basically consistent; while when the cosine value approaches negative one, it indicates that the two objectives have a serious mutual exclusion relationship, for example, improving response speed may come at the cost of drastically sacrificing equipment lifespan. The formula for calculating the gradient direction cosine value can be expressed as:

[0079] ;

[0080] In the formula, For the first The sub-objective function and the first The direction cosine result between the individual sub-objective functions; For the first The gradient vector of each sub-objective function in the current solution space coordinates; For the first The gradient vector of each sub-objective function in the current solution space coordinates.

[0081] Specifically, the hybrid optimization and budget calculation module further quantifies the conflict sensitivity between sub-objectives based on the calculated cosine values ​​of multiple directions. This conflict sensitivity is a quantitative indicator characterizing the difficulty of multi-objective coupling in the system. When this value exceeds a preset conflict threshold, it indicates that the system is in a non-steady-state condition where the weights are extremely unstable or the objectives are severely conflicting. In this case, if the weighted average mode continues to be used, the search trajectory will oscillate violently. Therefore, the system automatically converts the multi-objective weighted average mode to a constraint-guided optimization mode to fundamentally change the optimization environment.

[0082] It is also important to note that under the constraint-guided optimization mode, the system's search logic shifts from pursuing a weighted optimum of all objectives to pursuing the extreme value of the core objective while meeting safety baseline requirements. Specifically, the system selects the sub-objective with the largest weight coefficient from seven sub-objectives as the primary optimization objective based on the current status label identified by the multi-source hierarchical acquisition and edge preprocessing module. For example, when the system status label is "high equipment wear," the equipment lifespan guarantee indicator is elevated to the primary optimization objective; while when the status label is "carbon quota shortage," the carbon emission reduction indicator becomes the primary optimization objective.

[0083] For example, the hybrid optimization and budget calculation module maps the remaining six sub-objectives to restrictive boundary conditions with dynamic thresholds. This means that, apart from the primary optimization objective which requires maximization or minimization, the other objectives only need to be maintained within a preset safe operating range. This approach reduces the complexity of the seven-dimensional multi-objective optimization problem into a single-objective optimization problem with multi-dimensional constraints, greatly reducing the topological complexity of the solution space and laying the foundation for achieving deterministic convergence.

[0084] To strictly protect these restrictive boundary conditions from being violated during the search process, the hybrid optimization and budget calculation module introduces a dynamic violation penalty operator based on Lagrange multipliers during the iteration of the hybrid optimization algorithm. The core function of this dynamic violation penalty operator is to apply a large reverse force to the fitness function when the search trajectory attempts to deviate from the safe feasible region, forcibly guiding the particle swarm or genetic population back into the feasible region. It should also be noted that the calculation formula for the dynamic violation penalty operator can be expressed as:

[0085] ;

[0086] In the formula, The penalty term is fed back into the fitness function, the primary optimization objective; For the corresponding number The Lagrange multipliers of the restrictive boundary conditions are used to adjust the penalty intensity; For the first The value of the violation function under each restrictive boundary condition is zero when the candidate solution is within the boundary. It is a natural constant; This is a time penalty factor used to increase the severity of the penalty as the iteration progresses; This represents the current algorithm iteration step.

[0087] Specifically, the transient asynchronous compensation module works in conjunction with the hybrid optimization and pre-computation module. When the penalty operator is applied to the fitness function of the primary optimization objective, the gradient direction of the search algorithm is forcibly corrected. Even if the original weight coefficients drift in the background due to scene changes, the algorithm can still lock onto the direction guided by the primary optimization objective because the system has switched to constraint-guided optimization mode, thus achieving deterministic convergence in the environment of dynamic weight drift.

[0088] It is also important to note that this constraint-guided optimization mode is of great significance in the actual operation of FSRUs. For example, during peak load periods at the receiving terminal, if there is a surge in natural gas spot prices, traditional weighted algorithms may hesitate between cost-saving and supply assurance, causing the dispatching scheme to switch back and forth between increasing seawater pump capacity and starting auxiliary boilers, resulting in pipeline pressure fluctuations. This scheme, however, identifies weight priorities, setting supply assurance as the primary optimization objective and energy cost as the limiting boundary condition. As long as the cost does not exceed the budget limit, the algorithm will search for the fastest load-increasing path in a deterministic trajectory, ensuring the decisiveness of the control logic and the safety of the system.

[0089] For example, the hybrid optimization and pre-computation module, after completing constraint-guided optimization, outputs a heat source scheduling strategy with stronger robustness. Due to the introduction of conflict sensitivity assessment during iteration, the system can identify which targets are current bottlenecks and perform targeted strategy compensation. This gradient-based deep optimization approach enables the system provided by this invention to exhibit a higher level of intelligence than traditional controllers when dealing with extremely complex maritime operating environments.

[0090] The dynamic violation penalty operator also has adaptive learning capabilities. The system records the penalty trajectory under different constraint boundaries in real time and fine-tunes the Lagrange multipliers using historical data. The initial value. This means that as the running time accumulates, the system can more accurately predict which boundaries are likely to be touched, thereby delineating obstacle avoidance routes in the solution space in advance, further shortening the time span from algorithm startup to obtaining a deterministic convergent solution.

[0091] By dynamically transforming a multi-objective optimization problem into a single-objective constraint problem under extreme conditions, this embodiment effectively solves the inherent contradiction between performance optimization and stability that has long existed in the field of industrial control, and provides solid algorithmic support for the efficient operation of the FSRU regasification system.

[0092] Example 4:

[0093] like Figure 2 As shown, this fourth embodiment is a supplement to the process closed loop of the aforementioned system embodiments one, two and three. It focuses on describing the logical evolution and step coordination of an intelligent switching control method for FSRU regasification heat source based on multi-source coupling optimization in the actual operation process. This method aims to solve the mismatch between response accuracy and computation time consumption, as well as the contradiction between multi-objective trade-offs and convergence stability proposed in the background technology.

[0094] Specifically, the method provided in this embodiment consists of five core process steps: multi-source hierarchical acquisition and edge preprocessing, dual-mode adaptive modeling and transient compensation, hybrid optimization scheduling, predictive dynamic decision-making and multi-loop switching, and hardware / software collaborative acceleration and fault-tolerant processing. These steps are strictly aligned in the time dimension, ensuring determinism from bottom-level perception to top-level decision-making.

[0095] It is also important to note that the multi-source hierarchical acquisition and edge preprocessing steps, serving as the input to the entire control process, first collect five-dimensional basic data and dynamic operational constraint data (heat source, load, environment, heat storage, and anomalies) through various sensors and data interfaces. To optimize the core network's transmission efficiency, the system divides the five-dimensional basic data into primary core data, secondary auxiliary data, and tertiary redundant data for hierarchical transmission. Specifically, primary core data serves as a key basis for scheduling decisions and is identified in real time by the edge computing unit. When the parameter change rate of multiple consecutive sampling points exceeds a set threshold, it is determined to be an anomaly. Subsequently, the edge computing unit normalizes the original sampled values ​​by calculating the difference between the original sampled values ​​and the minimum value within the acquisition time period, and then quotienting this difference with the difference between the maximum and minimum values ​​within the same time period, thereby eliminating dimensional differences between different heat source parameters.

[0096] For example, the dual-mode adaptive modeling and transient compensation step directly addresses computational delay conflicts. This step dynamically switches between a pre-built lightweight coupled model and a high-precision coupled model based on monitored operating condition trigger signals such as regasification load change rate, ambient temperature change rate, or operational constraint changes. Specifically, during the transient period of switching from the lightweight coupled model to the high-precision coupled model, the method initiates an asynchronous parallel computation path, where the first control path maintains high-frequency output through the lightweight coupled model. Simultaneously, the transient asynchronous compensation module extracts historical deviation features and generates a dynamic prediction compensation operator, which is then superimposed on the lightweight output command to generate a transient prediction control command for real-time issuance. After the high-precision calculation is completed, the system corrects subsequent compensation parameters by calculating the alignment residual between the high-precision result and the prediction command, thereby ensuring a smooth transition under sudden operating conditions.

[0097] Specifically, after the model outputs physical constraints, the hybrid optimization scheduling step begins global optimization. This step first constructs a multi-objective optimization function that includes heat source utilization efficiency, energy cost, response time, thermal storage efficiency, fault tolerance rate, equipment lifespan assurance indicators, and carbon emission reduction indicators. Based on the current operational scenario label, the system adaptively adjusts the weights of each sub-objective. When the gradient direction cosine of the quantified conflict sensitivity between sub-objectives exceeds a preset conflict threshold, the method switches from a weighted mode to a constraint-guided optimization mode. In this mode, the system selects the sub-objective with the largest weight as the primary optimization objective, transforms the remaining sub-objectives into threshold-based restrictive boundary conditions, and guides the hybrid optimization algorithm to converge within the feasible region by introducing a dynamic violation penalty operator based on Lagrange multipliers. This process effectively solves the problem of search trajectory divergence caused by weight drift mentioned in the background technology, achieving the beneficial effects of improving heat source utilization efficiency and ensuring equipment lifespan.

[0098] The predictive dynamic decision-making and multi-loop switching steps are responsible for translating abstract optimization strategies into concrete physical actions. The system performs multi-scale forecasts of load and energy prices based on dual time windows. Short-term forecasts trigger instantaneous switching actions, while medium-term forecasts pre-adjust the charging and discharging status of the thermal storage tank. Based on the forecast results, the system dynamically allocates heat source priorities and controls the actuators to coordinate switching between independent loop mode, series loop mode, and parallel loop mode. In parallel loops, the flow ratio of each branch is adjusted by controlling valves to achieve a precise ratio of multiple heat sources to the thermal storage tank, thereby achieving an optimal balance between energy costs and load response.

[0099] The hardware-software collaborative acceleration and fault-tolerant processing steps serve as the support and guarantee for the method execution, permeating all the aforementioned process stages. This step utilizes a dedicated FPGA computing module to perform complex coupled model matrix operations, while simultaneously accelerating the iterative efficiency of multi-objective optimization through a multi-core CPU parallel computing architecture. Furthermore, the system identifies operational anomaly levels in real time and adapts differentiated fault-tolerant compensation strategies based on whether the anomaly is mild, moderate, or severe. Under severe anomaly conditions, the heat storage tank is prioritized for heat release to ensure basic load requirements are met.

[0100] In summary, the method steps described in this fourth embodiment are highly related to and mutually support the corresponding module functions in the three system embodiments mentioned above. Through the implementation of this method, the conflicts between response accuracy, computation time, and multi-objective stability mentioned in the background art are effectively resolved, ultimately achieving the industrial application goal of significantly improving the stability, energy efficiency, and carbon emission reduction capabilities of the FSRU regasification process without relying on extreme numerical comparisons.

[0101] Example 5:

[0102] Corresponding to the above embodiments, the present invention also proposes an electronic device.

[0103] like Figure 3 The diagram shows a structural schematic of an electronic device according to the present invention. The electronic device 100 includes a processor 101 and a memory 103. The processor 101 and the memory 103 are connected, for example, via a bus 102. Optionally, the electronic device 100 may further include a transceiver 104. It should be noted that in practical applications, the transceiver 104 is not limited to one unit, and the structure of this electronic device 100 does not constitute a limitation on the embodiments of the present invention.

[0104] Processor 101 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 101 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0105] Bus 102 may include a pathway for transmitting information between the aforementioned components. Bus 102 may be a PCI bus or an EISA bus, etc. Bus 102 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0106] The memory 103 stores a computer program corresponding to the intelligent switching control method for FSRU regasification heat source based on multi-source coupling optimization according to the above embodiments of the present invention. This computer program is executed by the processor 101. The processor 101 executes the computer program stored in the memory 103 to implement the content shown in the aforementioned method embodiments.

[0107] Among them, electronic devices 100 include, but are not limited to: mobile terminals such as laptops and PADs (tablet computers) and fixed terminals such as desktop computers. Figure 3 The electronic device 100 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0108] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A smart switching control system for FSRU regasification heat source based on multi-source coupling optimization, characterized in that, include: The multi-source hierarchical acquisition and edge preprocessing module is used to acquire five-dimensional basic data of heat source, load, environment, heat storage and anomaly, as well as dynamic operation constraint data, and to perform hierarchical transmission and normalization processing on the data. The dual-mode adaptive multi-source-thermal storage coupling model module is used to pre-build a lightweight coupling model and a high-precision coupling model, and dynamically switch between the lightweight coupling model and the high-precision coupling model according to the working condition trigger signal; The hybrid optimization and pre-calculation module is used to construct a multi-objective optimization function that includes heat source utilization efficiency, energy cost and equipment life, and outputs heat source scheduling strategy through hybrid optimization algorithm or typical operating condition pre-calculation library; The multi-objective optimization function includes a weighted summation of sub-objectives: heat source utilization efficiency, energy cost, system response time, thermal storage efficiency, fault tolerance rate, equipment lifespan assurance index, and carbon emission reduction index. The carbon emission reduction index is determined based on the proportional relationship between the proportion of green heat source utilization, remaining carbon quotas, and total carbon quotas. The weight coefficients of each sub-objective in the multi-objective optimization function are adaptively adjusted according to the current operational scenario label. The hybrid optimization and pre-computation module is further configured to: calculate the gradient direction cosine value of each sub-objective function in the solution space of the multi-objective optimization function, and quantify the conflict sensitivity between sub-objectives based on the cosine value; when the conflict sensitivity exceeds a preset conflict threshold, convert the multi-objective optimization function into a constraint-guided optimization mode; in the constraint-guided optimization mode, select the sub-objective with the largest weight as the only optimization term, convert the remaining sub-objectives into restrictive boundary conditions with dynamic thresholds, and introduce a penalty operator to guide the optimization algorithm to converge within the feasible region; The predictive dynamic decision-making and switching module is used to predict load and energy prices based on dual time windows, and dynamically adjust the priority of heat sources according to the prediction results to achieve coordinated switching of multi-loop heat sources. The hardware and software co-acceleration module is used to perform matrix operations of the coupled model through a dedicated computing unit and to execute the hybrid optimization algorithm through multi-core parallel computing. The graded anomaly tolerance processing module is used to identify different levels of operational anomalies and adapt the corresponding fault tolerance compensation strategy in combination with the dynamic operational constraint data.

2. The system according to claim 1, characterized in that, The hierarchical acquisition mechanism of the multi-source hierarchical acquisition and edge preprocessing module is as follows: The five-dimensional basic data is divided into primary core data, secondary auxiliary data, and tertiary redundant data. The primary core data includes heat source operating parameters, regasification load mutation signals, and heat storage tank status parameters, and its sampling frequency is higher than that of the secondary auxiliary data. The secondary auxiliary data includes ambient temperature and humidity, as well as operating parameters of non-critical equipment. The three-level redundant data includes historical operating condition retrospective data and equipment maintenance logs.

3. The system according to claim 2, characterized in that, The edge preprocessing includes: The edge computing unit is used to identify anomalies in the collected data. When the parameter change rate of multiple consecutive sampling points exceeds a set threshold, it is determined to be an anomaly. The core parameters are normalized by calculating the difference between the original data and the minimum value within the collection period, and then performing a quotient operation between the difference and the difference between the maximum and minimum values ​​within the same period to map the parameters to a preset interval.

4. The system according to claim 1, characterized in that, The switching logic of the dual-mode adaptive multi-source-thermal storage coupling model module is as follows: When the load fluctuation rate is not greater than the preset load threshold and the ambient temperature is within the preset stable range, the lightweight coupling model is invoked. When a sudden change in load, a sudden change in ambient temperature, a sudden change in operational constraints, or an abnormal signal is detected, the system switches to the high-precision coupling model. The switching process employs a weighted smooth transition mechanism, which gradually reduces the weight of the calculation results of the lightweight coupling model and correspondingly increases the weight of the calculation results of the high-precision coupling model during multiple calculation cycles in the initial stage of the switching.

5. The system according to claim 4, characterized in that, The high-precision coupling model includes: An energy price-heat source coupling model is used to map the cost-efficiency characteristics of various heat sources under different energy price ranges. The equipment loss-operating parameter coupling model is used to calculate the real-time loss level based on the heat source output power, operating time and number of start-stop cycles using a preset loss coefficient formula. A carbon emission reduction-green heat source coupling model is used to map the relationship between the proportion of green heat source utilization and the contribution value of carbon emission reduction.

6. The system according to claim 1, characterized in that, The predictive mechanism of the predictive dynamic decision-making and switching module is as follows: Short-term forecasting models are used to predict load and price fluctuations in the first time period in the future, which are then used to trigger instantaneous switching actions; The system state in the second time period is predicted by using a medium-term forecasting model, which is used to pre-adjust the heat storage tank's heat charging and discharging state and the priority ranking of heat sources. Wherein, the first duration is shorter than the second duration.

7. The system according to claim 1, characterized in that, The multi-loop heat source coordinated switching includes: Independent loop mode: a single heat source is used based on the principles of low cost and high green ratio. In the series loop mode, the coupling sequence between the heat source and the heat storage tank is adjusted according to the load fluctuation scenario; In the parallel loop mode, multiple heat sources and the heat storage tank can be powered simultaneously by adjusting the valve flow ratio.

8. The system according to claim 4, characterized in that, The system also includes a transient asynchronous compensation module: During the transient period when the lightweight coupling model switches to the high-precision coupling model, the asynchronous parallel computing path is started synchronously. The first path maintains the output of the lightweight coupling model, while the second path initiates the asynchronous computation of the high-precision coupling model. The transient asynchronous compensation module extracts historical deviation features and generates a dynamic prediction compensation operator. The operator is superimposed on the output instruction of the first path to generate a transient prediction control instruction for issuance. After the second path is calculated, the subsequent compensation parameters are corrected by calculating the alignment residual.

Images