Energy station efficient operation system based on AI modeling and operation method thereof

Abstract

The invention relates to the technical field of energy conservation and environmental protection, in particular to an AI modeling-based energy station efficient operation system and an operation method thereof, and the method comprises the following steps: collecting equipment operation data at an energy station site through a multi-source sensor, and uploading the equipment operation data to a cloud through an edge gateway; performing time synchronization, abnormal value elimination, missing value processing and standardization processing on the acquired operation data; carrying out working condition clustering on the operation state of the energy station based on the historical preprocessing data, and establishing a corresponding energy consumption baseline; and performing time sequence prediction on the energy station load and the equipment state based on the real-time data and the historical data, and identifying energy consumption abnormity. According to the invention, the accuracy of energy consumption assessment and abnormity identification can be effectively improved, the risk of sudden failure is reduced, invalid start and stop of equipment and energy waste are reduced, the energy utilization efficiency and the equipment operation reliability are improved, the linkage of operation management and operation and maintenance disposal is realized, and the intelligent operation level and comprehensive operation benefits of the energy station are integrally improved.

Description

Technical Field

[0001] This invention relates to an efficient operation system for energy stations based on AI modeling and its operation method, particularly to an efficient operation system for energy stations based on AI modeling and its operation method, belonging to the field of energy conservation and environmental protection technology. Background Technology

[0002] With the continuous expansion of large-scale public buildings, industrial parks, and complex projects, energy stations, as core facilities for centralized cooling, heating, and power supply, have a direct impact on overall energy consumption and operating costs due to their operational efficiency and stability. Existing energy stations typically rely on building automation systems or human experience for operation and management. While these systems can achieve basic data acquisition and equipment start-up and shutdown control, they still have the following shortcomings in practical applications: First, existing technologies are mostly based on rule control or static threshold control, which lack the ability to adapt to changes in complex operating conditions. They are difficult to maintain efficient operation under conditions such as load fluctuations, weather changes and equipment aging, which can easily lead to energy waste. Secondly, traditional energy management systems typically only have post-event statistical analysis capabilities and lack the ability to predict load change trends and equipment operation risks. Equipment failures are often only discovered after they occur, affecting system safety and continuous operation capabilities. Furthermore, the operation of energy stations involves the coordination of multiple types of equipment and multiple operating conditions. Existing systems have difficulty in effectively distinguishing and modeling different operating conditions, resulting in poor generalization ability of optimization strategies and difficulty in forming replicable and scalable operation optimization solutions. Furthermore, existing technologies lack effective linkage between operational data, equipment models, and operation and maintenance management, resulting in low digitalization levels. Operation and maintenance work relies on manual inspections, which is inefficient and prone to overlooking potential risks, and cannot form a closed-loop optimization mechanism. Therefore, there is an urgent need to improve an efficient energy station operation system and its operation method based on AI modeling to solve the above-mentioned problems. Summary of the Invention

[0003] The purpose of this invention is to provide an efficient energy station operation system and its operation method based on AI modeling, in order to solve the problems in the existing technology that there is a lack of effective linkage between operation data, equipment models and operation and maintenance management, low degree of digitalization, reliance on manual inspection for operation and maintenance work, low efficiency and easy to miss potential risks, and inability to form a closed-loop optimization mechanism.

[0004] To achieve the above objectives, the present invention provides the following technical solution: An efficient operation system for an energy station based on AI modeling and its operation method include the following steps: Step 1: Collect equipment operation data at the energy station site using multi-source sensors and upload it to the cloud via an edge gateway; Step 2: Perform time synchronization, outlier removal, missing value handling, and standardization on the collected runtime data; Step 3: Based on historical preprocessed data, perform operating condition clustering on the energy station's operating status and establish corresponding energy consumption baselines; Step 4: Based on real-time and historical data, perform time-series prediction of energy station load and equipment status, and identify energy consumption anomalies; Step 5: Generate start-up and shutdown strategies and operational parameter optimization schemes for energy station equipment based on the prediction results; Step 6: Perform safety constraint verification on the optimized scheme. After the verification is passed, send control commands to the field control system for execution. Step 7: Adjust the model parameters online based on the execution feedback to achieve a closed loop of operation optimization.

[0005] Furthermore, the operational data mentioned in step 1 includes power consumption, current, temperature, pressure, flow rate, equipment operating status, and vibration data. The edge gateway completes protocol parsing, local caching, and data forwarding.

[0006] Furthermore, in step 2, outlier removal is determined using statistical methods. Short-term missing data is filled in by interpolation, while long-term missing data is marked and a data quality alarm is triggered.

[0007] Furthermore, in step 3, the operating data is divided into operating conditions using a clustering algorithm, and corresponding energy consumption statistical characteristics and typical operating curves are established for each operating condition.

[0008] Furthermore, in step 4, a neural network model based on time-series data is used to predict the load and equipment failure trends of the energy station, and energy consumption anomalies are identified based on the operating condition baseline.

[0009] Furthermore, the optimization decision in step 5 includes equipment start-up and shutdown combination optimization and continuous operation parameter optimization, with the goal of generating a control scheme that minimizes energy consumption and meets operational constraints.

[0010] Furthermore, the safety constraints mentioned in step 6 include the minimum operating time of the equipment, the minimum downtime, the maximum number of start-stop cycles, and the allowable temperature and pressure range of the system. When the control scheme does not meet the safety constraints, it switches to manual confirmation mode.

[0011] Furthermore, in step 7, the parameters of the prediction model and the digital twin model are corrected by real-time running data, and model retraining is triggered when the model prediction accuracy is lower than a set threshold.

[0012] Furthermore, the method further includes associating and displaying the anomaly identification results and prediction results with the three-dimensional model of the energy station, and generating operation and maintenance tasks based on the results.

[0013] Furthermore, it includes a sensor subsystem for collecting operational data from the energy station; The edge processing subsystem is used for data aggregation, preprocessing, and uploading. A cloud-based processing platform is used for data storage, operational condition modeling, predictive analytics, and optimization decision-making. The control execution subsystem is used to receive optimization instructions and control the operation of field equipment; The visualization and operation and maintenance subsystem is used to display the operating status, anomaly information, and form an operation and maintenance closed loop. The system is configured to execute steps 1-7 of the method to achieve efficient, safe and intelligent operation of the energy station.

[0014] The present invention has at least the following beneficial effects: This invention establishes an AI-based operating condition model and prediction model by uniformly collecting and processing multi-source operating data from energy stations. It predicts load changes and equipment status under different operating conditions and optimizes equipment start-up and shutdown strategies and operating parameters under safety constraints, achieving dynamic perception and closed-loop optimized control of the energy station's operating status. This technical solution effectively improves the accuracy of energy consumption assessment and anomaly identification, reduces the risk of sudden failures, minimizes ineffective equipment start-up and shutdown and energy waste, improves energy utilization efficiency and equipment reliability, and simultaneously achieves linkage between operation management and maintenance, thereby enhancing the overall intelligent operation level and comprehensive operational efficiency of the energy station. Attached Figure Description

[0015] Figure 1 This is a flowchart of an efficient operation method for an energy station based on AI modeling, according to the present invention. Figure 2 This is a system block diagram of an efficient energy station operation system based on AI modeling, according to the present invention. Detailed Implementation

[0016] The following will describe in detail the implementation of this application with reference to the accompanying drawings and embodiments, so that the implementation process of how this application uses technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly.

[0017] This invention proposes an efficient operation system for energy stations based on artificial intelligence modeling and its usage method. Through the collaborative work of multi-source data acquisition, operating condition modeling, predictive analysis, and optimized control, it achieves intelligent and refined management of energy station operation. The specific technical effects are reflected in the following aspects: First, this invention systematically collects and preprocesses the operating data of energy stations and establishes a stable operating condition baseline based on clustering methods, so that energy stations have corresponding energy consumption reference models under different loads and operating conditions, thereby significantly improving the accuracy of energy consumption assessment and anomaly identification. Secondly, this invention introduces an artificial intelligence prediction model based on time-series data to predict the load change trend and equipment operating status of energy stations in advance, thereby enabling early identification of abnormal energy consumption and equipment failures, reducing the risk of sudden failures, and improving the reliability and safety of system operation. Furthermore, this invention constructs an optimized decision-making mechanism based on the prediction results. By jointly optimizing the equipment start-up and shutdown strategy and continuous operation parameters, it automatically generates the optimal operation plan under the premise of meeting safety constraints, thereby effectively reducing energy consumption and reducing ineffective equipment start-ups and shutdowns, and improving equipment service life. Furthermore, this invention continuously corrects the model parameters through an online update mechanism, enabling the digital model to dynamically evolve with changes in equipment operating status and working conditions, avoiding model failure, and forming a data-driven closed-loop optimization process. This invention combines operational analysis results with visualization and operation and maintenance management to achieve linkage between anomaly location, operation and maintenance task generation and execution feedback, thereby improving operation and maintenance efficiency, reducing reliance on manual labor, and achieving an overall improvement in the operational efficiency, stability and management level of energy stations. It can effectively solve the problems of high energy consumption, insufficient prediction capability and discontinuous optimization control in the operation of existing energy stations, and has good engineering practicality and promotion value.

[0018] Example 1 (Field Pilot Deployment) Applicable scenarios: On-site pilot deployment and integrated operation and maintenance of energy stations (refrigeration room and boiler room) in public health centers.

[0019] Objectives: Achieve real-time energy consumption monitoring, an accuracy rate of ≥80% in energy consumption anomaly diagnosis, an accuracy rate of ≥85% in fault prediction, and gradually approach 10% in energy consumption reduction target during the trial operation phase, while striving for ≥10% energy consumption reduction during the formal operation period; data synchronization delay target ≤10 seconds.

[0020] Hardware and Network List Sensors: Power and current acquisition: single-phase or three-phase energy meters connected in parallel with current transformers (CTs), one set of CTs for each unit's main circuit; Temperature: Pt100 or digital temperature probes for supply water, return water, steam, and machine room environmental points; Pressure: pump outlet pressure, system back pressure; Flow meter: circulating water flow; Valve position acquisition and inverter status: opening encoder and inverter communication port; Vibration sensor: at key motor bearings; Sampling frequency: 1 time per second for core equipment, 1 time per minute for secondary points; Edge gateway: industrial grade, supporting Modbus TCP / RTU, OPC UA, BACnet, with local caching (link failure caching for no less than 24 hours) and local rule alarm capabilities (threshold alarm response ≤ 2 seconds); Field PLC / BEMS: used to receive control commands and execute unit start / stop, inverter settings, valve opening adjustments, etc.; Server / cloud platform: time series database, model training and inference nodes, model repository, visualization front-end, API management services.

[0021] Software modules and algorithms Data fusion and preprocessing module: time synchronization, missing value strategy (short-term ≤5 minutes linear or cubic spline interpolation, long-term >5 minutes annotation and generation of data quality work orders), anomaly removal using 3σ and box plot methods; The load condition modeling module uses K-means clustering for load condition classification. The number of clusters k is determined by maximizing the profile coefficient. The clustering results preserve typical load curves and energy efficiency boundaries. Prediction module: Employs Attention-LSTM, a two-layer LSTM network with 128 hidden units and single-head attention. Training employs 5-fold cross-validation, and SMOTE or focal loss is used for imbalanced faulty samples. Early stopping is enabled during training. Anomaly identification module: Based on baseline comparison of operating conditions and combined with SVM discrimination, online energy consumption anomaly classification is achieved.

[0022] Optimize the scheduling module (two levels) Global combinatorial optimizer: Genetic algorithm, population size 50, number of generations 100, crossover rate 0.8, mutation rate 0.05. Fitness function F = 0.6 energy cost + 0.3 expected maintenance cost + 0.1 comfort penalty. Global optimization is used for unit start-up / shutdown and unit selection; Local Continuous Optimizer: MPC, prediction window 30 minutes, control step size 5 minutes, used for continuous optimization of water supply temperature and pump frequency.

[0023] Safety constraints: Enforce a minimum continuous operating time of 10 minutes, a minimum downtime of 5 minutes, a maximum of 10 start-stop cycles per day, and upper and lower limits for temperature and pressure provided by the equipment manufacturer. If any constraint is violated, the optimization suggestion will be converted to manual confirmation and a work order will be generated.

[0024] BIM visualization and operation and maintenance linkage: Anomalies or predictions are highlighted in the 3D model and work orders are automatically generated. Work orders include fault categories, suggested handling, and spare parts lists.

[0025] Implementation steps Equipment installation and access: Deploy sensors according to the point table, connect the edge gateway to the PLC / BEMS, complete protocol adaptation, and verify data reporting. Target data synchronization delay ≤ 10 seconds.

[0026] Data Acquisition and Preprocessing: Initial data collection and preprocessing were completed after 2 months of operation, data quality was checked and sensor issues were corrected.

[0027] Offline model training: Attention-LSTM and SVM were trained using data from the first two months and backtested on a laboratory simulation platform.

[0028] Joint debugging: Deploy the model to the cloud inference node, and integrate the debugging optimizer with PLC, BIM display and operation and maintenance.

[0029] Trial operation: One month of trial operation, recording energy consumption, fault warnings, and actual fault conditions for cross-validation. Target: Energy consumption anomaly diagnosis ≥80%, fault prediction ≥85%.

[0030] Formal Operation Validation: 3 months of formal operation to evaluate energy consumption reduction rate, false alarm rate, and user satisfaction. Target energy consumption reduction ≥10%, system false alarm rate ≤5%.

[0031] verify Data synchronization delay ≤10 s; data processing accuracy ≥95%.

[0032] The accuracy rate for diagnosing energy consumption anomalies is ≥80%; the accuracy rate for fault prediction is ≥85%.

[0033] During normal operation, energy consumption is reduced by ≥10%; false alarm rate is ≤5%; and user satisfaction is ≥85%.

[0034] Example 2 (Research and Development Verification and Transfer Learning) Applicable scenarios: Development benches and small-scale test stations for algorithm development, joint simulation of physical models, and multi-site deployment. This embodiment is used for rapid algorithm iteration, generating synthetic fault samples, and forming a transferable model package.

[0035] Background and Objectives: To run physical and behavioral models in parallel on a small laboratory testbed, forming a dual-model validation process of "physical model comparison + data-driven model," generating a pre-trained model package that can be transferred to the field, shortening field parameter tuning time and reducing reliance on a large number of field failure samples. Objectives: ≥75% energy consumption anomaly diagnosis and ≥80% fault prediction in the laboratory, providing an initial version of the model for field deployment.

[0036] The test bench and simulation environment constitute Bench hardware: one small chiller, two circulating pumps, a heat load simulator (electric heater), a flow throttling device, and a collection of programmable valves and sensors (refer to the site checklist but with reduced quantities).

[0037] Simulink physical model: Establish energy balance and fluid resistance models for the chiller and pump system to generate baseline data for comparison.

[0038] Data generation and fault simulation Normal operating condition data: The test bench is operated under various external temperatures and loads, and at least 500 hours of operating data are recorded as training data.

[0039] Fault scenario simulation: Typical faults such as motor overload, valve jamming, partial pipeline blockage, and flow sensor drift are sequentially injected, and the temporal characteristics before and after the fault are recorded to form a controllable fault sample library. This method is used to supplement the insufficient training of sparse fault samples in the field.

[0040] Algorithm development process and parameters Operating conditions were defined: after merging laboratory and simulation data, K-means clustering was used, with k selected by the silhouette coefficient. Typical curves and energy efficiency boundaries were generated for each cluster.

[0041] Behavioral model training: First, a two-layer, 128-hidden-unit Attention-LSTM model was trained on laboratory data. Then, transfer learning was used to transfer the model weights to the field model, and mini-batch fine-tuning was performed on the field data. The training strategy used 5-fold cross-validation and SMOTE oversampling for fault classes.

[0042] Optimizer simulation: Genetic algorithm and MPC are used in the simulation environment for joint simulation. The parameters of the genetic algorithm are the same as in Example 1. The impact of different start-stop strategies on COP and energy consumption is evaluated to screen the optimal template for initial on-site scheduling.

[0043] Model transfer and field implementation steps The model is trained on the test bench and a model package is generated, including model weights, preprocessor configuration, test case templates, and initial optimizer configuration.

[0044] When integrating the system on-site, the model package is first deployed and then fine-tuned using data from the first two weeks of the event. The fine-tuning strategy is to train with small learning rate increments to prevent overfitting.

[0045] Run in recommended mode for two weeks, and then enable automatic execution after confirmation by operations and maintenance personnel.

[0046] verify In the laboratory stage, the accuracy rate of energy consumption anomaly diagnosis is ≥75%, and the fault prediction rate is ≥80%.

[0047] Deliverables to the field: pre-trained model package, working condition template library, simulation report, fault sample library, and transfer learning fine-tuning report.

[0048] The foregoing description illustrates and describes several preferred embodiments of the present invention. However, as previously stated, it should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the inventive concept described herein through the foregoing teachings or techniques or knowledge in related fields. Any modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A method for efficient operation of an energy station based on AI modeling, characterized in that, Includes the following steps: Step 1: Collect equipment operation data at the energy station site using multi-source sensors and upload it to the cloud via an edge gateway; Step 2: Perform time synchronization, outlier removal, missing value handling, and standardization on the collected runtime data; Step 3: Based on historical preprocessed data, perform operating condition clustering on the energy station's operating status and establish corresponding energy consumption baselines; Step 4: Based on real-time and historical data, perform time-series prediction of energy station load and equipment status, and identify energy consumption anomalies; Step 5: Generate start-up and shutdown strategies and operational parameter optimization schemes for energy station equipment based on the prediction results; Step 6: Perform safety constraint verification on the optimized scheme. After the verification is passed, send control commands to the field control system for execution. Step 7: Adjust the model parameters online based on the execution feedback to achieve a closed loop of operation optimization.

2. The method for efficient operation of an energy station based on AI modeling according to claim 1, characterized in that: The operational data mentioned in step 1 includes power consumption, current, temperature, pressure, flow rate, equipment operating status, and vibration data. The edge gateway completes protocol parsing, local caching, and data forwarding.

3. The method for efficient operation of an energy station based on AI modeling according to claim 1, characterized in that: The outlier removal in step 2 is determined using statistical methods. Short-term missing data is filled in by interpolation, while long-term missing data is marked and a data quality alarm is triggered.

4. The method for efficient operation of an energy station based on AI modeling according to claim 1, characterized in that: In step 3, the operating data is divided into operating conditions using a clustering algorithm, and corresponding energy consumption statistical characteristics and typical operating curves are established for each operating condition.

5. The method for efficient operation of an energy station based on AI modeling according to claim 1, characterized in that: In step 4, a neural network model based on time series data is used to predict the load and equipment failure trends of the energy station, and energy consumption anomalies are identified based on the operating condition baseline.

6. The method for efficient operation of an energy station based on AI modeling according to claim 1, characterized in that: The optimization decision in step 5 includes equipment start-up and shutdown combination optimization and continuous operation parameter optimization, with the goal of generating a control scheme that minimizes energy consumption and meets operational constraints.

7. The method for efficient operation of an energy station based on AI modeling according to claim 1, characterized in that: The safety constraints mentioned in step 6 include the minimum operating time of the equipment, the minimum downtime, the maximum number of start-stop cycles, and the allowable temperature and pressure range of the system. When the control scheme does not meet the safety constraints, it switches to manual confirmation mode.

8. The method for efficient operation of an energy station based on AI modeling according to claim 1, characterized in that: In step 7, the parameters of the prediction model and the digital twin model are corrected by real-time running data. When the model prediction accuracy is lower than the set threshold, model retraining is triggered.

9. The method for efficient operation of an energy station based on AI modeling according to claim 1, characterized in that: The method further includes associating and displaying the anomaly identification results and prediction results with the three-dimensional model of the energy station, and generating operation and maintenance tasks based on the results.

10. An efficient operation system for an energy station based on AI modeling, as described in any one of claims 1-9, characterized in that: Includes a sensor subsystem for collecting operational data from the energy station; The edge processing subsystem is used for data aggregation, preprocessing, and uploading. A cloud-based processing platform is used for data storage, operational condition modeling, predictive analytics, and optimization decision-making. The control execution subsystem is used to receive optimization instructions and control the operation of field equipment; The visualization and operation and maintenance subsystem is used to display the operating status, anomaly information, and form an operation and maintenance closed loop. The system is configured to execute steps 1-7 of the method to achieve efficient, safe and intelligent operation of the energy station.

Images