Port micro-wind power intelligent regulation method and system based on multi-source data

Abstract

The present disclosure provides a port micro-wind power generation intelligent regulation method and system based on multi-source data, which comprises the following steps: obtaining port multi-source data, collecting meteorological data by deploying sensors in the port area through a multi-source data directional collection unit, collecting operation state data by deploying sensors in the micro-wind power generation equipment, and collecting power generation data by deploying sensors in the port power grid; screening key influence characteristics, correlating the meteorological data, operation state data and power generation data, classifying the data to generate type labels, extracting statistical characteristics and calculating correlation degree values, and screening out meteorological characteristics and operation state characteristics that affect port micro-wind power generation; establishing a linkage prediction model, establishing a micro-wind power generation power prediction model, an equipment state evaluation model and a power grid load prediction model based on the screened characteristics through a three-model linkage prediction unit, outputting corresponding prediction data, and generating difference result data, and then formulating and executing regulation strategies.

Description

Technical Field

[0001] This disclosure pertains to the field of port micro-wind power generation, specifically a smart control method for port micro-wind power generation based on multi-source data. Background Technology

[0002] There are three core problems in the current port micro-wind power generation control: First, the multi-source data collection lacks specificity, with a variety of meteorological, equipment, and power grid data collection devices, parameter ranges that do not match the port environment, and collection frequencies that do not take into account the port's operational rhythm; Second, feature selection lacks quantitative standards, relying solely on experience to remove redundant data, resulting in redundant input features or missing key features in subsequent prediction models; Third, the control strategy is static, leading to problems such as insufficient power supply to core equipment and frequent unnecessary controls. Summary of the Invention

[0003] In view of this, this disclosure proposes a port micro-wind power generation intelligent control method and system based on multi-source data, which is used to solve the technical problems in the prior art, such as lack of specificity in multi-source data acquisition, lack of quantitative standards for feature screening, insufficient power supply to core equipment, and frequent unnecessary control.

[0004] This disclosure provides a smart control method for port micro-wind power generation based on multi-source data, applied to a port micro-wind power generation control system. The control system includes a multi-source data directional acquisition unit, a feature screening and analysis unit, a three-model linkage prediction unit, a difference comparison unit, and a strategy formulation and execution unit. The method includes the following steps:

[0005] To acquire multi-source data from the port, sensors are deployed in the port area to collect meteorological data, sensors are deployed in the micro-wind power generation equipment to collect operational status data, and sensors are deployed in the port power grid to collect power generation data through multi-source data directional acquisition units.

[0006] Key influencing features are screened. Through the feature screening and analysis unit, the meteorological data, operational status data and power generation data are correlated and analyzed. First, the data are classified and type labels are generated. Then, statistical features are extracted and correlation values ​​are calculated to screen out the meteorological features and operational status features that affect the port's micro-wind power generation.

[0007] A linkage prediction model is established. Based on the selected features, a micro-wind power generation prediction model, an equipment status assessment model, and a power grid load prediction model are established through a three-model linkage prediction unit, and the corresponding prediction data is output.

[0008] The difference result data is generated by comparing the actual power generation, equipment status, and grid load with the predicted data point by point and quantifying the deviation.

[0009] Formulate and execute control strategies. Through the strategy formulation and execution unit, set the priority of control scenarios based on the difference result data and current meteorological data, generate intelligent control instructions and issue them for execution. After execution, collect new data to verify the effect. If the target is not met, trigger secondary control.

[0010] This disclosure also provides a port micro-wind power generation control system, which performs the aforementioned control method, the system comprising:

[0011] The multi-source data directional acquisition unit includes at least an ultrasonic wind speed and direction sensor, an electrochemical salt spray sensor, a piezoelectric vibration sensor, an electromagnetic voltage transformer, and a Rogowski coil power sensor, used to acquire port meteorological data, equipment operating status data, and power grid generation data.

[0012] The feature filtering and analysis unit is used to generate classification labels, extract statistical features and calculate the correlation of collected multi-source data, filter key influencing features and remove redundant data.

[0013] The three-model linkage prediction unit integrates a micro-wind power generation prediction model, an equipment status assessment model, and a grid load prediction model to output predicted data on power generation, equipment status, and grid load.

[0014] The difference comparison unit is used to acquire the current actual operating data, compare it with the predicted data, quantify the deviation, and generate difference result data;

[0015] The strategy formulation and execution unit is used to set the priority of the control scenario, generate intelligent control instructions, and send them to the field controller via Modbus or TCP protocol. At the same time, it collects data after control to verify the effect and triggers secondary control.

[0016] In combination with the above technical solutions, the positive effects of the intelligent control method and system for port micro-wind power generation based on multi-source data provided in this disclosure are as follows:

[0017] By deploying sensors in the port area, micro-wind power generation equipment, and port power grid through multi-source data directional acquisition units, meteorological data, equipment operation status data, and power generation data are collected in a directional manner. This achieves matching between data acquisition and the port micro-wind power generation scenario, avoiding the problems of chaotic acquisition equipment models and data types that are out of touch with control requirements in existing technologies. It ensures that the collected data directly serves subsequent feature screening and control decisions, providing basic data support with strong scenario adaptability and high effectiveness for the whole process control.

[0018] By using the feature screening and analysis unit, this disclosure first classifies multi-source data to generate type labels, and then filters key features through a quantitative process of extracting statistical features and calculating correlation values. This replaces the experience-based screening method in the prior art, which avoids the loss of key influencing features and eliminates redundant data, making the input features of the subsequent linkage prediction model more accurate.

[0019] The application strategy formulation and execution unit, by setting the priority of control scenarios, can prioritize the power supply needs of core port scenarios when the differential result data reflects supply and demand imbalances. This avoids the problem of insufficient power supply to core equipment caused by the lack of priority in control in existing technologies, and ensures that critical power needs can be met first when power generation capacity is limited or load fluctuates.

[0020] This disclosure, after generating control commands, verifies the control effect by collecting new data. Only when the target is not met is a secondary control triggered, forming a closed loop of command execution, effect verification, and secondary optimization. This avoids the problems of no feedback and blind and frequent adjustments in existing technologies, reduces the interference of unnecessary control on system stability, and ensures that the control effect meets the target through secondary control, thereby improving the accuracy and operational efficiency of port micro-wind power generation control.

[0021] Furthermore, the power generation, equipment status, and grid load forecast data output by the three-model linkage forecasting unit, combined with the point-by-point comparison and deviation quantification by the difference comparison unit, enable the control decision to be based on forecasts and difference analysis, and to identify supply and demand contradictions and changes in equipment status in advance, replacing the lagging passive control in the existing technology. At the same time, deviation quantification provides a scientific basis for determining the control intensity, further improving the rationality of control decisions. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating the exemplary method implementation steps of this disclosure.

[0024] Figure 2 This is a schematic diagram illustrating the secondary regulation process exemplified in this disclosure. Detailed Implementation

[0025] The technical solutions of this disclosure will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without creative effort are within the scope of protection of this disclosure.

[0026] This disclosure provides a method and system for intelligent control of port micro-wind power generation based on multi-source data. By clarifying the equipment selection and deployment rules for data acquisition, establishing a quantitative feature screening process, and constructing a scenario-based linkage prediction model and dynamic control mechanism, the stability and economy of port micro-wind power generation are improved.

[0027] For details, please refer to Figure 1 and Figure 2 This disclosure provides an intelligent control method for port micro-wind power generation based on multi-source data, applied to a port micro-wind power generation control system. The control system includes a multi-source data directional acquisition unit, a feature screening and analysis unit, a three-model linkage prediction unit, a difference comparison unit, and a strategy formulation and execution unit. The control method includes:

[0028] To acquire multi-source data from the port, sensors are deployed in the port area to collect meteorological data, sensors are deployed in the micro-wind power generation equipment to collect operational status data, and sensors are deployed in the port power grid to collect power generation data through multi-source data directional acquisition units.

[0029] Key influencing features are screened. Through the feature screening and analysis unit, the meteorological data, operational status data and power generation data are correlated and analyzed. First, the data are classified and type labels are generated. Then, statistical features are extracted and correlation values ​​are calculated to screen out the meteorological features and operational status features that affect the port's micro-wind power generation.

[0030] A linkage prediction model is established. Based on the selected features, a micro-wind power generation prediction model, an equipment status assessment model, and a power grid load prediction model are established through a three-model linkage prediction unit, and the corresponding prediction data is output.

[0031] The difference result data is generated by comparing the actual power generation, equipment status, and grid load with the predicted data point by point and quantifying the deviation.

[0032] Formulate and execute control strategies. Through the strategy formulation and execution unit, set the priority of control scenarios based on the difference result data and current meteorological data, generate intelligent control instructions and issue them for execution. After execution, collect new data to verify the effect. If the target is not met, trigger secondary control.

[0033] In this way, by deploying sensors in the port area, the micro-wind power generation equipment, and the port power grid through multi-source data directional acquisition units, meteorological data, equipment operation status data, and power generation data are collected in a directional manner. This achieves the matching of data acquisition with the port micro-wind power generation scenario, avoiding the problems of chaotic acquisition equipment models and data types that are out of touch with control requirements in existing technologies. It ensures that the collected data directly serves subsequent feature screening and control decisions, providing basic data support with strong scenario adaptability and high effectiveness for the whole process control.

[0034] Meanwhile, by applying the feature screening and analysis unit, this disclosure first classifies multi-source data to generate type labels, and then filters key features through a quantitative process of extracting statistical features and calculating correlation values. This replaces the experience-based screening method in the prior art, which avoids the loss of key influencing features and eliminates redundant data, making the input features of the subsequent linkage prediction model more accurate.

[0035] Furthermore, by setting the priority of control scenarios, the application strategy formulation and execution unit can prioritize the power supply needs of the core port scenarios when the differential result data reflects supply and demand imbalance. This avoids the problem of insufficient power supply to core equipment caused by the lack of priority in control in existing technologies, and ensures that critical power demand can be met first when power generation capacity is limited or load fluctuates.

[0036] Furthermore, after generating the control command, this disclosure verifies the control effect by collecting new data. Only when the target is not met is a secondary control triggered, forming a closed loop of command execution, effect verification, and secondary optimization. This avoids the problems of no feedback and blind and frequent adjustments in the existing technology, reduces the interference of unnecessary control on system stability, and ensures that the control effect meets the target through secondary control, thereby improving the accuracy and operational efficiency of port micro-wind power generation control.

[0037] Furthermore, the power generation, equipment status, and grid load forecast data output by the three-model linkage forecasting unit, combined with the point-by-point comparison and deviation quantification by the difference comparison unit, enable the control decision to be based on forecasts and difference analysis, and to identify supply and demand contradictions and changes in equipment status in advance, replacing the lagging passive control in the existing technology. At the same time, deviation quantification provides a scientific basis for determining the control intensity, further improving the rationality of control decisions.

[0038] Based on this, this disclosure utilizes multi-source data targeted acquisition technology, combined with key feature screening methods, to integrate and process three types of heterogeneous data: environmental, equipment, and power grid data. This effectively overcomes data integration bottlenecks and successfully solves the problems of format incompatibility and information silos in traditional data integration. Ultimately, it provides accurate and comprehensive data support for regulatory decisions, completely avoiding the limitations of traditional single-dimensional data in terms of coverage and information depth, making decision-making more scientific and reliable.

[0039] This disclosure achieves coordinated operation of three models—a power generation fluctuation prediction model, an equipment fault diagnosis model, and a load peak-valley prediction model—upgrading the control mode from traditional post-event response to proactive prediction. This technology can identify potential hazards such as abnormal bearings in equipment in advance, significantly shortening the fault detection and handling cycle. It not only significantly increases the average annual power generation of the equipment but also enhances the stability and reliability of equipment operation, reducing losses caused by unplanned downtime.

[0040] This disclosure proposes a dynamic control strategy adapted to port scenarios, specifically addressing the unique environmental factors of ports such as low wind speed, turbulent flow, and salt spray corrosion, thus overcoming the limitations of traditional fixed-parameter control. This strategy effectively solves the problems of energy waste caused by low-wind-speed shutdowns and equipment damage due to high-wind-speed overloads in traditional models. While ensuring stable equipment operation, it optimizes energy utilization efficiency and enhances the adaptability of port energy systems to complex environments.

[0041] This disclosure utilizes precise load forecasting technology, combined with scenario-based power curtailment and multi-entity collaborative strategies, to dynamically balance and regulate the power grid's supply and demand relationship. This technology effectively avoids power supply fluctuations caused by imbalances between grid load and power generation, ensuring the power supply stability of core port equipment such as container cranes. While reducing the risk of power outages, it also lowers the overall energy consumption costs of ports, providing strong technical support for the green transformation of port energy systems.

[0042] In one feasible implementation, the step of acquiring multi-source port data includes:

[0043] The meteorological data includes wind speed, wind direction, temperature, air pressure, and salt spray concentration. The salt spray concentration is collected by an electrochemical gas sensor by detecting the concentration of Cl⁻ ions, while the wind speed and wind direction are collected by an ultrasonic sensor by calculating the time difference between transmitting or receiving ultrasonic waves.

[0044] The operating status data includes generator speed, output power, equipment temperature, vibration amplitude, and cumulative equipment operating time. The vibration amplitude is collected using a piezoelectric vibration sensor, and the output power is calculated by detecting the phase difference between voltage and current using a power sensor.

[0045] The power generation data includes voltage, power, and load. The voltage is collected using an electromagnetic voltage transformer, the power is collected using a Rogowski coil power sensor, and the load is calculated based on the voltage and power data.

[0046] In this way, by clearly collecting wind speed, wind direction, temperature, air pressure, and salt spray concentration, the core environmental factors affecting port micro-wind power generation are fully covered. Compared with the limitations of traditional monitoring that only collects wind speed and temperature, the addition of salt spray concentration and air pressure, such as salt spray concentration as a unique corrosive factor in coastal ports, and air pressure as an influence on air density and thus related to power generation, can comprehensively reflect the impact of the environment on power generation equipment and avoid the one-sidedness of control decisions caused by the lack of environmental parameters.

[0047] It should be noted that the salt spray concentration is collected by an electrochemical gas sensor that detects Cl⁻ ion concentration, which is accurately adapted to the high salt spray environment of ports and coastal areas. This solves the problem of insufficient accuracy of traditional indirect humidity measurement methods for inferring salt spray. It can capture data related to salt spray corrosion risk in real time, providing a direct basis for equipment corrosion prevention and control. Wind speed and direction are collected by an ultrasonic sensor that calculates the time difference between transmitting or receiving ultrasonic waves. It has a fast response speed and strong anti-interference ability, and can accurately capture the instantaneous changes in wind speed and direction caused by the complex turbulent field formed by port buildings and cranes. This avoids the shortcomings of traditional mechanical cup-type sensors, such as response lag and susceptibility to turbulence interference, and ensures that the environmental data is highly matched with the actual wind field conditions.

[0048] In addition, the parameters disclosed herein are comprehensive, covering generator speed, output power, equipment temperature, vibration amplitude, and cumulative operating time. This breaks through the limitations of traditional monitoring that only focuses on electrical parameters, and can simultaneously grasp the equipment's output and health status, avoiding unplanned equipment downtime due to the failure to monitor early signs of mechanical failure.

[0049] Meanwhile, the vibration amplitude is collected using a piezoelectric vibration sensor, which has high sensitivity and can capture micro-vibrations at the 0.01mm level. It can monitor abnormal vibrations of key mechanical components such as generator bearings and gearboxes in real time and provide early warning of mechanical failures. The output power is collected by detecting the phase difference between voltage and current using a power sensor. Compared with the traditional method of directly reading the meter data, it can eliminate the power calculation error caused by voltage and current asynchrony, improve accuracy, ensure that the actual output data of the power generation equipment is true and reliable, and provide an accurate basis for judging the operating efficiency of the equipment.

[0050] In this disclosure, voltage is acquired using an electromagnetic voltage transformer, which has strong overload resistance and good linearity, and can adapt to voltage fluctuations in the port power grid caused by the start-up and shutdown of equipment such as container cranes and refrigerated containers; power is acquired using a Rogowski coil power sensor, which has a wide frequency response range and no magnetic saturation problem, and can accurately acquire power data under complex loads in the port power grid, avoiding the problem of acquisition distortion of traditional sensors under high load and strong interference conditions.

[0051] This disclosure clarifies that the load is calculated based on collected voltage and power data, rather than being directly estimated. Since the voltage and power data are obtained from high-precision sensors adapted to port operating conditions, it ensures that the calculated grid load can accurately reflect the port's power supply and demand situation. This avoids the problem of incorrect prediction of grid supply and demand imbalance caused by inaccurate load data, and provides accurate data support for subsequent grid load and power generation balance regulation.

[0052] In one feasible implementation, the method for generating type labels by classifying data in the step of screening key influencing features includes:

[0053] The meteorological data is processed by selecting raw meteorological data or derived data. The derived data includes turbulence intensity and wind energy density. The data is divided into low wind speed, medium wind speed and high wind speed types according to wind speed range using a clustering algorithm. At the same time, high-density areas of data are identified as typical meteorological types, and meteorological type labels are extracted for each time point.

[0054] The processing of the operating status data involves inputting operating status parameters such as generator speed and equipment temperature, and using a combination of supervised and unsupervised modes to classify the equipment operating conditions and extract the operating status type label for each time point.

[0055] In this way, the generation of classification labels for meteorological data incorporates original meteorological data and derived data such as turbulence intensity and wind energy density. Combined with clustering algorithms, low, medium, and high wind speed types are divided according to wind speed ranges. This accurately adapts to the complex wind field characteristics formed by the dense construction and cranes in ports. Compared with the traditional coarse classification that only relies on a single wind speed data, the derived data can quantify the wind field turbulence interference and actual wind energy potential. The wind speed types divided by clustering can clearly identify the power generation adaptation scenarios under different wind conditions. Furthermore, identifying high-density areas of data as typical meteorological types can focus on the wind conditions that occur frequently in ports. This ensures that the meteorological type label at each time point is highly matched with the actual power generation environment, avoiding the inability of subsequent feature selection to accurately associate the impact of wind conditions on power generation due to fuzzy meteorological data classification.

[0056] To generate classification labels for operational status data, a combination of supervised and unsupervised classification methods is used by inputting core parameters such as generator speed and equipment temperature. This effectively solves the limitations of single-mode classification. The supervised mode can calibrate the operational condition classification standard based on historical fault data to ensure that the labels match the actual operating status of the equipment, such as normal or sub-healthy. The unsupervised mode can capture unlabeled abnormal operating condition types. The combination of the two can comprehensively cover the operating scenarios of port micro-wind power generation equipment, accurately extract the operational status type label at each time point, avoid the label bias caused by traditional classification based solely on experience, and provide an accurate operational condition classification basis for subsequent screening of operational status features that affect power generation, ensuring that feature screening is closely related to the actual operating status of the equipment.

[0057] In one feasible implementation, the extracted statistical features in the step of screening key influencing features include: basic state features, physical related features, and operational state features.

[0058] The basic state characteristics include the mean, peak value, standard deviation, and skewness of the data, which are used to reflect the basic distribution characteristics of the data;

[0059] The physical characteristics include wind energy density and the duration of the characteristics, wherein the wind energy density is calculated based on wind speed and air density;

[0060] The operational characteristics include the matching degree between blade angle and wind direction, the average bearing temperature, and the standard deviation of bearing temperature, which are used to correlate the relationship between equipment operation and power generation efficiency.

[0061] In this way, by extracting basic state features and physically relevant features, a multi-dimensional quantitative characterization of port micro-wind power generation data is achieved, solving the problem that traditional data processing only focuses on single numerical values ​​and cannot reflect the inherent laws of the data. The mean and peak values ​​in the basic state features reflect the overall level and extreme cases of the data, the standard deviation reflects the data fluctuation range, and the skewness reveals the data distribution bias. The combination of these four features comprehensively characterizes the basic distribution characteristics of the data, avoiding feature analysis bias caused by one-sided data descriptions. The wind energy density in the physically relevant features is calculated based on wind speed and air density, directly related to the core energy source of micro-wind power generation. The feature duration reflects the stability of wind conditions or equipment status. Together, these two features combine the data with the physical principles of power generation, ensuring that the extracted features can truly map the environmental energy potential and state persistence, providing a physically logical basis for subsequent correlation of power generation.

[0062] Meanwhile, the precise extraction of operational status features establishes a bridge between equipment operation and power generation efficiency, resolving the disconnect between equipment parameters and power generation efficiency in traditional feature selection. The matching degree between blade angle and wind direction directly affects wind energy capture efficiency, and as a feature, it can intuitively reflect the equipment's adaptability to the wind field. The average bearing temperature reflects the normal operating state of the equipment's mechanical components, while the standard deviation of bearing temperature can capture abnormal temperature fluctuations, such as precursors to faults. The combination of these two features comprehensively reflects the impact of the equipment's mechanical health status on power generation stability. This type of feature directly links equipment operating parameters with power generation efficiency, ensuring that the subsequently selected key features accurately pinpoint how equipment status affects power generation, providing effective support for model prediction of the impact of equipment status on power generation and the development of targeted control strategies.

[0063] In one feasible implementation, the specific process for calculating the relevance value and screening in the step of screening key influencing features includes:

[0064] The correlation degree value was calculated using the Spearman rank correlation coefficient for meteorological characteristics and the correlation degree value was calculated using mutual information value for operational status characteristics.

[0065] Features with correlation values ​​greater than a preset value are retained. The preset value is determined based on historical port data, wherein the Spearman coefficient of meteorological features is preset to be ≥0.6 and the mutual information value of operational status features is preset to be ≥0.5.

[0066] If the port is a high-salt-spray environment, the preset value of the salt spray concentration-temperature covariance is reduced to 0.4 and used as a supplementary feature in the screening.

[0067] Redundant features were removed by using a variance inflation factor (VIF) of less than 10. The power generation prediction model was then trained using the selected features to ensure that the accuracy of the test set was better than that of the unselected full-feature model.

[0068] In this way, by employing appropriate correlation calculation methods for meteorological and operational characteristics respectively, combined with preset values ​​verified based on historical port data, precise screening of key influencing features is achieved. This solves the problems of traditional single calculation methods being unable to adapt to different types of features and the screening bias caused by the subjective experience of preset values. The Spearman rank correlation coefficient is used for meteorological features, which can effectively capture the nonlinear relationship between wind speed, salt spray concentration, and power generation, avoiding the limitations of linear calculation methods. Mutual information values ​​are used for operational characteristics, which can accurately quantify the correlation between generator speed, bearing temperature, and power generation, ensuring that no key operational parameters are overlooked. Furthermore, the preset values ​​of Spearman coefficient ≥ 0.6 for meteorological features and mutual information value ≥ 0.5 for operational characteristics are determined based on historical port data verification, rather than subjective settings, ensuring that the selected features do indeed have a significant impact on power generation. Adjusting the preset value of salt spray concentration-temperature covariance to 0.4 for high salt spray environments further adapts to the unique coastal environment of ports, ensuring that key environmental coupling features are not misscreened in this scenario, thus improving the scenario adaptability of feature screening.

[0069] Meanwhile, by eliminating redundant features using a variance inflation factor (VIF) < 10, and training the model with the selected features, the accuracy of the test set is ensured to be better than that of the unselected full-feature model. This effectively solves the problems of excessive model complexity and low prediction accuracy caused by redundant features in traditional feature selection. The VIF < 10 standard can accurately identify multicollinearity between meteorological and operational features, such as the redundant correlation between wind speed and wind energy density, eliminating invalid redundant information and reducing the computational cost of subsequent model training. The requirement that the test set accuracy is better than that of the full-feature model directly verifies the effectiveness of this selection process. Compared with the prediction deviation caused by the interference of redundant information in the full-feature model, the selected feature set focuses more on the core influencing factors, making the input of the power generation prediction model more accurate. This provides a high-quality feature foundation for subsequent linkage prediction and regulation strategy formulation, avoiding regulatory decision errors caused by feature redundancy or key missing features.

[0070] In summary, in a more specific application of this disclosure, meteorological data collection at ports can employ ultrasonic sensors to calculate wind speed and direction by transmitting or receiving ultrasonic waves based on the time difference; thermistor sensors to collect temperature data; piezoresistive or piezoelectric sensors to collect air pressure data; electrochemical gas sensors to collect salt spray concentration data by detecting Cl⁻ ion concentration; and capacitive or resistive sensors to collect humidity data.

[0071] Sensors are deployed on power generation equipment to collect operational status data. Vibration sensors can collect vibration amplitude, power sensors can calculate real-time power through voltage and current phase differences, speed sensors can collect equipment speed, and temperature sensors can collect equipment temperature. At the same time, the cumulative operating time and design life of the equipment are also collected.

[0072] Port power grids deploy various sensors to collect power generation data. Voltage data can be collected through electromagnetic voltage transformers or fiber optic voltage sensors, while power data can be collected through Hall effect power sensors or Rogowski coil power sensors to calculate load data. Multi-source data acquisition, by integrating different types and dimensions of data sources, can overcome the limitations of single data sources, bringing multiple core advantages to system operation, decision analysis, and value creation.

[0073] This study analyzes meteorological data, operational status data, and corresponding power generation data. Raw meteorological data or derived data, including turbulence intensity and wind energy density, are selected. A clustering algorithm is used to classify wind speed ranges into low, medium, and high wind speed types, while high-density areas are identified as typical meteorological types. Meteorological type labels are extracted for each time point. Operational status parameters are input, and operational status type labels are extracted for each time point using a combination of supervised and unsupervised methods. By clustering wind speed ranges and extracting meteorological and operational status labels, a patterned correlation analysis between meteorological conditions and equipment operating conditions can be achieved. This allows for accurate identification of typical operating scenarios, providing refined data label support for equipment energy efficiency optimization, fault early warning, and scheduling strategy formulation.

[0074] For each meteorological type and operational status type, statistical features are extracted as representative indicators for that type. These statistical features specifically include: basic status features, including mean, peak value, standard deviation, and skewness; physical correlation features, including wind energy density and the duration of the feature; and operational status features, including the matching degree between blade angle and wind direction, and the mean and standard deviation of bearing temperature. By extracting multi-dimensional statistical features, a quantitative characterization of meteorological types and equipment operating conditions can be achieved, providing structured and differentiated data support for equipment energy efficiency analysis, fault mode identification, and operational strategy optimization, thereby improving the accuracy of scenario-based analysis and the scientific basis of decision-making.

[0075] The meteorological type features and operational status features at each time point are aligned with the power generation during the same period to form a dataset. Different measurement methods are selected based on the feature type to calculate the correlation value. The calculation method is as follows: for all features of meteorological type and operational status type, a correlation index is calculated according to the type; features with an index greater than a preset value are retained while redundant features are removed. Redundant features are removed based on the criterion of VIF < 10; the selected features are used to train the power generation prediction model, and the accuracy of the test set is compared to whether it is better than the unselected full-feature model. Through feature correlation analysis and selection, redundant information can be removed while key influencing factors are retained, optimizing the input feature set of the power generation prediction model. This improves model training efficiency, and comparative verification ensures that the selected features significantly improve prediction accuracy and generalization ability.

[0076] The methods for quantifying the correlation are as follows: For meteorological characteristics, the Spearman rank correlation coefficient can be used, which is suitable for nonlinear relationships, to calculate the correlation between the cube of wind speed and power generation; for operational characteristics, mutual information can be used to evaluate the correlation between bearing temperature fluctuations and power generation decline. The mutual information calculation uses a sliding time window with a window size of 20 minutes, which can capture faster and more instantaneous correlations.

[0077] Features with correlation values ​​greater than preset values ​​in each data type were selected as those influencing power generation. The preset values ​​were determined based on historical port data: the Spearman coefficient for meteorological features was preset to ≥0.6, and the mutual information value for operational features was preset to ≥0.5. Supplementary preset values ​​could be adjusted through small-sample testing. The preset value for the salt spray concentration-temperature covariance in high-salt-spray ports was reduced to 0.4. The core features selected were: meteorological dimensions including wind speed cube and wind direction matching duration; equipment dimensions including bearing temperature standard deviation and blade angle deviation cumulative time; and environmental coupling dimensions including salt spray concentration-temperature covariance.

[0078] Based on the selected characteristics, we established a micro-wind power generation prediction model, an equipment condition assessment model, and a power grid load prediction model.

[0079] In one feasible implementation, the output rules and calculation logic of the micro-wind power generation prediction model in the step of establishing the linkage prediction model include: outputting prediction results according to time periods based on the port power consumption scenario; specifically:

[0080] For short-term power consumption adjustment scenarios, it outputs the power generation forecast for the next 15 minutes; for regular power consumption planning scenarios, it outputs the power generation forecast for the next 1 hour.

[0081] The calculation expression for the micro-wind power generation prediction model:

[0082] ;

[0083] Where P is the predicted power, ρ is the air density, v is the wind speed, θ is the wind direction, TI is the turbulence intensity, t is the ambient temperature, h is the cumulative operating time of the equipment, hmax is the design life of the equipment, a is the fitting coefficient, and b is the aging coefficient of the equipment. This is a correction term for wind direction influence. This is a correction term for the influence of turbulence intensity. This is a correction term for the influence of ambient temperature.

[0084] Among them, the turbulence intensity TI is the ratio of the standard deviation of the wind speed to the average wind speed; the fitting coefficient a and the equipment aging coefficient b need to be obtained by using regression algorithms such as support vector machine and neural network to fit historical operating data.

[0085] Therefore, it can be concluded that the calculation expression of the micro-wind power generation prediction model includes a wind direction influence correction term. ), Turbulence intensity influence correction term ( ) and correction items for the influence of ambient temperature represents the fitting coefficient.

[0086] In one feasible implementation, the output content and calculation method of the equipment status assessment model in the step of establishing the linkage prediction model include:

[0087] The output data are the device's health index (HI) and predicted remaining device lifespan (h1).

[0088] The Health Index (HI) is calculated using the following formula:

[0089] ;

[0090] Where w represents the weight of each parameter, including vibration amplitude characteristics, power efficiency characteristics, speed stability characteristics, and temperature characteristics. The weights of each parameter are assigned based on historical fault data through statistical or machine learning algorithms, and the total weight is 1.

[0091] Remaining equipment life ( The calculation formula is:

[0092] ;

[0093] Where h represents the cumulative operating time of the equipment. HI stands for Health Index, representing the design life of the equipment.

[0094] In one feasible implementation, the output format and parameter settings of the power grid load forecasting model in the step of establishing the linkage forecasting model include:

[0095] Output the port power grid load curve for the next 24 hours, and mark the peak and valley periods and corresponding loads in the curve. The peak and valley period division is combined with the peak period data of port container operations.

[0096] The model calculation expression is:

[0097]

[0098] in To predict the load value at time t, This represents the actual load value at the previous moment. This represents the load value at the same time 24 hours ago. To predict the temperature at any given time, For reference temperature;

[0099] Reference temperature T t The annual average temperature is used, combined with historical meteorological data of the port, and dynamically adjusted according to the seasons. The regression coefficient... It was obtained through training with historical load and temperature data.

[0100] In one feasible implementation, the specific method of deviation quantification in the step of generating difference result data includes:

[0101] The absolute error formula |actual value - predicted value| is used to calculate the numerical deviation, reflecting the absolute difference between the actual data and the predicted data.

[0102] The deviation percentage is calculated using the relative error formula (|actual value - predicted value| / predicted value) × 100%, reflecting the relative severity of the deviation.

[0103] The results are quantified using both absolute and relative errors to generate difference data that includes the deviation value, deviation percentage, and deviation duration.

[0104] For example, current power generation, equipment status, and grid load are obtained and compared with predicted data to generate discrepancy results. The discrepancy results can be quantified using formulas for absolute error and relative error to determine the deviation between predicted and actual values.

[0105] The formulated intelligent control strategy refers to: setting the priority of control targets, setting intelligent control strategies according to levels and scenarios, and generating control instructions when specific scenarios include power generation being lower than the predicted value, grid load being higher than power generation, and equipment health rapidly declining.

[0106] The primary objective is to ensure grid stability, such as keeping grid frequency fluctuations stable (±0.2Hz under normal operating conditions and ±0.5Hz under extreme operating conditions) and voltage fluctuations within rated values ​​(±5% of rated voltage under normal operating conditions and ±10% of rated voltage under emergency operating conditions).

[0107] The secondary objective focuses on equipment reliability, requiring that the health index of key equipment does not exceed the set value, such as HI≥0.8 for normal operation, HI≥0.6 for triggering an early warning, and HI<0.6 for shutdown and maintenance; the predicted remaining lifespan is not less than the safety threshold, such as ≥3 months of safe operation, and ≤1 month for mandatory replacement;

[0108] The three-tiered target aims to improve power generation efficiency and control power generation deviation; a deviation of ≤5% is considered normal, while deviations >5% require optimization and adjustment. Priorities are dynamically adjusted based on the port's real-time operational status to achieve rational allocation and utilization of resources.

[0109] The scenario where power generation is lower than the predicted value is triggered when the power generation deviation exceeds 5% for 30 consecutive minutes, and the wind speed is between the cut-in wind speed and the rated wind speed. If the wind is light, the blade angle is adjusted first to improve the efficiency of capturing light winds, and the blade electric heating system is activated to prevent icing. If the wind speed is normal but turbulence is present, the equipment damping control is increased, and the yaw system response speed is reduced to minimize mechanical losses. If the equipment health is low, the generator load rate is reduced simultaneously to avoid over-operation of the equipment.

[0110] The scenario where the grid load exceeds the power generation is as follows: When the error in the power load forecast exceeds the set value (e.g., >15%), the corresponding response strategy for this scenario is activated. First, port power equipment is prioritized, with priority given to ensuring power supply to core operational equipment in the container crane and refrigerated container area. Power rationing is tiered based on the size of the load gap: a gap <10% is considered minor, and 10%-30% is considered major. For minor gaps, port lighting brightness is reduced, and non-essential ventilation systems are suspended. For major gaps, temperature setpoints in the refrigerated container area are adjusted, and charging of automated guided vehicles is suspended. In extreme cases, diesel generators are started and external grid support is requested, while port management is notified to coordinate the shutdown of non-critical operations.

[0111] Scenarios indicating a rapid decline in equipment health include: when the rate of decline in the equipment health index exceeds a set value (e.g., >0.1 / hour), or when the vibration amplitude exceeds a warning threshold (e.g., 50% above the baseline value), targeted strategies are implemented. The fault type is determined based on fault characteristics: if the vibration frequency exhibits a first-harmonic anomaly, check the coupling alignment and increase bearing lubrication frequency; if it is a second-harmonic fault, focus on investigating gearbox tooth surface wear and monitor the metal particle content in the lubricating oil. Based on the equipment's remaining life assessment results, if the remaining life is less than 3 days, immediately shut down for maintenance and replace with backup equipment; if the remaining life is less than 7 days, reduce the equipment load to 50% and strengthen online monitoring; otherwise, adjust equipment operating parameters and arrange preventative maintenance.

[0112] Reference Figure 2 As shown, after the command is executed, the system will collect new power generation, equipment status, and grid load data within 5 minutes and verify the effect against a multi-dimensional verification index system. If the key indicators fail to reach the qualified threshold, secondary regulation will be triggered.

[0113] The second round of regulation will formulate targeted strategies based on the specific indicators that were not met in the first round of regulation. If the power generation recovery is insufficient, the blade angle will be further adjusted, such as from 15° to 10° under light wind conditions, and the system will be switched to maximum power point tracking mode. If the equipment health continues to deteriorate, the equipment load will be reduced again, from 100% to 70%, and the equipment maintenance priority will be raised to emergency. If the grid frequency or voltage stability does not meet the standards, the discharge power of the energy storage system will be increased to improve the demand-side response level.

[0114] The secondary control command is also issued and executed after parameter verification and protocol encapsulation, with a 10-minute extended observation period. If the target is still not met after the secondary control, a three-level progressive control will be initiated: If the first failure occurs, the backup control logic will be activated, an extreme weather adaptation algorithm will be adopted, short-term turbulence fluctuations will be ignored, and equipment stability will be the primary objective. The parameter adjustment range will be expanded, and the blade angle adjustment range will be expanded to 1.5 times that of the first time during the secondary control. When multiple devices are coordinated, units closer to high wind speed areas will be prioritized, and the load will be allocated according to the current health status ratio, with units with HI≥0.8 bearing 60% of the load. If the second failure occurs, the multi-device coordinated control mechanism will be activated, and external power grid support will be requested. If multiple failures occur consecutively, an emergency response will be triggered, the operation of some non-critical equipment will be suspended, and an expert team will be notified to conduct remote diagnosis.

[0115] After each adjustment, the system optimizes the adjustment strategy based on the difference between the actual effect and the target. It analyzes the relationship between parameter adjustments and effect feedback during the adjustment process, summarizes effective adjustment experience, optimizes the control parameter range and priority rules, and forms a closed-loop optimization mechanism. Through continuous accumulation and learning, the accuracy and effectiveness of the adjustment strategy are improved, gradually enhancing the overall operational efficiency and stability of the port micro-wind power generation system.

[0116] In one specific embodiment, the air density is 1.225 kg / m³, the wind speed is 5 m / s, the wind direction is 30°, the turbulence intensity is 10%, the ambient temperature is 28°C, the cumulative operating time of the equipment is 10,000 hours, the design life of the equipment is 100,000 hours, the fitting coefficient is 1, the equipment aging coefficient is 1, and the predicted power is 2447 kW.

[0117] The vibration amplitude w1 is 0.4, the average bearing temperature w2 is 0.3, the standard deviation of temperature w3 is 0.2, and the speed fluctuation w4 is 0.1. The vibration amplitude is 0.8 mm / s, the average bearing temperature is 65℃, the standard deviation is 2.5℃, the speed fluctuation is 0.5%, the health index is 0.82, and the predicted remaining life is 73,800 hours.

[0118] The load at the previous moment was 5.2MW, and the same load 24 hours ago was 4.8MW. Model parameters The predicted load value at the time of prediction is 6.32MW.

[0119] The current power is 2600kW, the remaining lifespan is 90,000 hours, and the load value is 6MW.

[0120] By comparing the predicted power, predicted remaining lifespan, and predicted load values ​​with the data obtained at the current time, the deviation is obtained, and it is determined that the power generation is lower than the predicted value, the grid load is higher than the power generation, and the equipment health is declining rapidly. Control instructions are generated according to different scenarios.

[0121] This disclosure also provides a port micro-wind power generation control system that executes the aforementioned intelligent control method. The system includes:

[0122] The multi-source data directional acquisition unit includes at least an ultrasonic wind speed and direction sensor, an electrochemical salt spray sensor, a piezoelectric vibration sensor, an electromagnetic voltage transformer, and a Rogowski coil power sensor, used to acquire port meteorological data, equipment operating status data, and power grid generation data.

[0123] The feature filtering and analysis unit is used to generate classification labels, extract statistical features and calculate the correlation of collected multi-source data, filter key influencing features and remove redundant data.

[0124] The three-model linkage prediction unit integrates a micro-wind power generation prediction model, an equipment status assessment model, and a grid load prediction model to output predicted data on power generation, equipment status, and grid load.

[0125] The difference comparison unit is used to acquire the current actual operating data, compare it with the predicted data, quantify the deviation, and generate difference result data;

[0126] The strategy formulation and execution unit is used to set the priority of the control scenario, generate intelligent control instructions, and send them to the field controller via Modbus or TCP protocol. At the same time, it collects data after control to verify the effect and triggers secondary control.

[0127] In this way, all the effects of the aforementioned intelligent control methods can be achieved, which will not be elaborated here.

[0128] Through the above description of the embodiments, those skilled in the art can clearly understand that the various embodiments of this application can be implemented by means of software or software combined with necessary general-purpose hardware platforms, and of course, they can also be implemented by hardware functions; based on this understanding, the technical solution of this application, in essence or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions to cause a computer device, such as including but not limited to a personal computer, server, or network device, to execute all or part of the steps of any embodiment of the method of this application.

[0129] The foregoing describes exemplary embodiments of this application. It should be understood that the above exemplary embodiments are not restrictive but illustrative, and the scope of protection of this application is not limited thereto. It should be understood that those skilled in the art can make modifications and variations to the embodiments of this application without departing from the spirit and scope of this application, and such modifications and variations should be within the scope of protection of this application.

Claims

1. A port micro-wind power generation intelligent regulation method based on multi-source data, applied to a port micro-wind power generation regulation system, characterized in that, The control system includes a multi-source data targeted acquisition unit, a feature screening and analysis unit, a three-model linkage prediction unit, a difference comparison unit, and a strategy formulation and execution unit. The method includes the following steps: To acquire multi-source data from the port, sensors are deployed in the port area to collect meteorological data, sensors are deployed in the micro-wind power generation equipment to collect operational status data, and sensors are deployed in the port power grid to collect power generation data through multi-source data directional acquisition units. Key influencing features are screened. A feature screening and analysis unit performs correlation analysis on the meteorological data, operational status data, and power generation data. First, the data is classified and type labels are generated. Then, statistical features are extracted and correlation values ​​are calculated to screen out meteorological and operational status features affecting port light wind power generation. The correlation value is calculated using the Spearman rank correlation coefficient for meteorological features and mutual information for operational status features. Features with correlation values ​​greater than preset values ​​are retained: the preset value for the Spearman coefficient of meteorological features is ≥0.6, and the preset value for the mutual information value of operational status features is ≥0.

5. If the port has a high salt spray environment, the preset value for the salt spray concentration-temperature covariance is reduced to 0.4 and used as a supplementary feature in the screening. Redundant features are removed using a variance inflation factor <10. The screened features are used to train a power generation prediction model, ensuring that the accuracy of the test set is better than the unscreened full-feature model. A linkage prediction model is established. Based on the selected features, a micro-wind power generation prediction model, an equipment status assessment model, and a power grid load prediction model are established through a three-model linkage prediction unit, and the corresponding prediction data is output. The micro-wind power generation prediction model is used to predict power based on air density, wind speed, wind direction, turbulence intensity, ambient temperature, cumulative equipment operating time and design life, and is corrected by fitting coefficients and equipment aging coefficients. The fitting coefficients include constant terms, wind direction influence correction terms, turbulence intensity influence correction terms and ambient temperature influence correction terms. Turbulence intensity is the ratio of the standard deviation of wind speed to the average wind speed. The fitting coefficients and equipment aging coefficients are obtained by fitting historical operating data using regression algorithms such as support vector machines or neural networks. The equipment condition assessment model outputs the equipment's health index and predicted remaining lifespan. The health index is calculated based on the weighted product of vibration amplitude, power efficiency, rotational speed stability, and temperature characteristics. The remaining lifespan is determined based on the equipment's cumulative operating time, design life, and health index. The weights of each parameter are allocated based on historical fault data through statistical or machine learning algorithms, and the total weight is 1. The power grid load prediction model outputs the port power grid load curve for the next 24 hours, and marks the peak and valley periods and corresponding loads in the curve. The peak and valley periods are divided by combining the port container operation peak period data. The load value is determined by regression coefficients based on historical load data, temperature data and reference temperature. The reference temperature is taken as the annual average temperature and dynamically adjusted according to the season by combining the port's historical meteorological data. The regression coefficients are obtained by training with historical load and temperature data. The difference result data is generated by comparing the actual power generation, equipment status, and grid load with the predicted data point by point and quantifying the deviation. A control strategy is formulated and implemented. Through the strategy formulation and execution unit, based on the differential result data and current meteorological data, the priority of control scenarios is set, intelligent control instructions are generated and issued for execution, and new data is collected after execution to verify the effect. If the target is not met, secondary control is triggered. The control target priority is set as follows: Level 1 target: frequency fluctuation ±0.2Hz under normal operating conditions, ±0.5Hz under extreme operating conditions; Level 2 target: HI≥0.8 for normal operation, HI≥0.6 for triggering an early warning, HI<0.6 for shutdown and maintenance; Level 3 target: power generation deviation ≤5%.

2. The method of claim 1, wherein, In the step of acquiring multi-source port data: The meteorological data includes wind speed, wind direction, temperature, air pressure, and salt spray concentration. The salt spray concentration is collected by an electrochemical gas sensor by detecting the concentration of Cl⁻ ions, while the wind speed and wind direction are collected by an ultrasonic sensor by calculating the time difference between transmitting or receiving ultrasonic waves. The operating status data includes generator speed, output power, equipment temperature, vibration amplitude, and cumulative equipment operating time. The vibration amplitude is collected using a piezoelectric vibration sensor, and the output power is calculated by detecting the phase difference between voltage and current using a power sensor. The power generation data includes voltage, power, and load. The voltage is collected using an electromagnetic voltage transformer, the power is collected using a Rogowski coil power sensor, and the load is calculated based on the voltage and power data.

3. The method of claim 1, wherein the step of modulating comprises: In the step of screening key influencing features, the methods for generating type labels for data classification include: The meteorological data is processed by selecting raw meteorological data or derived data. The derived data includes turbulence intensity and wind energy density. The data is divided into low wind speed, medium wind speed and high wind speed types according to wind speed range using a clustering algorithm. At the same time, high-density areas of data are identified as typical meteorological types, and meteorological type labels are extracted for each time point. The processing of the operating status data involves inputting operating status parameters such as generator speed and equipment temperature, and using a combination of supervised and unsupervised modes to classify the equipment operating conditions and extract the operating status type label for each time point.

4. The method of claim 3, wherein the step of regulating comprises, In the step of screening key influencing features, the extracted statistical features include: basic state features, physical related features, and operational state features; The basic state characteristics include the mean, peak value, standard deviation, and skewness of the data, which are used to reflect the basic distribution characteristics of the data; The physical characteristics include wind energy density and the duration of the characteristics, wherein the wind energy density is calculated based on wind speed and air density; The operational status characteristics include the matching degree between blade angle and wind direction, the average bearing temperature, and the standard deviation of bearing temperature, which are used to correlate the relationship between equipment operation and power generation efficiency.

5. The method of claim 1, wherein the step of modulating comprises: In the step of establishing the linkage prediction model, the output rules and calculation logic of the micro-wind power generation prediction model include: The forecast results are output according to the time period of the port's electricity consumption scenario; specifically: For short-term power consumption adjustment scenarios, it outputs the power generation forecast for the next 15 minutes; for regular power consumption planning scenarios, it outputs the power generation forecast for the next 1 hour.

6. The method of claim 1, wherein the step of modulating comprises: In the step of generating difference result data, the specific methods of deviation quantification include: The absolute error formula |actual value - predicted value| is used to calculate the numerical deviation, reflecting the absolute difference between the actual data and the predicted data. The deviation percentage is calculated using the relative error formula (|actual value - predicted value| / predicted value) × 100%, reflecting the relative severity of the deviation. The results are quantified using both absolute and relative errors to generate difference data that includes the deviation value, deviation percentage, and deviation duration.

7. A port micro-wind power generation control system, characterized in that, The system implementing the port micro-wind power generation intelligent control method based on multi-source data as described in any one of claims 1-6, the system comprising: The multi-source data directional acquisition unit includes at least an ultrasonic wind speed and direction sensor, an electrochemical salt spray sensor, a piezoelectric vibration sensor, an electromagnetic voltage transformer, and a Rogowski coil power sensor, used to acquire port meteorological data, equipment operating status data, and power grid generation data. The feature selection and analysis unit is used to generate classification labels, extract statistical features, and calculate correlations from the collected multi-source data. It selects key influencing features and removes redundant data. The correlation value is calculated using the Spearman rank correlation coefficient for meteorological features and mutual information for operational status features. Features with correlation values ​​greater than preset values ​​are retained: the preset Spearman coefficient for meteorological features is ≥0.6, and the preset mutual information value for operational status features is ≥0.

5. If the port is in a high-salt-fog environment, the preset value of the salt-fog concentration-temperature covariance is reduced to 0.4 and used as a supplementary feature for selection. Redundant features are removed based on a variance inflation factor <10. The selected features are used to train the power generation prediction model, ensuring that the accuracy of the test set is better than the unselected full-feature model. The three-model linkage prediction unit integrates a micro-wind power generation prediction model, an equipment status assessment model, and a grid load prediction model to output predicted data on power generation, equipment status, and grid load. The micro-wind power generation prediction model is used to predict power based on air density, wind speed, wind direction, turbulence intensity, ambient temperature, cumulative equipment operating time and design life, and is corrected by fitting coefficients and equipment aging coefficients. The fitting coefficients include constant terms, wind direction influence correction terms, turbulence intensity influence correction terms and ambient temperature influence correction terms. Turbulence intensity is the ratio of the standard deviation of wind speed to the average wind speed. The fitting coefficients and equipment aging coefficients are obtained by fitting historical operating data using regression algorithms such as support vector machines or neural networks. The equipment condition assessment model outputs the equipment's health index and predicted remaining lifespan. The health index is calculated based on the weighted product of vibration amplitude, power efficiency, rotational speed stability, and temperature characteristics. The remaining lifespan is determined based on the equipment's cumulative operating time, design life, and health index. The weights of each parameter are allocated based on historical fault data through statistical or machine learning algorithms, and the total weight is 1. The power grid load prediction model outputs the port power grid load curve for the next 24 hours, and marks the peak and valley periods and corresponding loads in the curve. The peak and valley periods are divided by combining the port container operation peak period data. The load value is determined by regression coefficients based on historical load data, temperature data and reference temperature. The reference temperature is taken as the annual average temperature and dynamically adjusted according to the season by combining the port's historical meteorological data. The regression coefficients are obtained by training with historical load and temperature data. The difference comparison unit is used to acquire the current actual operating data, compare it with the predicted data, quantify the deviation, and generate difference result data; The strategy formulation and execution unit is used to set the priority of control scenarios, generate intelligent control instructions, and send them to the field controller via Modbus or TCP protocol. At the same time, it collects data after control to verify the effect and trigger secondary control. Among them, the priority of control targets is set, including: Level 1 target: frequency fluctuation ±0.2Hz under normal operating conditions and ±0.5Hz under extreme operating conditions; Level 2 target: HI≥0.8 for normal operation, HI≥0.6 to trigger an early warning, and HI<0.6 to perform shutdown and maintenance; Level 3 target: power generation deviation ≤5%.

Images