Active harmonic suppression and energy management method and system for a ship power grid

Abstract

The application relates to the technical field of harmonic suppression, and specifically discloses a ship power grid harmonic active suppression and energy management method and system, which comprises the following steps: collecting basic electric parameters of a ship power grid, energy equipment operation parameters and transient characteristics of nonlinear loads; based on the collected multidimensional data, a deep learning hybrid model containing an attention mechanism is used to predict the harmonic trend and energy demand of the ship power grid, and the model parameters are updated online and iteratively according to the prediction error. The application realizes the transformation from passive compensation to forward intervention by deeply coupling the harmonic active suppression and energy management; based on the accurate prediction of the harmonic trend and energy demand of the power grid, the system can dynamically coordinate the strategy of harmonic control and energy distribution, significantly improves the overall energy efficiency and operation economy of the ship power grid under the premise of guaranteeing that the power quality strictly meets the standard.

Description

Technical Field

[0001] This invention relates to the field of harmonic suppression technology, and in particular to a method and system for active harmonic suppression and energy management of ship power grids. Background Technology

[0002] Currently, active power filters (APFs) and other devices are commonly used in shipboard power grids for harmonic mitigation, and independent energy management systems are in place. However, existing technologies often treat harmonic suppression and energy management as two relatively independent issues, resulting in limited overall system efficiency. On the one hand, common harmonic suppression strategies often rely on passive compensation for existing harmonics, lacking the ability to actively predict harmonic trends under complex ship operating conditions. This leads to system response lag, making it difficult to effectively intervene at the source of harmonic generation, resulting in insufficient mitigation accuracy and real-time performance. On the other hand, traditional energy dispatch models typically only focus on the power balance between generation and load, rarely incorporating the energy consumption of APF devices themselves (such as IGBT switching losses) into the overall energy efficiency optimization objective. This results in additional energy losses, contradicting the urgent need for high-efficiency operation on ships. This disconnected architecture makes it difficult for the system to maximize energy efficiency while ensuring power quality.

[0003] Furthermore, to improve adaptability, some advanced systems attempt to introduce predictive models with self-learning capabilities or dynamically adjust control parameters based on real-time data. However, in complex scenarios such as shipboard electrical networks where data is easily affected by environmental interference and operating conditions frequently change, the inherent contradictions of this highly responsive adaptive mechanism are amplified. This manifests in the following ways:

[0004] When a system attempts to sensitively follow load transients to optimize harmonic suppression, overreaction to measurement data jitter or transient communication delays can easily cause high-frequency oscillations in control commands across multiple strategies, such as frequent switching of compensation bandwidth or repeated corrections of power allocation commands. This instability not only exacerbates losses in critical power electronic components, such as IGBTs in the APF, but may even trigger local oscillations, threatening the stable operation of the power grid, thus creating a design dilemma where pursuing accurate prediction introduces control instability. Summary of the Invention

[0005] This invention aims to at least partially solve one of the technical problems in related technologies. Therefore, the objective of this invention is to propose a method and system for active harmonic suppression and energy management of ship electrical systems, in order to improve the robustness and safety of ship electrical systems under various complex operating conditions.

[0006] To achieve the above objectives, a first aspect of the present invention provides a method for active harmonic suppression and energy management in ship electrical networks, comprising the following steps:

[0007] S1. Collect basic electrical parameters of the ship's power grid, operating parameters of energy equipment, and transient characteristics of nonlinear loads;

[0008] S2. Based on the collected multi-dimensional data, a deep learning hybrid model including an attention mechanism is used to predict the harmonic trend and energy demand of the ship's power grid, and the model parameters are updated online iteratively according to the prediction error.

[0009] S3. Monitor the device-level energy consumption of the active power filter (APF) in real time, combine the predicted harmonic trends and energy demands, calculate the harmonic-energy dynamic priority coefficient, and use the model predictive control algorithm based on the coefficient to plan the power allocation of each energy device.

[0010] S4. Based on the power allocation results, coordinate the distributed master-slave APFs to suppress harmonics and execute local pre-control strategies when communication delays are detected.

[0011] To achieve the above objectives, a second aspect of the present invention provides a shipboard electrical grid harmonic active suppression and energy management system, comprising:

[0012] The data acquisition module is configured to perform multi-dimensional data acquisition, collecting basic electrical parameters of the ship's electrical network, operating parameters of energy equipment, and transient characteristics of nonlinear loads; the transient characteristics of the nonlinear loads include at least the rate of change of the inverter's modulation frequency. and the derivative of the traction torque of the electric propulsion device ;

[0013] The adaptive prediction module, communicatively connected to the data acquisition module, is configured to predict the harmonic trends and energy demands of the ship's power grid based on the acquired multi-dimensional data using a deep learning hybrid model incorporating an attention mechanism, and to iteratively update the model parameters online according to the prediction error; wherein, the attention mechanism employs a correlation weighting function. Weighting of transient characteristics of nonlinear loads;

[0014] The coupled energy scheduling module, communicatively connected to the operating condition adaptive prediction module, is configured to monitor the device-level energy consumption of the active power filter (APF) in real time and calculate the harmonic-energy dynamic priority coefficient based on the energy demand prediction results. Based on this coefficient, a model predictive control algorithm is used to plan the power allocation of each energy device; wherein, the device-level energy consumption of the APF includes IGBT switching losses. and reactor copper loss ;

[0015] The collaborative harmonic suppression module, which is communicatively connected to the coupled energy dispatch module, is configured to coordinate the distributed master-slave APFs to suppress harmonics based on the power allocation results, and to execute a local pre-control strategy when a communication delay is detected. The system also includes a harmonic feature library, which stores typical harmonic sequence sets of inverters with different modulation methods and topologies under different load rates, for the collaborative harmonic suppression module to dynamically query and perform precise compensation for high-energy harmonic sequence sets based on the query results.

[0016] To achieve the above objectives, a third aspect of the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory. When the computer program is executed by the processor, it implements the above-described method for active harmonic suppression and energy management of ship power grids.

[0017] The active harmonic suppression and energy management method and system for ship power grids of this invention realizes the transformation from passive compensation to proactive intervention by deeply coupling active harmonic suppression and energy management. The system can dynamically coordinate harmonic control and energy allocation strategies based on accurate prediction of power grid harmonic trends and energy demand, significantly improving the overall energy efficiency and operational economy of the ship power grid while ensuring strict compliance with power quality standards. At the same time, by introducing an adaptive mechanism with anti-interference capabilities and a local intelligent decision-making strategy, the system can smoothly transition and maintain a stable and reliable suppression effect when facing data noise or communication anomalies, thereby ensuring the robustness and safety of the ship power system under various complex operating conditions. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the active harmonic suppression and energy management method for ship electrical grids provided by the present invention.

[0019] Figure 2 This is a schematic diagram of multi-dimensional data waveforms in the active harmonic suppression and energy management method for ship power grids provided by the present invention;

[0020] Figure 3 This is a schematic diagram of the deep learning hybrid model in the active harmonic suppression and energy management method for ship power grids provided by this invention;

[0021] Figure 4 This is a schematic diagram of the calculation of harmonic-energy dynamic priority coefficients in the active harmonic suppression and energy management method for ship power grids provided by the present invention;

[0022] Figure 5 This is a schematic diagram comparing the power distribution before and after MPC optimization in the active harmonic suppression and energy management method for ship power grids provided by this invention;

[0023] Figure 6This is a schematic diagram of the division of labor for harmonic compensation in the active harmonic suppression and energy management method for ship power grids provided by the present invention;

[0024] Figure 7 This is a schematic diagram of the dynamic response of the two-stage compensation strategy switching in the active harmonic suppression and energy management method for ship power grids provided by the present invention;

[0025] Figure 8 This is a schematic diagram illustrating the implementation of the ship electrical network harmonic active suppression and energy management system provided by the present invention;

[0026] Figure 9 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0027] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0028] The following description, with reference to the accompanying drawings, describes a method, system, and electronic device for active harmonic suppression and energy management of marine power grids according to embodiments of the present invention.

[0029] Example 1:

[0030] The present invention provides a method for active harmonic suppression and energy management of ship power grids. This method aims to solve complex problems in modern ship power systems, such as power quality deterioration, conflict between harmonics and energy management, and reduced system operating economy caused by large-scale nonlinear loads and the access of new energy sources.

[0031] The method described in this embodiment is particularly suitable for advanced vessels with hybrid power grids of medium-voltage DC MVDC or low-voltage AC LVAC, such as electric propulsion vessels, offshore engineering vessels, luxury cruise ships, or special-purpose vessels. These shipboard power grids typically exhibit the following characteristics: diverse power sources, including but not limited to diesel generator sets, gas turbine generator sets, high-capacity lithium-ion battery energy storage units (ESS), and possibly photovoltaic or fuel cells; a high proportion of power electronic converters on the load side, particularly megawatt-class frequency converters used for main propulsion and side thrust, which are major sources of nonlinear harmonics; and, simultaneously, stringent requirements for both operational economy and power quality, which are often mutually constraining.

[0032] This method is applied to the central energy management system (EMS) of a ship's electrical system or a dedicated ship power quality and energy management controller. This system or controller is connected to various subsystems on board via a high-speed shipboard local area network, such as industrial Ethernet, CAN-FD, or a DNV-certified marine network. For example... Figure 1 As shown below, the specific steps of the method in this embodiment will be described in detail. The method includes:

[0033] The first step is to perform multi-dimensional data collection.

[0034] The purpose of this step is to obtain a complete and real-time synchronized snapshot of the ship's electrical network across the three dimensions of electrical, energy, and control. The data collected in this step can be specifically divided into three categories:

[0035] Category 1 data: Basic electrical parameters of the ship's electrical network.

[0036] These data are used to characterize the macroscopic power quality and power flow status of the power grid. Data acquisition points are typically located at the busbars of the main switchboard, the outlets of each generator, the outlets of the energy storage converter (PCS), the point of common coupling (PCC), and major nonlinear loads, such as the inlet of the drive inverter. Acquisition equipment can include high-precision power quality analyzers, smart meters, or remote terminal units (RTUs) integrating a high-speed data acquisition (DAQ) module.

[0037] To accurately analyze harmonics, the sampling rate should be much higher than the Nyquist frequency. For example, for harmonic analysis below 400 Hz, the sampling rate is preferably 20 kHz or higher.

[0038] The basic electrical parameters specifically include: the effective value (RMS) of the bus voltage, the effective value of the current, the grid frequency, the three-phase imbalance, active power (P), reactive power (Q), apparent power (S), power factor (PF), and the content of each harmonic. Total Harmonic Distortion (THD).

[0039] The second type of data: operating parameters of energy equipment.

[0040] These data are used to characterize the state and capability of a ship's energy sources. For diesel generator sets, the parameters collected include: their output active power, speed, fuel consumption rate, operating temperature, excitation voltage, etc.

[0041] For an energy storage unit (ESS), the collected parameters are usually from the battery management system (BMS), including: the actual state of charge (SOC) of the energy storage unit, the terminal voltage of the battery cluster, the charging and discharging current, the highest and lowest temperatures of the cells, and the state of health (SOH), etc.

[0042] If the system also includes new energy sources such as photovoltaics, then it is also necessary to collect its DC side irradiance and AC side output power.

[0043] The third type of data: transient characteristics of nonlinear loads.

[0044] This is the key data for achieving forward-looking prediction in this embodiment. Traditional electrical parameters, such as P, Q, and THD, are all lagging indicators, meaning that by the time they change, harmonic impacts have already occurred. Transient characteristics, on the other hand, are leading indicators or causal indicators that characterize impending load changes; they directly reflect the dynamic commands of the internal control system of the nonlinear load.

[0045] In this embodiment, these transient features include at least the following two:

[0046] 1. The derivative of the traction torque of an electric propulsion device, its symbol can be expressed as: .

[0047] Traction torque derivative of electric propulsion device Defined as the difference between the current traction torque of the electric propulsion device and the traction slip in the previous sampling period, divided by the sampling period.

[0048] In practice, the traction torque T is not measured directly from the propeller shaft, which can be delayed and costly. Instead, it is preferably read directly from the ship's joystick or the control command bus of the dynamic positioning DP system.

[0049] For example, when a crew member pushes the control stick from 30% thrust to 80% thrust, the rate of change of this thrust command is captured by the system as a high amplitude. The physical significance of this feature lies in the fact that it represents the ship's operational intention. A high Values ​​such as those used for emergency evasive maneuvers or acceleration during departure are the strongest indication of impending severe power demand and strong harmonic shocks.

[0050] 2. The rate of change of the frequency modulation of the inverter, which can be represented by the symbol: .

[0051] Frequency conversion rate of inverter modulation Defined as the difference between the current modulation frequency of the inverter and the modulation frequency of the previous sampling period, divided by the sampling period.

[0052] In practical implementation, the modulation frequency f, such as the carrier frequency of SPWM or the switching frequency of SVPWM, is a control loop parameter inside the frequency converter. This data needs to be read from the main controller of the frequency converter via a high-speed communication interface. The physical meaning of this characteristic is that it represents the control dynamics of the frequency converter. When the frequency converter responds to drastic changes in torque, i.e. When forced to rapidly adjust its internal control strategy, its modulation frequency will change instantaneously. The dramatic increase directly indicates that the composition of the harmonic spectrum, especially the high-frequency harmonics, is about to undergo drastic changes.

[0053] All of this multi-dimensional data is synchronized with time protocols such as PTP or NTP to ensure timestamp consistency, and then sent to a central management controller to provide input for the predictive model in the second step.

[0054] like Figure 2 The diagram illustrates the collected multidimensional data waveforms. The three waveforms represent different types of electrical data collected from the ship's electrical network: voltage waveform, current waveform, and frequency variation. Through real-time acquisition and processing of this data, the system can monitor the operating status of the power grid and predict future harmonic trends and energy demands.

[0055] Voltage waveform (blue curve): This curve shows the variation in grid voltage, which includes the fundamental frequency (50Hz) and higher harmonic components (such as 150Hz) generated by nonlinear loads. Such high-frequency harmonics are generated in the voltage waveform due to rapid frequency changes caused by the control of frequency converters and electric propulsion systems. This graph shows that the grid voltage is not only the fundamental frequency but also contains additional higher harmonics, reflecting power quality problems caused by nonlinear loads in the ship's power grid.

[0056] Current waveform (green curve): The current waveform shows the variation in grid current, corresponding to the basic electrical energy transfer in the system. This waveform is an ideal sine wave, reflecting the normal operating state of the load. By acquiring this data, the system can monitor the load's power demand and for any additional current fluctuations, which is crucial for further harmonic suppression and energy management.

[0057] Frequency variation (red curve): Frequency fluctuations reflect changes in the grid frequency caused by nonlinear loads in the ship's electrical network. These fluctuations are particularly pronounced during transient load changes, such as acceleration or torque regulation. This fluctuation is monitored in real-time using the transient characteristics of the nonlinear loads, and this data is fed back into the model for further prediction and control.

[0058] Through this high-precision data acquisition and real-time processing, the system can implement proactive power quality management in the ship's power grid, ensuring the stable operation of the grid and effectively reducing unnecessary energy losses.

[0059] The second step is to perform the adaptive prediction step based on the operating condition.

[0060] This step receives the multi-dimensional data collected in the first step, and its core task is to build upon the previous step: predicting future trends from the current state. This step needs to predict two objectives simultaneously: future harmonic trends and future energy demands.

[0061] This embodiment uses a deep learning hybrid model that includes an attention mechanism to perform this prediction task. The reason for using a hybrid model is that the input data is multimodal.

[0062] For example, it can preferably be composed of a convolutional neural network (CNN) and a long short-term memory network (LSTM) connected in series or in parallel: CNN layer: used to process raw data with spatial features, for example, the high-frequency voltage and current waveforms collected are regarded as one-dimensional images, and the CNN can automatically extract the morphological features of waveform distortion (i.e. harmonics).

[0063] LSTM layers: Used to process data with time-series characteristics. Examples include changes in SOC, fuel consumption, and load power P. and These are all sequences with strong temporal correlation. The memory gating mechanism of LSTM makes it adept at capturing long-term dependencies in these sequences.

[0064] This hybrid model will incorporate all data from the first step (V, I, P, SOC, T, ... , These components (such as harmonics and energy demand) are combined to form a high-dimensional input vector, which is then input into the model. The model then outputs two key predictions: harmonic trends and energy demand.

[0065] Regarding the attention mechanism: This is key to achieving adaptation in this embodiment. In ship operation condition A, such as stable cruising, the main factors affecting future harmonics and energy consumption are likely SOC and fuel rate; while in condition B, such as emergency acceleration, Clearly, this is the most important feature, overriding everything else. The introduction of the attention mechanism allows the model to learn to dynamically assign different levels of attention or weights to different input features under different conditions without human intervention.

[0066] Optionally, the attention mechanism in this embodiment employs a correlation weighting function. To weight the transient characteristics of nonlinear loads, the correlation weighting function is used. The calculation formula can be expressed as:

[0067]

[0068] in, : Represents the first The final association weights for transient characteristics of nonlinear loads, for example, It can be considered as Category 1. This can be considered as the second type. This weight is dynamically changed; it will be used to amplify or reduce the influence of the transient feature when it enters the next layer of the neural network.

[0069] : The number of categories representing the transient characteristics of nonlinear loads, considering only the weights. At least 2;

[0070] : Represents the first The absolute value of the current change in a transient-like feature. For example, if The current value is +50 Nm / s, then It's 50. The physical meaning of this value is how drastic the change is.

[0071] : Represents the first The contribution coefficient of transient-like features to harmonics. This coefficient is not a physical constant, but a parameter learned through model training; its physical meaning is how significant this type of change is. For example, historical data shows... The change in [a certain factor] has a very strong causal relationship with the total THD, therefore its corresponding [factor / condition]... The value will be very large; and if The changes have a relatively small impact. The value will be smaller.

[0072] Based on the calculations using the above formula, it can be seen that a feature Influence It depends on the drasticness of its change at this moment. It also depends on the importance it has proven to be in history. This weighting method ensures that the model captures the key points. For example, when When a surge occurs, its It will spike rapidly, causing the predictive model to immediately... The data is the primary basis for updating its [relationship / policy]. The prediction.

[0073] It is also important to note that ship equipment ages, and mechanical characteristics can drift. An offline-trained model may become ineffective after a year of operation. Therefore, this step also includes an online iterative update mechanism.

[0074] Specifically, the system will continuously compare the predicted harmonic trends. and the actual values ​​collected later Calculate the prediction error This error This feedback is used to adjust the predictive model as a penalty or reward signal, through online learning techniques such as backpropagation or Kalman filters, making small, continuous adjustments to the model's internal parameters. This gives the model adaptive and self-learning capabilities, ensuring its long-term predictive accuracy.

[0075] like Figure 3 The diagram illustrates the structure of a deep learning hybrid model, demonstrating how the entire system processes data from the ship's electrical grid and makes decisions using a deep learning hybrid model and the MPC algorithm to optimize harmonic suppression and energy management of the grid.

[0076] Data Input Module (Blue Box): The first module in the diagram represents multi-dimensional data input from the ship's electrical network, including voltage, current, and frequency. These data represent the basic electrical parameters of the power grid. By monitoring this data in real time, the system can understand the operating status of the power grid, especially how it is affected by nonlinear loads, thus providing raw data for further processing.

[0077] Data Preprocessing Module (Green Box): The second module represents the data preprocessing stage. In this stage, the input raw data is cleaned and processed to remove noise, normalize the data, and perform necessary transformations and feature extraction. Through this process, the system can ensure that only the most relevant data is input into the prediction model, thereby improving the accuracy and efficiency of predictions.

[0078] Predictive Model Module (red box): The third module is the predictive model, which uses deep learning technology to predict future harmonic trends and energy demand. This model combines multi-dimensional data (such as voltage, current, frequency, etc.) with a deep learning hybrid model including an attention mechanism, enabling accurate predictions in complex power grid environments. The predictive model's role is to analyze and predict potential harmonic problems and energy demand changes in real time, thereby providing early warnings for energy dispatch and harmonic suppression.

[0079] Control Algorithm Module (Purple Box): Finally, the system optimizes the power allocation of energy equipment based on the harmonic trends and energy demands provided by the predictive model using the Model Predictive Control (MPC) algorithm. The MPC algorithm dynamically adjusts the operating status of each energy device in the ship's power grid by simulating the results of different control strategies, thereby maximizing energy efficiency and minimizing harmonics.

[0080] Overall, Figure 3 The structure in the diagram demonstrates the key processes of the ship's power grid harmonic active suppression and energy management system. Through the combination of real-time data acquisition, preprocessing, deep learning prediction, and control algorithms, the system can accurately predict changes in power grid load, adjust energy management strategies in a timely manner, and optimize energy efficiency and power quality.

[0081] The third step is to execute the coupled energy scheduling step.

[0082] This step receives the predicted information from the second step, namely harmonic trends and energy demand, and performs a dynamic optimization process aimed at determining an optimal power allocation scheme. The coupling of this step lies in the fact that it must simultaneously balance two seemingly conflicting objectives:

[0083] First, power quality: it is essential to... Suppressed at the upper limit of national standards Below, such as below 5%;

[0084] Second, operational economy: fuel consumption must be minimized. and make reasonable use of .

[0085] The means to achieve these two goals are: active power filters (APF) (used to suppress harmonics) and energy devices (generators, energy storage, used to supply energy).

[0086] However, the APF itself is also an energy-consuming device. The APF cancels harmonics by injecting compensating current, a process that consumes active power. Running the APF at full power to reduce THD from 2% to 1% would result in a 10% increase in fuel consumption, which is economically unacceptable.

[0087] Therefore, the first key issue to be addressed in this step is to accurately determine the operating cost of the APF.

[0088] Therefore, this step first requires real-time monitoring of the device-level power consumption of the active power filter (APF), which directly reflects the operating cost of the APF. Device-level power consumption includes at least the following two main components:

[0089] 1. IGBT switching losses, which can be represented by the symbol: This is the energy loss generated by APF power semiconductor devices during high-speed switching, and its calculation method can be expressed as:

[0090]

[0091] in, : Represents the instantaneous voltage between the collector and emitter of the IGBT power transistor in the APF;

[0092] : Represents the instantaneous collector current flowing through the IGBT;

[0093] : Represents a complete sampling period.

[0094] In digital control systems, the above integration operation is typically implemented as high-speed sampling and summation, where the two instantaneous values... and The voltage and current are measured directly by the voltage and current sensors inside the APF.

[0095] 2. Reactor loss, its symbol can be represented as This is the heat loss generated by the AC or DC side filter / coupling reactor of the APF due to the flowing current, which can be calculated as follows:

[0096]

[0097] in, : Represents the power loss of the reactor;

[0098] : Represents the effective value (RMS value) of the operating current flowing through the reactor, which is measured by the current sensor inside the APF;

[0099] : Represents the DC resistance of the reactor coil or its equivalent resistance. This is an inherent parameter of the device, which is calibrated and stored in the controller when the APF leaves the factory.

[0100] Subsequently, by using the real-time total energy consumption of the APF By performing real-time calculations, the system accurately determines the energy cost required to suppress harmonics.

[0101] Next, this step performs its core task: calculating the harmonic-energy dynamic priority coefficients. .Should The coefficient is a dimensionless scalar whose value dynamically reflects which is more urgent under the current system conditions: power quality or energy economy.

[0102] For example, harmonic-energy dynamic priority coefficients The calculation formula can be expressed as:

[0103]

[0104] in, : That is, dynamic priority coefficient, The higher the value, the worse the current state of the system or the further it deviates from the ideal state;

[0105] Total harmonic distortion rate of the power grid as monitored in real time by the first step S1;

[0106] The system must adhere to the upper limit of harmonic values, which is a hard constraint, such as 5% as specified by the classification society;

[0107] The rated state of charge of the energy storage unit, i.e., the optimal amount of energy the system wants the stored energy to maintain, for example, 90%;

[0108] The actual state of charge of the energy storage unit as monitored in real time by the first step S1;

[0109] : Instantaneous fuel consumption rate of the generator set (g / kWh) monitored in real time by the first step S1;

[0110] The economic operating fuel consumption rate of a generator set is a known minimum fuel consumption rate at a specific power point, serving as a benchmark for the system's energy efficiency.

[0111] These represent harmonic weight, SOC deviation weight, and fuel consumption weight, respectively. These three coefficients are strategy parameters that can be set by the operator.

[0112] Further analysis reveals the dynamic priority coefficients of harmonics and energy. The calculation formula consists of three core pressure terms:

[0113] The first term is the harmonic stress term: This reflects the urgency of power quality. When near When this ratio is close to 1.0, the pressure increases; if If it is set very high, then The value will be very sensitive to harmonics;

[0114] The second item is the energy storage pressure item: This reflects the urgency of energy reserves. When far below For example, when the battery is low on charge, this item will have a large positive value, indicating increased pressure and a need for system charging; while if If the setting is very high, the system will prioritize ensuring that the energy storage is at full capacity.

[0115] The third item is the fuel pressure item: This reflects the urgency of operational economy. When the generator operates in the inefficient zone, It will be much higher than This results in a ratio much greater than 1, increasing pressure; if If the system is set too high, it will go to any lengths, such as frequently starting and stopping the energy storage, to bring the generator back to normal. Run point.

[0116] Finally, this step is based on this The coefficients are used to plan the power allocation of each energy device using the Model Predictive Control (MPC) algorithm. MPC is an advanced control algorithm whose core principles are rolling optimization and forward-looking approach.

[0117] In this embodiment, the MPC algorithm will perform the following operations:

[0118] 1. Objective Function: The core of MPC is an objective function. In this embodiment, this It is set to minimize the time step over the next N time steps. The cumulative integral of the value, i.e. ;

[0119] 2. Constraints: MPC must adhere to a series of constraints during optimization, such as... ;

[0120] 3. Prediction Model: MPC needs a model to predict: if I perform action A, what will happen to K in the future? This model is the harmonic and energy prediction model of the second step S2.

[0121] 4. Rolling Optimization: The MPC controller calculates the complete optimal control sequence for the next 5 minutes, but it only executes the first action in the sequence. Then, in the next control cycle, it reacquires the new data for S1, reruns the prediction for S2, recalculates the K value for S3, and performs the optimization for the next 5 minutes again.

[0122] In this way, the MPC algorithm utilizes Using the coefficient as the optimization objective and the S2 prediction model as a crystal ball, an optimal power allocation scheme is finally calculated.

[0123] like Figure 4 A schematic diagram illustrating the calculation of harmonic-energy dynamic priority coefficients is shown. Figure 4 In the figure, the horizontal axis represents time, and the vertical axis represents the rate of change of standardized operating parameters, which respectively represent the harmonic distortion rate of the ship's electrical network, the energy storage state of charge deviation, and the fuel consumption pressure. The black dashed line represents the dynamic priority coefficient calculated by the system.

[0124] Figure 4 The red curve shows the relative trend of the total harmonic distortion rate of the power grid. This trend reflects the urgency of power quality. When the red curve rises, it indicates that the current power grid is under strong nonlinear interference and active filters are needed to compensate with higher power.

[0125] The blue curve represents the degree of deviation of the energy storage SOC. It rises sharply in the first 20 seconds, indicating that the energy storage is in a rapid discharge state, which is a typical situation when the system is dealing with transient load impacts.

[0126] The green curve represents the unit power consumption of the generator. Its periodic changes indicate that the system load is experiencing periodic fluctuations, which corresponds to the common scenario of alternating operation of propulsion and shipboard equipment in actual ship navigation.

[0127] The weighted output of these three factors is the priority coefficient shown by the black dashed line in the figure. It can be seen that this coefficient reaches its peak at about 15 seconds. Combined with the model predictive control mechanism, the system will prioritize the full power compensation of the low-order dominant harmonics and improve the energy storage scheduling capability based on this value, thereby ensuring that energy use efficiency is controlled while guaranteeing power quality.

[0128] For example, scenario 1: S2 predicts the next minute. A surge, predicted by MPC and It will surge, leading to The value is about to explode;

[0129] The optimization results of MPC are:

[0130] 1. Immediately instruct the energy storage to prepare for high-power discharge in order to mitigate the power surge;

[0131] 2. Immediately instruct the APF to enter full-power compensation mode, because High weighting necessitates the suppression of harmonics;

[0132] 3. Instruct the generator to smoothly increase its output.

[0133] Scenario 2: S2 predicts stable cruise for the next 5 minutes, but S1 detects... Very low, at this time The second value (energy storage pressure) is very high;

[0134] The optimization results of MPC are:

[0135] 1. Slightly increase the generator output to a level slightly higher than the power required for cruising;

[0136] 2. Charge the excess power command stored energy;

[0137] 3. At the same time because With a high weighting, MPC will ensure that the generator's output point is exactly at the economical fuel consumption point. ;

[0138] 4. The energy requirements of the instruction APF are maintained at just enough. The minimum compensation power is used to save the APF's own energy consumption.

[0139] like Figure 5As shown, the power allocation changes of three types of core energy equipment in the ship's power grid before and after the implementation of model predictive control are illustrated. The upper part shows the original scheduling behavior before optimization, and the lower part shows the allocation results after adopting the predictive control algorithm.

[0140] The red curve represents the output power of the diesel generator. Before optimization, it fluctuated significantly, changing by nearly 100 kilowatts per minute. This frequent load fluctuation caused fuel consumption to deviate from the economic operating point, leading to increased operating costs and wasted fuel. The blue curve represents the output power of the energy storage system. Before optimization, although it provided some regulation, the response was delayed and the feedback was unstable, failing to fully utilize its peak-shaving and valley-filling efficiency. The green curve represents the operating power of the active power filter. Before optimization, its fluctuation frequency was too high, the control strategy lacked judgment on future harmonic trends, and the compensation behavior was blind and inefficient.

[0141] In contrast, Figure 5 The lower half of the diagram shows the optimized results. We can see that the diesel generator output is smoother, and the power setting is closer to the high-efficiency fuel zone, aligning with an operation strategy focused on optimal fuel consumption. The energy storage unit exhibits positive and negative power switching behavior at different times, indicating its use in dynamically following and predicting load fluctuations, demonstrating forward-looking adjustment capabilities. This reflects the control and scheduling optimization strategy introduced in step three after model prediction. The active filter's power curve shows a slight increase in amplitude but a decrease in frequency, indicating that its scheduling is no longer triggered by single real-time detection but rather by proactive control based on prediction results. In particular, there is a tendency to increase power before instantaneous load disturbances occur, demonstrating a shift from passive compensation to active intervention.

[0142] The fourth step is to perform the coordinated harmonic suppression step.

[0143] This step performs cooperative harmonic suppression based on the power allocation results output by S3's MPC algorithm. This step includes two aspects: cooperative and pre-control.

[0144] 1. Coordinate the distributed deployment of master-slave APFs for harmonic suppression.

[0145] On large ships, the APF (Automatic Power Quality) is not a single device, but a distributed system. Typically, there is a master APF on the main switchboard, responsible for the overall power quality at the PCC (Power Control Center) point; and multiple slave APFs near large harmonic sources such as propulsion inverters, responsible for local compensation. In this embodiment, "coordination" refers to:

[0146] The total compensation command calculated in the third step S3 is sent to the master APF; the master APF is responsible for decomposing this total task. It obtains the compensation status of each slave APF and the harmonic situation of its location in real time through high-speed communication.

[0147] The master APF sends specific, differentiated compensation current commands to each slave APF based on a cooperative algorithm, such as the optimal allocation of harmonic contributions at each point.

[0148] This master-slave collaboration ensures that the system can always operate at the lowest possible cost. To achieve the optimal .

[0149] 2. Execute local pre-control policies when communication delays are detected.

[0150] This is a necessary measure to ensure system robustness. Communication delay refers to the interruption, packet loss, or delay in the high-speed communication link between the central EMS and the main APF, or between the main APF and the slave APF; if this happens, the slave APF cannot stop and wait, otherwise the harmonics will instantly become uncontrollable.

[0151] The local pre-control strategy in this embodiment refers to:

[0152] Trigger: When an instruction from the main APF is detected in the APF, but the APF fails to refresh within the specified time;

[0153] Execution: Immediately switch from collaborative mode to local autonomous mode from APF;

[0154] Autonomous Strategy: In local autonomous mode, the slave APF no longer waits for commands from its superior. Instead, it independently measures the harmonic current at its connection points using its local current sensors and performs local compensation to the best of its ability based on a built-in control algorithm that prioritizes safety over economy. Control algorithms may include constant DC bus voltage control or simple FFT harmonic extraction.

[0155] Recovery: Once communication is restored, the APF will immediately (or under the master control command) smoothly exit the local autonomous mode and return to the cooperative mode.

[0156] This mechanism ensures that even in the extreme case of communication failure, the power quality of the ship's electrical grid will not suffer a catastrophic collapse.

[0157] like Figure 6 The bar chart illustrates the task allocation of the three active power filters (APFs) in a distributed deployment across various harmonic orders. The horizontal axis represents typical odd harmonic orders, and the vertical axis represents the normalized compensation amplitude, characterizing the control load and target focusing of each APF in a specific harmonic frequency band.

[0158] Figure 6The blue area represents APF1, whose compensation capability is mainly concentrated on the fifth to eleventh harmonics, which are the most common and detrimental low-order dominant harmonics in the operation of ship propulsion frequency converter systems. Because the APF is deployed close to the load, it can efficiently and quickly suppress these harmonics on-site, ensuring power stability in the low-frequency band.

[0159] The red area represents the main active damping factor (APF), primarily concentrated in the 13th to 23rd harmonic frequency band. This band often falls within the region where the grid impedance may generate resonance peaks. The main APF is responsible for active damping control at high impedance frequencies, thereby improving the overall system stability and vibration resistance.

[0160] The yellow section represents APF2, which handles the compensation for the 25th and higher high-frequency harmonics. Most of these harmonics are induced by high carrier frequency modulation strategies. Although their amplitudes are low, they may affect the operation of precision instruments. Therefore, an auxiliary APF with a narrower bandwidth is needed to filter them, which is far away from the interference source.

[0161] Overall, Figure 6 This clearly reflects the strategy of each APF to divide the compensation frequency band according to the grid harmonic spectrum distribution and impedance characteristics in the collaborative mode, confirming the collaborative suppression mechanism driven by grid impedance analysis and characteristic harmonic database, and highlighting the engineering innovation value of this technical solution in achieving high-efficiency, low-redundancy, and low-energy-consumption operation in the multi-point power supply system of large ships.

[0162] This first embodiment successfully incorporates the ship's operational intent, control dynamics, system status, and operational strategies into a unified optimization framework based on model predictive control (MPC). This transforms ship power grid management from the old model of passive governance and independent management to a new model of prediction-driven, goal-coupled, and global optimization, ultimately maximizing operational energy efficiency while ensuring the safety of ship navigation.

[0163] Example 2:

[0164] The core difference and innovation of this embodiment lies in providing a more technically complete fourth step, namely, the implementation method of the cooperative harmonic suppression step.

[0165] In the cooperative harmonic suppression step S4 of Example 1, the system relies on a central energy management system (EMS) or main controller to execute the MPC optimization algorithm of the third step S3, and sends the calculated optimal power allocation results and cooperative suppression instructions to the distributed active power filter (APF) execution units on the ship via a communication network.

[0166] However, as an independent system operating in a harsh environment, a ship's internal communication network, whether it is industrial Ethernet or marine bus, may face communication delays, data packet loss, or even temporary interruptions caused by electromagnetic interference, vibration, or line failures.

[0167] In traditional centralized control schemes, if central commands are delayed or interrupted, the local APF (Automatic Power Controller) execution unit will become uncontrollable. If the APF continues to execute old commands from the previous cycle, it may cause compensation overshoot or lag, resulting in resonance with the power grid; if the APF chooses to shut down for protection, it will cause instantaneous harmonic runaway and severe deterioration of power quality. Both of these situations are unacceptable for propulsion and navigation systems that rely on power quality.

[0168] The technical problem to be solved in this second embodiment is the aforementioned technical problem. In step S4 of this embodiment, a communication delay pre-control strategy is further added on top of the basic collaborative function. The core purpose of this strategy is to ensure that, in the extreme case of any APF execution unit losing connection with the central controller, it can immediately and seamlessly switch to an intelligent, locally autonomous mode aimed at self-survival and basic grid stability, achieving smooth system degradation rather than catastrophic collapse.

[0169] Specifically, the fourth step S4 of this embodiment, the cooperative harmonic suppression step, has its internal logic divided into two complementary operating modes:

[0170] Mode 1: Centralized Coordination and Scheduling Mode (Default Mode)

[0171] When the communication link is normal, step S4 in this embodiment is the same as in embodiment one. Each APF acts as a slave and strictly executes the harmonic compensation commands issued by the central controller, i.e., the MPC algorithm output in step S3. The central controller has a global view, and its commands aim at maximizing the overall system energy efficiency and optimizing global harmonics.

[0172] Mode 2: Local Autonomous Pre-Control Mode

[0173] 1. Mode triggering conditions:

[0174] Each distributed APF execution unit in this embodiment has a built-in heartbeat monitor that continuously monitors the instruction refresh status from the central controller or the main APF. When the APF detects that it has not received a valid new instruction within a preset time threshold, the system determines that a communication delay or interruption has occurred. At this time, the APF will immediately and automatically trigger, switching from the central collaborative scheduling mode to the local autonomous pre-control mode.

[0175] 2. The core of the local autonomy strategy: bandwidth adaptation based on its own state.

[0176] Once in a local self-governance model, the primary task of the APF will no longer be economic, but rather survival and criticality:

[0177] Survivability: The APF must ensure that its power semiconductor devices, such as IGBTs, do not overheat and burn out, and that its DC-side energy storage capacitors or local energy storage units do not experience voltage collapse due to overcompensation.

[0178] Key point: The APF must prioritize suppressing harmonics that pose the greatest threat to the power grid, such as low-order harmonics, while ensuring the grid's survival.

[0179] To achieve the above goals, in its local autonomous mode, the APF will no longer blindly attempt to compensate for all harmonics. Instead, it will dynamically calculate a bandwidth adaptive coefficient based on the APF's real-time energy consumption and DC-side energy storage status. Bandwidth adaptive coefficient The calculation formula can be defined as:

[0180]

[0181] in, This refers to the bandwidth adaptive coefficient. It is a value between 0 and... The dimensionless value between these values. It comprehensively assesses the current health status and compensatory potential of the APF. The higher the value, the more spare capacity the APF has, and the more compensation bandwidth it can handle and the heavier compensation tasks it can handle.

[0182] This represents the real-time total energy consumption of the APF, a value monitored in real time by the APF's local controller. This total energy consumption is preferably the IGBT switching losses within the APF. and reactor copper loss the sum of It directly reflects the thermal stress level of the APF.

[0183] : Represents the rated maximum power consumption of the APF, which is an inherent parameter calibrated at the factory and represents the limits of the APF's heat dissipation and design;

[0184] : Represents the current state of charge of the APF DC-side energy storage unit. This value reflects the energy reserves of the APF when performing transient compensation.

[0185] : Represents the rated state of charge of the DC-side energy storage unit, i.e., its full energy storage value;

[0186] and : These represent energy consumption weight and energy storage weight, respectively. These two coefficients are pre-set strategy parameters used to adjust the conservative or aggressive nature of the APF. For example, if If the value is set too high, the APF will be extremely sensitive to its own heat generation (thermal stress) and will preferentially reduce its derating in order to survive.

[0187] In this formula: the first term (thermal stress constraint term): This reflects the heat margin of the APF. When the real-time total energy consumption of the APF... Approaching its rated maximum power consumption At this point, the ratio is close to 1, causing this term to approach 0. At this time, regardless of the energy storage level, this term will... The value is lowered. This is a mandatory self-protection mechanism to prevent the APF from overheating and burning out due to overcompensation during local autonomy;

[0188] Second item (energy reserve constraint): This reflects the energy margin of the APF. When the APF stores energy on the DC side... When the ratio decreases due to continuous compensation, the ratio will also decrease. The value is lowered. This prevents the DC bus voltage from collapsing due to insufficient energy reserves in the APF, thus avoiding control failure.

[0189] 3. Subsequently, execution based on Priority-based suppression strategy for coefficients:

[0190] After calculating the current After the coefficients are applied, the APF's local controller will execute a priority-based harmonic suppression strategy, as follows:

[0191] First priority (mandatory): Suppress low-order dominant harmonics. The APF local controller will always (as long as) The APF (Automatic Power Grid Controller) uses its controlled resources to suppress the most harmful and concentrated low-order dominant harmonics to the power grid. In a typical marine power grid, this usually refers to the 5th, 7th, 11th, and 13th harmonics. This is the core task of the APF in its local autonomous mode, aiming to maintain the basic stability of the power grid.

[0192] Second priority ( (Dependency-dependent execution): To suppress higher-order or non-characteristic harmonics, the APF's local controller will set a... Threshold of coefficient ;

[0193] when This indicates that the APF has sufficient heat margin and energy margin, i.e. Lower, and The power quality is relatively high. At this point, the APF controller determines that it has sufficient capacity. While completing its first priority task, which is to suppress low-order harmonics, it will also activate additional suppression channels for high-order harmonics or non-characteristic interharmonics to provide better power quality.

[0194] when This indicates that the APF's margin is insufficient, and the system is under high load or energy deficit. At this time, the APF will perform a smooth degradation operation. It will actively shut down the second-priority compensation channel, that is, stop compensating for higher harmonics, and use all of the APF's limited compensation capacity solely to ensure the suppression of first-priority low harmonics.

[0195] 4. Mode recovery:

[0196] In local autonomous mode, APF will continuously attempt to reconnect to the central controller. Once the heartbeat signal is restored, APF will not immediately switch back to central mode, but will execute a smooth transition procedure. Over several cycles, it will gradually fade out its locally generated compensation instructions while smoothly introducing new instructions issued by the central controller, and finally switch back to central cooperative scheduling mode without impact.

[0197] This second embodiment provides a communication delay pre-control strategy for the S4 suppression step in embodiment one. This strategy introduces a bandwidth adaptive coefficient based on the APF's own thermal capability and energy reserves. Combined with priority-based harmonic suppression logic, the APF (Automatic Power Controller) possesses intelligent local autonomy when disconnected from central control. This addresses the core pain point of communication uncertainty in large ship power grids, ensuring high reliability and stability under extreme conditions. It achieves a smooth degradation capability from best-effort to bottom-line assurance, demonstrating significant engineering application value.

[0198] Example 3:

[0199] The core difference and innovation of this embodiment lies in the fact that, in order to address the spatial harmonic coupling and network resonance problems that may occur when distributed active power filters (APFs) work together, a spatial collaborative control implementation method based on grid impedance is provided for the collaborative harmonic suppression step.

[0200] In large shipboard power grids, especially those systems employing multi-zone distribution, long cable connections, and multiple Active Power Filters (APFs) deployed on different busbars (including the main APF on the main busbar and auxiliary APFs on branch busbars near large loads), the compensation behavior between APFs is not isolated. The compensation current injected by one APF at a certain point flows through the grid impedance, affecting not only its own connection point but also coupling to the connection points of other APFs, forming complex spatial harmonic coupling paths.

[0201] When the equivalent impedance of a certain harmonic in the power grid exhibits high impedance characteristics at a specific frequency, the active power supply (APF) may induce resonance when compensating for that harmonic, leading to harmonic amplification rather than suppression. The traditional solution is to simply have all APFs compensate for all harmonics, which wastes resources and exacerbates coupling and potential resonance risks.

[0202] The technical problem to be solved in this third embodiment is the aforementioned one. Its core lies in enabling the distributed APF system to perform targeted and differentiated harmonic suppression by acquiring and analyzing the frequency characteristics of the power grid in real time, thereby achieving precise spatial coordination. The spatial coordinated control strategy in the fourth step S4 of this embodiment includes the following core components:

[0203] 1. Measure and calculate the equivalent impedance of the power grid under different harmonic sequences in real time.

[0204] Measuring equivalent impedance is a prerequisite for implementing spatial coordination. In shipboard electrical networks, the equivalent impedance of the network varies due to the continuous changes in generators, converters, and loads. It is a dynamically changing function, and this embodiment requires the main APF to perform this critical analysis task; while the impedance resonant frequency point refers to the equivalent impedance. The frequency point at which a peak occurs at a specific harmonic frequency is prone to triggering power grid resonance.

[0205] The main APF (Automatic Power Distribution Filter) is located at the main distribution bus of the system and has the highest awareness of the overall grid status. This main APF injects a small-amplitude sweep frequency excitation current into the grid while simultaneously monitoring the voltage response generated by this excitation in real time. The injected current... and the measured voltage Frequency domain analysis, such as using the discrete Fourier transform, can be performed at various harmonic frequencies. Calculate the corresponding equivalent impedance. .

[0206] The main APF is based on the current injection method to obtain the equivalent impedance of the power grid under different harmonic orders in real time. equivalent impedance The calculation formula can be defined as:

[0207]

[0208] in, : Represents the equivalent impedance of the power grid at the point of common coupling (PCC) or the main bus. It is a complex number that includes the magnitude and phase angle of the impedance.

[0209] This represents the voltage change at the PCC point after the main APF injects a sweep current; this change occurs at a specific harmonic frequency. Measured at the location;

[0210] : Represents the frequency sweep excitation current injected into the power grid by the main APF, which is present at various harmonic frequencies. It has a known, predetermined amplitude and phase.

[0211] This formula follows the basic principle of Ohm's law. By measuring the voltage response caused by a known excitation current in the power grid, the main APF can accurately plot the impedance spectrum of the ship's power grid in the range of hundreds to thousands of hertz. The peak position of this impedance spectrum indicates the potential resonant frequency.

[0212] 2. APF master-slave task allocation based on impedance information.

[0213] The equivalent impedance spectrum of the entire system is calculated using the main APF. Then, it uses this information as the core parameter of the coordinated control command and sends it to all auxiliary APFs through a high-speed communication network.

[0214] The cooperative inhibition module in this step will, according to The characteristics of the spectrum decompose the entire harmonic suppression task into two differentiated sub-tasks:

[0215] The main APF's task (suppressing resonance risk): The main APF's control strategy is no longer to fully compensate for all harmonics; instead, the main APF will prioritize concentrating the compensation bandwidth on the equivalent impedance. The order of harmonics with extremely large amplitudes;

[0216] When impedance At a certain harmonic frequency When a peak value appears at a certain frequency, it indicates that the power grid is highly susceptible to resonance at that frequency. In this case, the main APF will proactively adjust its compensation target.

[0217] At the impedance peak: the main APF will employ an active damping control strategy. Its compensation objective is not to eliminate harmonic currents, but rather to simulate a low-damping resistor in parallel with the high impedance peak. This is equivalent to the main APF smoothing and filling the impedance valleys of the power grid at that resonant frequency, fundamentally eliminating the risk of resonance.

[0218] At off-peak times: the compensation intensity of the main APF will be appropriately reduced to save its own operating energy consumption, and the remaining compensation tasks will be allocated to the auxiliary APF.

[0219] The task of the auxiliary APF (on-site pollution elimination): The auxiliary APF is deployed at the interface of the main nonlinear load, such as the electric propulsion inverter. The auxiliary APF receives data from the main APF... After spectrum analysis, targeted in-situ compensation will be performed. The auxiliary APF will avoid compensating for the equivalent impedance. The harmonic sequence with extremely large amplitude is compensated for by focusing its bandwidth on the remaining non-resonant harmonic sequences generated by the local load.

[0220] For example, if Analysis shows that the 17th harmonic is the main resonant point, so the main APF is responsible for damping control of the 17th harmonic; while the auxiliary APF will skip the compensation of the 17th harmonic and instead focus on eliminating harmonics of other orders such as the 5th, 7th or 29th generated by the local load.

[0221] 3. Dynamic collaboration and effects.

[0222] The final effect of this embodiment is that, through This real-time analysis of spatial coupling parameters and task allocation enables the primary APF and auxiliary APF to achieve differentiated and complementary synergy:

[0223] The main APF acts as a system stabilizer through an active damping strategy, eliminating the root cause of resonance at a global level; the auxiliary APF acts as a local purifier through a selective compensation strategy, effectively removing harmonic pollution generated locally at its access point.

[0224] This spatial coordination method avoids blind competition or mutual interference among multiple APFs due to overlapping compensation targets, greatly improving the compensation efficiency and system stability of the entire distributed APF system. This scheme solves the long-standing problem of 1+1<2 faced by distributed harmonic suppression systems, achieving optimal resource allocation.

[0225] This third embodiment, building upon the predictive scheduling framework of the first embodiment, provides a spatially coordinated advanced solution for step S4 by fully integrating analysis and control strategies based on the grid's equivalent impedance. The main APF uses the current injection method to calculate the equivalent impedance in real time and identify resonant points. Subsequently, the system decomposes the harmonic suppression task: the main APF is responsible for eliminating the resonance risk of high impedance sequences, while the auxiliary APF is responsible for eliminating harmonic pollution of local low impedance sequences. This precise targeted allocation fundamentally improves the suppression efficiency of the distributed APF system, ensuring high-quality power even in complex shipboard power grid environments and effectively avoiding systemic resonance faults.

[0226] Example 4:

[0227] This embodiment further proposes a precise compensation and dynamic suppression scheme based on a harmonic feature library, targeting the inherent laws governing harmonic generation in power electronic equipment in ship power grids.

[0228] Because the power electronic equipment such as frequency converters and inverters used extensively in shipboard power grids have harmonic emission characteristics that are closely related to modulation methods, topology, and operating load rates, traditional broad-spectrum proportional compensation methods lack specificity and are inefficient. This embodiment constructs a multi-dimensional harmonic characteristic knowledge base and performs forward-looking queries and planning based on load forecasting results, achieving a leap from passive response to proactive and precise suppression. Simultaneously, it innovatively proposes a two-stage suppression strategy targeting both transient and steady-state load processes, significantly improving the efficiency and economy of harmonic suppression.

[0229] 1. Construction and content of the harmonic feature library.

[0230] This paper constructs an inverter harmonic characteristic library, a structured database used in the system to store typical harmonic order sets under different inverter types, modulation methods, and load rates. Its design aims to systematically store and manage the typical harmonic emission characteristics of various commonly used power electronic converters in ships under different operating conditions. The library is built based on extensive offline testing, simulation analysis, and mining of historical operating data, covering mainstream inverter types in ship power grids, including but not limited to propulsion inverters, pump drives, and auxiliary power supplies.

[0231] The feature library is first categorized and archived according to the inverter's modulation method. Common modulation methods, such as sinusoidal pulse width modulation (PWM) and space vector pulse width modulation (SVM), have significantly different harmonic spectrum distributions due to their different switching pulse generation mechanisms. For example, the harmonic energy of sinusoidal PWM devices is mainly concentrated near integer multiples of the switching frequency, while SVM may produce different sideband harmonic structures.

[0232] Secondly, the feature library is further subdivided according to the inverter's topology. Two-level topology, as the most basic structure, has a relatively classic harmonic spectrum. Three-level topologies, such as neutral-point clamping topologies, effectively reduce the harmonic distortion rate of the output voltage due to the increased number of output levels. Their harmonic spectrum composition differs fundamentally from that of two-level topologies, primarily manifested in better harmonic performance at lower switching frequencies. The feature library accurately records the characteristic harmonic modes of these different topologies.

[0233] Most importantly, the feature library establishes a precise mapping relationship between load factor and harmonic sequence set. Load factor is the most important operating parameter affecting harmonic amplitude and composition. For each included inverter model, under a specific modulation method and topology, the feature library stores its typical harmonic sequence set at different load factors.

[0234] The harmonic order set refers to the set of harmonic orders that are most likely to occur and have the highest energy at a given operating point, which can be represented as: ,in This represents the number of major harmonics at this load rate.

[0235] For example, a typical harmonic sequence set of a six-pulse frequency converter at a 50% load rate may be the 5th, 7th, 11th, and 13th harmonics; while when the load rate increases to 90%, the harmonic sequence set may remain unchanged, but the amplitude ratio of each harmonic will change, and it may even activate higher harmonics such as the 17th and 19th.

[0236] In order to manage harmonic data more precisely and support dynamic suppression strategies, this embodiment further divides the inverter harmonic feature library into two logic sub-libraries: steady-state harmonic sub-library and transient harmonic sub-library.

[0237] (1) The steady-state harmonic sub-library is used to store the mapping relationship between load rate and steady-state characteristic harmonic sequence set. Here, steady state refers to the inverter operating at a certain constant or slowly changing load rate, and its harmonic emission also reaches a relatively stable periodic state. This sub-library is the fundamental basis for performing steady-state accurate compensation.

[0238] (2) The transient harmonics sub-library focuses on harmonic phenomena during rapid load changes. This sub-library stores the mapping relationship between the load gradient during the transition from the current load rate to the target load rate and the transient interharmonic spectrum. The load gradient is defined as the ratio of the load rate change to the change time, which quantifies the severity of the load change. When the load rate undergoes a rapid transition, the dynamic response process of the inverter's control system will trigger aperiodic current surges, thereby generating a large number of interharmonics and subharmonics. These harmonic components are not limited to integer multiples of the power frequency, and their spectrum is continuously distributed. By recording the typical interharmonic frequency band distribution under different load gradients, the transient harmonics sub-library provides key data support for dealing with harmonic disturbances during transient processes.

[0239] 2. Dynamic query and precise compensation based on feature library.

[0240] During system online operation, the harmonic characteristic library is not a static reference, but an active component that dynamically participates in decision-making. At the beginning of each scheduling cycle, the system dynamically queries the harmonic characteristic library by combining the power demand forecast and load forecast results for the next scheduling cycle obtained in step S2.

[0241] The query process is as follows: First, based on load forecasting, the system learns which major inverter loads will be put into operation or change their operating status in the next cycle, and estimates their target load rates. Then, the system uses the load type, modulation method, topology, and predicted target load rate as indexes to query the steady-state harmonic sub-library and the transient harmonic sub-library in parallel.

[0242] From the steady-state harmonic sub-library, the system obtains the sequence set of steady-state high-energy harmonics most likely to be generated by these loads under the current predicted operating conditions. The steady-state high-energy harmonic sequence set refers to the set of characteristic harmonics with the highest amplitude at the corresponding load rate, obtained statistically from the steady-state harmonic sub-library of the harmonic feature library. This harmonic sequence set indicates the target harmonics that require focused attention after the load stabilizes.

[0243] From the transient harmonic sub-library, the system calculates the load gradient based on the difference between the current load rate and the target load rate, and then queries the transient interharmonic frequency bands that may be excited under this gradient. This prepares the system for dealing with harmonic impacts during load switching or power adjustments.

[0244] After obtaining the query results, the collaborative harmonic suppression module instructs the corresponding active filters to execute a hierarchical compensation strategy. For the high-energy harmonic order set obtained from the steady-state harmonic sub-library, the active filters are instructed to prioritize precise compensation for that specific harmonic order set with adaptive bandwidth. Adaptive bandwidth means that the bandwidth of the compensator is not fixed, but dynamically adjusted according to the order and distribution of the harmonics to be compensated, to ensure complete coverage of the target harmonic group while avoiding interference with irrelevant frequency bands. This targeted compensation, like precision guidance, concentrates limited switching frequencies and compensation energy on the harmonics that contribute the most to the total harmonic distortion rate, resulting in extremely high efficiency.

[0245] Meanwhile, the system does not completely abandon the monitoring and suppression of other non-characteristic harmonics. For other subharmonics not included in the high-energy order set, the active filter uses a fixed, relatively narrow, low bandwidth for background suppression. This background suppression mode has low power consumption and mainly serves as a safety net, preventing some accidental or small-amplitude harmonic components from going out of control. Through this clear-cut and focused compensation strategy, the system achieves an optimal balance between overall energy efficiency and compensation effect.

[0246] 3. Two-stage suppression program.

[0247] Based on the detailed harmonic feature library and a deep understanding of the load dynamic process, this embodiment proposes and implements a complete two-stage suppression planning process. This plan clearly divides harmonic suppression into two time-series stages: a transient suppression stage and a steady-state suppression stage, which respectively address different periods of the load change process.

[0248] (1) Transient Suppression Phase: The two-stage suppression planning is immediately initiated when the system detects, through prediction algorithms or direct commands, that the load rate of a critical load is about to change rapidly. Just before the actual load rate transition, the system prioritizes querying the transient harmonic sub-library. The key input for this query is the difference between the current load rate and the target load rate, which is directly related to the load gradient. Based on this gradient value, the system obtains the main distribution frequency bands of the upcoming transient interharmonics from the transient harmonic sub-library.

[0249] Immediately, the system issued an emergency command to the active filter responsible for compensation in that area, requiring it to switch immediately from the current steady-state compensation mode to broadband compensation mode. In this mode, the compensation frequency response bandwidth of the active filter is significantly amplified, enabling it to cover the interharmonic frequency band indicated by the transient harmonic sub-library. This broadband compensation mode acts like a large net, designed to capture and cancel as many variable-frequency, widely distributed transient interharmonics and subharmonics as possible during the brief period of drastic load changes, quickly smoothing out voltage and current surges and distortions, and maintaining the stability of the power grid during the transition period. The core objective of this stage is to provide a safety net, ensuring the system safely weathers the transient process.

[0250] (2) Steady-state suppression stage: After the load rate has completed its transition and has been running stably for a period of time, the system confirms that it has stabilized near the target load rate by monitoring the actual load rate. At this time, the cooperative harmonic suppression module automatically triggers the switching of the working mode. The system stops querying the transient harmonic sub-library and instead queries the steady-state harmonic sub-library.

[0251] Using the current stable actual load rate as an index, the system retrieves the corresponding steady-state high-energy harmonic sequence set under this stable operating condition from the steady-state harmonic sub-library. This sequence set is more concentrated and regular than the spectrum in the transient stage.

[0252] Subsequently, the system instructs the active filter to switch from broadband compensation mode back to adaptive bandwidth mode. However, the adaptive bandwidth at this point is optimized for the newly queried set of steady-state high-energy harmonic sequences. The compensator concentrates its main energy and control resources on precisely compensating for these characteristic harmonic sequences. This mode acts like a precision scalpel, meticulously cleaning the power grid after the transient process subsides, minimizing harmonic distortion. The core objective at this stage is optimization, pursuing higher operational economy while ensuring suppression effectiveness.

[0253] like Figure 7 By using a three-layer time-series curve overlay method, the study fully demonstrates how the two-stage suppression strategy responds in real time and dynamically adjusts its compensation behavior during rapid load transitions in the ship's electrical network.

[0254] Figure 7The blue curve at the top represents the evolution of the inverter load rate over time. It can be seen that the load rate starts to climb rapidly from 30% at the 20th second and stabilizes at around 70% at about the 35th second. This transition process simulates a typical scenario of a ship's propulsion system switching from cruise to acceleration mode.

[0255] The red curve, located in the middle layer, represents the compensation frequency response bandwidth coefficient of the active filter. This parameter directly determines the filter's coverage of the harmonic spectrum. The system immediately queries the transient harmonic sub-library and triggers broadband compensation mode the instant the load rate begins to transition. Figure 7 As can be seen, the red curve reaches its highest value at the 20th second, indicating that the filter has entered full-spectrum coverage. At this time, the system uses maximum compensation power to cope with unpredictable interharmonic impacts. After the 40th second, the system detects that the load rate has stabilized, automatically switches to querying the steady-state harmonic sub-library and calling the adaptive narrowband compensation mode. The red curve drops significantly to about 0.5, indicating that the filter compensation resources are concentrated on the characteristic harmonic frequency band, achieving fine-grained control.

[0256] The black curve at the bottom represents the real-time change in the total harmonic distortion rate of the power grid. During the initial load transition, the curve spikes significantly to approximately 6%, which reflects the power quality degradation caused by interharmonic emergence under transient conditions. However, due to the timely activation of broadband compensation as a fallback measure, the distortion rate is quickly suppressed and falls back within ten seconds. After entering the steady-state phase, the black curve stabilizes at around 2% with slight fluctuations, indicating that the power quality is effectively maintained and meets national standards under the adaptive precision compensation strategy.

[0257] Overall, the graph further clarifies the smooth switching capability between the transient suppression stage and the steady-state suppression stage, reflecting the dual control objectives of the system in terms of both rapid fallback and long-term optimization during dynamic load changes. It also confirms the significant technical advantages of the predictive compensation strategy driven by the harmonic feature library in improving suppression response speed, reducing filter energy loss, and ensuring power quality stability.

[0258] In summary, this embodiment achieves proactive and precise management of harmonics in ship power grids, particularly those from power electronic loads, by constructing an intelligent harmonic feature library and performing dynamic querying and two-stage planning based on prediction. It not only significantly improves the efficiency and effectiveness of harmonic suppression and reduces the operating losses of active filters, but also effectively enhances the stability and power quality of the power grid when facing frequent changes in operating conditions through suppression strategies that adapt to dynamic load processes.

[0259] like Figure 8 As shown, this invention proposes an active harmonic suppression and energy management system for ship power grids, comprising:

[0260] The data acquisition module is configured to perform multi-dimensional data acquisition, collecting basic electrical parameters of the ship's electrical network, operating parameters of energy equipment, and transient characteristics of nonlinear loads; the transient characteristics of the nonlinear loads include at least the rate of change of the inverter's modulation frequency. and the derivative of the traction torque of the electric propulsion device ;

[0261] The adaptive prediction module, communicatively connected to the data acquisition module, is configured to predict the harmonic trends and energy demands of the ship's power grid based on the acquired multi-dimensional data using a deep learning hybrid model incorporating an attention mechanism, and to iteratively update the model parameters online according to the prediction error; wherein, the attention mechanism employs a correlation weighting function. Weighting of transient characteristics of nonlinear loads;

[0262] The coupled energy scheduling module, communicatively connected to the operating condition adaptive prediction module, is configured to monitor the device-level energy consumption of the active power filter (APF) in real time and calculate the harmonic-energy dynamic priority coefficient based on the energy demand prediction results. Based on this coefficient, a model predictive control algorithm is used to plan the power allocation of each energy device; wherein, the device-level energy consumption of the APF includes IGBT switching losses. and reactor copper loss ;

[0263] The collaborative harmonic suppression module, which is communicatively connected to the coupled energy dispatch module, is configured to coordinate the distributed master-slave APFs to suppress harmonics based on the power allocation results, and to execute a local pre-control strategy when a communication delay is detected. The system also includes a harmonic feature library, which stores typical harmonic sequence sets of inverters with different modulation methods and topologies under different load rates, for the collaborative harmonic suppression module to dynamically query and perform precise compensation for high-energy harmonic sequence sets based on the query results.

[0264] Example 5:

[0265] Corresponding to the above embodiments, the present invention also proposes an electronic device.

[0266] like Figure 9 The diagram shows a structural schematic of an electronic device according to the present invention. The electronic device 100 includes a processor 101 and a memory 103. The processor 101 and the memory 103 are connected, for example, via a bus 102. Optionally, the electronic device 100 may further include a transceiver 104. It should be noted that in practical applications, the transceiver 104 is not limited to one unit, and the structure of this electronic device 100 does not constitute a limitation on the embodiments of the present invention.

[0267] Processor 101 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 101 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0268] Bus 102 may include a pathway for transmitting information between the aforementioned components. Bus 102 may be a PCI bus or an EISA bus, etc. Bus 102 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0269] The memory 103 stores a computer program corresponding to a general page-turning data recursive query and processing method according to the above embodiments of the present invention. This computer program is controlled and executed by the processor 101. The processor 101 executes the computer program stored in the memory 103 to implement the content shown in the aforementioned method embodiments.

[0270] Among them, electronic devices 100 include, but are not limited to: mobile terminals such as laptops and PADs (tablet computers) and fixed terminals such as desktop computers. Figure 9 The electronic device 100 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0271] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for active harmonic suppression and energy management in ship electrical networks, characterized in that, Includes the following steps: S1. Collect basic electrical parameters of the ship's power grid, operating parameters of energy equipment, and transient characteristics of nonlinear loads; S2. Based on the collected multi-dimensional data, a deep learning hybrid model including an attention mechanism is used to predict the harmonic trend and energy demand of the ship's power grid, and the model parameters are updated online iteratively according to the prediction error. S3. Monitor the device-level energy consumption of the active power filter (APF) in real time, combine the predicted harmonic trends and energy demands, calculate the harmonic-energy dynamic priority coefficient, and use the model predictive control algorithm based on the coefficient to plan the power allocation of each energy device. The device-level power consumption of the active power filter (APF) includes IGBT switching losses. and reactor copper loss The calculation methods are as follows: ， ， in, This refers to the collector-emitter voltage of the IGBT. This refers to the collector current of the IGBT. The sampling period; This is the operating current of the reactor. The DC resistance of the reactor; S4. Based on the power allocation results, coordinate the distributed master-slave APFs to suppress harmonics and execute local pre-control strategies when communication delays are detected.

2. The method according to claim 1, characterized in that, In step S1, the transient characteristics of the nonlinear load include at least the rate of change of the inverter modulation frequency. and the derivative of the traction torque of the electric propulsion device ; Wherein, the frequency change rate of the inverter modulation frequency Defined as the difference between the current modulation frequency of the inverter and the modulation frequency of the previous sampling period divided by the sampling period; The derivative of the traction torque of the electric propulsion device Defined as the difference between the current traction torque of the electric propulsion device and the traction torque of the previous sampling period, divided by the sampling period.

3. The method according to claim 1, characterized in that, In step S2, the attention mechanism employs a correlation weighting function. The correlation weighting function weights the transient characteristics of the nonlinear load. for: ， in, For the first Association weights for transient characteristics of nonlinear loads. For the first The change in transient-like characteristics For the first The contribution coefficient of transient-like characteristics to harmonics. The number of categories representing the transient characteristics of nonlinear loads. For the first The change in transient-like characteristics For the first The contribution coefficient of transient-like characteristics to harmonics.

4. The method according to claim 1, characterized in that, In step S3, the harmonic-energy dynamic priority coefficient The calculation formula is: ， in, This represents the actual total harmonic distortion (THD). This is the upper limit of the national standard. This represents the rated state of charge of the energy storage unit. This represents the actual state of charge. This represents actual fuel consumption. For economical operation and fuel consumption; , , These are harmonic weights, SOC deviation weights, and fuel consumption weights, respectively.

5. The method according to claim 1, characterized in that, In step S4, the execution of the local pre-control strategy includes: When the communication delay between the main control unit and the APF exceeds a preset threshold, the APF initiates local control: IGBT switching losses based on current APF and the actual state of charge of the energy storage unit The compensation frequency response bandwidth of the APF is dynamically adjusted. when Below the threshold or When the frequency response bandwidth is below the threshold, the compensation frequency response bandwidth is reduced to prioritize the suppression of low-frequency and high-energy characteristic harmonics. when Below the threshold and When within the rated range, the frequency response bandwidth is amplified to perform broad-spectrum harmonic suppression.

6. The method according to claim 1, characterized in that, In step S4, the harmonic suppression performed by the coordinated distributed deployment of the master-slave APF includes: The main APF receives compensation feedback signals from the APFs in each region and the total harmonic distortion (THD) of the region in real time. ; If a certain area is detected If the temperature remains above the preset limit, the main APF will issue instructions to the secondary APF in that area: Based on the equivalent grid impedance of the region where the APF is located Dynamically calculate the optimal compensation current amplitude for the suppression frequency band. and phase angle ; The compensation frequency response bandwidth from the APF is adjusted to focus on the harmonics near the impedance resonant frequency point in this region, where the impedance resonant frequency point refers to the equivalent grid impedance. The frequency point at which a peak value occurs at a specific harmonic frequency.

7. The method according to claim 1, characterized in that, The S4 step also includes: Establish a harmonic characteristic library containing inverters with different modulation methods and topologies, and store their typical harmonic order sets under different load rates. ; Based on the power demand and load forecast results for the next scheduling cycle, the harmonic feature library is dynamically queried to obtain the set of high-energy harmonic sequences most likely to occur under the current operating conditions. Based on the query results, the APF prioritizes using adaptive bandwidth to accurately compensate for the high-energy harmonic order set, while using fixed low bandwidth to suppress the background of other harmonics.

8. The method according to claim 7, characterized in that, The harmonic feature library is further divided into a steady-state harmonic sub-library and a transient harmonic sub-library; The steady-state harmonic sub-library is used to store the mapping relationship between load rate and steady-state characteristic harmonic order set; The transient harmonic sub-library is used to store the mapping relationship between the load gradient transitioning from the current load rate to the target load rate and the transient harmonic spectrum; When dynamically querying the harmonic characteristic library based on the power demand and load forecast results for the next scheduling cycle, a two-stage suppression planning process is executed, including: Transient suppression phase: When a change in load rate is detected, the transient harmonic sub-library is queried first based on the difference between the current load rate and the target load rate to obtain the frequency band of the transient interharmonics that will be generated. It also instructs the APF to immediately adopt a broadband compensation mode to cover the transient interharmonic frequency band; Steady-state suppression phase: After monitoring that the actual load rate has stabilized at the target load rate, the system automatically switches to querying the steady-state harmonic sub-library. The system instructs the APF to switch to adaptive bandwidth to accurately compensate for the steady-state high-energy harmonic sequence set. The steady-state high-energy harmonic sequence set refers to the set of characteristic harmonics with the highest amplitude at the corresponding load rate, which is obtained by statistical analysis of the steady-state harmonic sub-library.

9. A shipboard electrical grid harmonic active suppression and energy management system, characterized in that, The system is used to perform the active harmonic suppression and energy management method for ship electrical grids according to any one of claims 1-8, the system comprising: The data acquisition module is configured to perform multi-dimensional data acquisition, collecting basic electrical parameters of the ship's electrical network, operating parameters of energy equipment, and transient characteristics of nonlinear loads; the transient characteristics of the nonlinear loads include at least the rate of change of the frequency converter modulation frequency. and the derivative of the traction torque of the electric propulsion device ; The adaptive prediction module, communicatively connected to the data acquisition module, is configured to predict the harmonic trends and energy demands of the ship's power grid based on the acquired multi-dimensional data using a deep learning hybrid model incorporating an attention mechanism, and to iteratively update the model parameters online according to the prediction error; wherein, the attention mechanism employs a correlation weighting function. Weighting of transient characteristics of nonlinear loads; The coupled energy scheduling module, communicatively connected to the operating condition adaptive prediction module, is configured to monitor the device-level energy consumption of the active power filter (APF) in real time and calculate the harmonic-energy dynamic priority coefficient based on the energy demand prediction results. Based on this coefficient, a model predictive control algorithm is used to plan the power allocation of each energy device; wherein, the device-level energy consumption of the APF includes IGBT switching losses. and reactor copper loss ; The collaborative harmonic suppression module, which is communicatively connected to the coupled energy dispatch module, is configured to coordinate the distributed master-slave APFs to suppress harmonics based on the power allocation results, and to execute a local pre-control strategy when a communication delay is detected. The system also includes a harmonic feature library, which stores typical harmonic sequence sets of inverters with different modulation methods and topologies under different load rates, for the collaborative harmonic suppression module to dynamically query and perform precise compensation for high-energy harmonic sequence sets based on the query results.

Images