An advanced power control method suitable for marine photovoltaic power generation system

Abstract

The present invention discloses an advanced power control method suitable for a marine photovoltaic power generation system, which relates to the field of photovoltaic power control technology, and includes the following steps: collecting multi-dimensional data during the navigation process of a ship, and using the multi-dimensional data as input information; based on the collected multi-dimensional data, constructing a CNN-LSTM-AM ultra-short-term photovoltaic power prediction model. The present invention introduces the CNN-LSTM-AM ultra-short-term marine photovoltaic power prediction model, which can accurately predict the photovoltaic power generation power in the next second, and perform power allocation accordingly, thereby realizing accurate prediction and dynamic adjustment of photovoltaic power. By real-time monitoring of photovoltaic power generation power, battery SOC, and key parameters of DC bus voltage, the system can dynamically adjust the power control strategy to ensure stable tracking of the inverter output current, effectively avoiding the problems of power deviation and energy accumulation or insufficiency, and significantly improving the stability and reliability of the system.

Description

Technical Field

[0001] The invention relates to the technical field of photovoltaic power control, and in particular to an advanced power control method suitable for a marine photovoltaic power generation system. Background Art

[0002] In today's era of rapid development of green energy technology, marine photovoltaic power generation systems, as an important part of ship power supply, are increasingly receiving widespread attention and application. However, due to the particularity of the ship's operating environment, such as changes in light intensity, temperature fluctuations, wind speed changes, and the ship's own motion state (such as speed, heading, swing angle, etc.) during navigation, the output power of the marine photovoltaic power generation system is subject to great uncertainty. This uncertainty not only affects the stable operation of the ship's power system, but may also have an adverse effect on the ship's navigation safety and energy efficiency.

[0003] As an energy storage device for photovoltaic power generation systems, the charging and discharging performance of batteries is also crucial to the stable operation of the system. However, the charging and discharging response time of batteries is usually longer than the inertia time of photovoltaic fluctuations, which makes it easy for batteries to produce power deviations when compensating for photovoltaic grid-connected fluctuating power. When the photovoltaic penetration rate is large, this power deviation will cause energy accumulation or shortage in the photovoltaic grid-connected system, causing the DC bus voltage to be in a dynamic fluctuation state, seriously affecting the stable operation of the system. Therefore, an advanced power control method suitable for marine photovoltaic power generation systems is proposed. Summary of the invention

[0004] The present invention aims to provide an advanced power control method applicable to a marine photovoltaic power generation system to solve the problems raised in the above-mentioned background technology.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] An advanced power control method applicable to a marine photovoltaic power generation system comprises the following steps:

[0007] Step 1: Collect multi-dimensional data during the ship's voyage, including speed, heading, light intensity, temperature, wind speed, ship position and swing angle, and use the multi-dimensional data as input information for subsequent photovoltaic power prediction. Ensure the accuracy and consistency of the data by cleaning, sorting and standardizing these data;

[0008] Step 2: Based on the collected multi-dimensional data, a CNN-LSTM-AM (convolutional neural network-long short-term memory network-attention mechanism) ultra-short-term photovoltaic power prediction model is constructed, where CNN is used to extract the spatial features of the data, LSTM is used to capture the long-term dependencies of the time series, and AM is used to enhance the model's attention to key information;

[0009] Step 3: Use the constructed ultra-short-term photovoltaic power prediction model to predict the power of the marine photovoltaic power generation system in the next second, and input the prediction results into the energy management system. Combined with the battery state of charge (SOC), the photovoltaic power is allocated to determine the optimal grid-connected power. And battery compensation power ;

[0010] Step 4: Set the battery charging and discharging strategy according to the power allocation result of the energy management system;

[0011] Step 5: Use a dual-loop power control strategy to accurately control the power output of the inverter;

[0012] Step 6: Monitor the key parameters of photovoltaic power generation, battery SOC and DC bus voltage in real time, and dynamically adjust the power control strategy based on the monitoring data to cope with the real-time changes in photovoltaic power generation and system requirements.

[0013] A further improvement of the technical solution of the present invention is that in step 1, the process of collecting multi-dimensional data is as follows:

[0014] Deploy and calibrate sensors and monitoring equipment on board ships to comprehensively collect multi-dimensional data on ship speed, heading, light intensity, temperature, wind speed, ship position and sway angle during navigation, and ensure that all sensors and monitoring equipment are calibrated to ensure the original accuracy of the data;

[0015] Clean the collected multi-dimensional data, check for missing values, outliers and erroneous data in the data set, and process them. Missing values ​​can be filled by interpolation methods, outliers and erroneous data need to be deleted or corrected, and noise data should be identified and removed.

[0016] Organize the cleaned data, sort the collected data by time series to ensure the time sequence of the data, classify and mark the data, classify the data of different speeds and directions separately, and ensure that each data point contains a timestamp and other necessary metadata for subsequent analysis and processing;

[0017] The sorted data is standardized to unify the dimension and scope of the data to facilitate subsequent data analysis and comparison. The multi-dimensional data after aggregation processing is obtained to obtain a multi-dimensional ship data sequence table and stored in the data warehouse.

[0018] A further improvement of the technical solution of the present invention is that in step 2, the construction process of the ultra-short-term photovoltaic power prediction model is:

[0019] The multi-dimensional data sequence table after cleaning, sorting and standardization is used as the data set of input information, and the data set is divided into training set and test set to ensure the effective training and verification of the model;

[0020] Design the architecture of the CNN-LSTM-AM (Convolutional Neural Network-Long Short-Term Memory Network-Attention Mechanism) model, where CNN is used to extract spatial features of data, LSTM is used to capture long-term dependencies of time series, and AM is used to enhance the model's attention to key information;

[0021] Receive multi-dimensional data input, use multiple convolution kernels in the convolution layer of CNN (convolutional neural network) to extract the spatial features of the data, and downsample the output of the convolution layer in the pooling layer to reduce the dimension of the data while retaining important features. In the fully connected layer, map the convolution and pooling features to a fixed-size vector to provide input for the subsequent LSTM layer;

[0022] The input gate of LSTM (Long Short-Term Memory Network) is used to control the inflow of input information, and the forget gate is used to determine the information to be forgotten, that is, which information is discarded from the state of the previous time step. The output gate is used to determine the output information based on the input of the current time step and the hidden state of the previous time step. The cell state is updated through the operation of the input gate, forget gate and output gate to pass information in the entire sequence. Multiple LSTM units are stacked in the LSTM layer to capture long-term dependencies in time series data, such as the potential impact of navigation status such as ship speed and heading on photovoltaic power generation;

[0023] AM (attention mechanism) is introduced to dynamically adjust the contribution of different input features to the prediction results. According to the output of the LSTM layer, the attention score of each time step is calculated, and the output of the LSTM layer is weighted and summed according to the attention score to obtain the final context vector. The context vector is combined with the final output of the LSTM layer to enhance the model's attention to key information.

[0024] The CNN-LSTM-AM model is trained using relevant data from the training set. The prediction error is minimized by continuously adjusting the model parameters. During the training process, the generalization ability of the model is evaluated based on relevant data from the test set combined with the cross-validation method to avoid overfitting.

[0025] A further improvement of the technical solution of the present invention is that in step 3, the process of determining the optimal grid-connected power and the battery compensation power is as follows:

[0026] After processing the multi-dimensional data at the current moment, input it into the trained CNN-LSTM-AM model to perform ultra-short-term photovoltaic power forecasting and output the photovoltaic power forecast value for the next second;

[0027] The photovoltaic power prediction value for the next second predicted by the CNN-LSTM-AM model Input into the energy management system as the predicted value of photovoltaic power generation and check the current state of charge (SOC) of the battery;

[0028] With the goal of keeping the maximum fluctuation rate of photovoltaic grid-connected power lower than 10% of the rated power, the photovoltaic power is reasonably allocated and the optimal grid-connected power is calculated based on the predicted value of photovoltaic power generation and the state of charge (SOC) of the battery. And battery compensation power ;

[0029] Clearly define the battery charging and discharging requirements. When the power drops instantly, the battery needs to discharge the photovoltaic inverter power at a slope of 10% of the rated power, so that the photovoltaic inverter power drops smoothly to 0. During instantaneous increase, the battery needs to charge the photovoltaic inverter power at a slope of 10% of the rated power, so that the photovoltaic inverter power can be smoothly increased from 0 to the maximum power value;

[0030] According to the calculated optimal grid-connected power And battery compensation power , the energy management system generates corresponding control instructions.

[0031] A further improvement of the technical solution of the present invention is that: the optimal grid-connected power The calculation expression is:

[0032] ;

[0033] in, For the optimal grid-connected power, is the rated power of the photovoltaic power generation system, is the predicted value of photovoltaic power generation, is the maximum allowable fluctuation rate of photovoltaic grid-connected power, , The value range is ;

[0034] The battery compensation power The calculation expression is:

[0035] ;

[0036] in, is the battery compensation power, which means the photovoltaic power part used to compensate for grid power fluctuations and meet the battery charging and discharging needs. is the predicted value of photovoltaic power generation, For the optimal grid-connected power, The value range is .

[0037] A further improvement of the technical solution of the present invention is that in step 4, the setting process of the battery charging and discharging strategy is:

[0038] Obtain the grid-connected power demand at the current moment through the energy management system , and the predicted value of photovoltaic power generation Grid-connected power demand Make comparisons;

[0039] According to the comparison results, the charge and discharge strategy is set. Greater than , the photovoltaic power is higher than the grid-connected power demand, and the battery should be charged. Less than , the photovoltaic power is lower than the grid-connected power demand, and the battery should be discharged. The discharge rate should be controlled within the acceptable range of the battery to ensure that it can continuously and stably provide power. During the discharge process, the battery parameters such as SOC and temperature should be closely monitored to prevent over-discharge or overheating;

[0040] Combined with the charging and discharging strategies, the charging power is calculated separately and discharge power , ensure that the charging and discharging power does not exceed the maximum charging and discharging current of the battery , , ,in, is the nominal voltage of the battery;

[0041] Based on the calculated charge and discharge power, a control instruction is generated and sent to the battery management system. The control instruction includes the power value and duration parameters of the charge or discharge.

[0042] A further improvement of the technical solution of the present invention is that: the charging power The calculation expression is:

[0043] ;

[0044] in, is the charging power, is the predicted value of photovoltaic power generation, The grid-connected power requirement is is the maximum charging power of the battery, SOC is the state of charge;

[0045] The discharge power The calculation expression is:

[0046] ;

[0047] in, is the discharge power, is the maximum discharge power of the battery.

[0048] A further improvement of the technical solution of the present invention is that in step 5, the process of controlling the inverter power output by the dual-loop power control strategy is:

[0049] Set the reference value of the inverter output current according to the system power demand and voltage conditions , and measure the inverter output current in real time , calculate the current error signal , that is, the difference between the reference current and the actual current, ;

[0050] Use PI (proportional-integral) controller to output modulation signal D, control the switching state of the inverter, and thus adjust the output current. According to the modulation signal D, adjust the switching state of the inverter to change the output current so that it tracks the reference current, continuously feed back the actual current value, and compare it with the reference value to achieve precise control. , where D is the modulating signal, is the proportional gain, is the integral gain, is the error signal of the current;

[0051] According to the voltage requirements of the system, set the reference value of the inverter output voltage , real-time measurement of the inverter output voltage , and calculate the voltage error signal , that is, the difference between the reference voltage and the actual voltage, ;

[0052] Use PI (proportional-integral) controller to output reference power , and then adjust the power, thereby adjusting the voltage, and according to the reference power, adjust the output power of the inverter to change the energy balance of the system, and continuously feedback the actual power value and compare it with the reference value to achieve stable control;

[0053] The reference current of the inner loop control and the reference power of the outer loop control are sent to the inverter as control signals. The inverter adjusts its working state according to the control signals to achieve accurate power output.

[0054] A further improvement of the technical solution of the present invention is that the calculation formula for adjusting the output current is:

[0055] ;

[0056] in, is the reference current, is the proportional gain, is the integral gain, is the reference voltage, is the output voltage of the inverter, is the integral term of the error, which is used to eliminate the steady-state error;

[0057] The calculation formula of the adjustment power is:

[0058] ;

[0059] in, is the reference power, used for outer loop control, is the proportional gain, is the integral gain, To set the power, i.e. the energy demand of the system, is the actual output power.

[0060] A further improvement of the technical solution of the present invention is that in step 6, the process of dynamically adjusting the power control strategy is:

[0061] Through sensors and monitoring systems, key parameters such as photovoltaic power generation, battery SOC (state of charge) and DC bus voltage are acquired in real time, and the collected data is transmitted to the energy management system for analysis and processing;

[0062] Analyze monitoring data, identify the real-time change trend of photovoltaic power generation, changes in battery SOC, and stability of DC bus voltage, and predict future changes in photovoltaic power generation, evaluate whether the current power control strategy can meet system requirements and whether there are potential unstable factors;

[0063] According to the data analysis and prediction results, the parameters of the inner and outer loop control strategies are dynamically adjusted, and the control signals of the inverter and battery bidirectional rectifier are adjusted to adapt to the changes in photovoltaic power generation and system requirements;

[0064] Send the adjusted control instructions to the inverter and battery management system to execute the new power control strategy, and monitor the performance of the system after the new control strategy is implemented, including the changes in photovoltaic power generation, battery SOC and DC bus voltage, to evaluate whether the adjusted control strategy is effective and whether it meets the requirements of system stability and performance;

[0065] Feedback the monitoring results to the data analysis stage to form a closed-loop control process, continuously optimize the control strategy, and record the history of all monitoring data and control strategy adjustments for subsequent analysis and optimization.

[0066] Due to the adoption of the above technical solution, the present invention has the following technical advances compared with the prior art:

[0067] The present invention provides an advanced power control method suitable for a marine photovoltaic power generation system. By introducing a CNN-LSTM-AM ultra-short-term marine photovoltaic power prediction model, the photovoltaic power generation power in the next second can be accurately predicted, and power allocation is performed accordingly, thereby realizing accurate prediction and dynamic adjustment of photovoltaic power. By real-time monitoring of key parameters of photovoltaic power generation power, battery SOC and DC bus voltage, the system can dynamically adjust the power control strategy to ensure stable tracking of the inverter output current, effectively avoid the problems of power deviation and energy accumulation or shortage, and significantly improve the stability and reliability of the system.

[0068] The present invention provides an advanced power control method suitable for a marine photovoltaic power generation system. The photovoltaic power is reasonably distributed through an energy management system, and the optimal matching of the grid-connected power PT and the battery compensation power PB is achieved. It not only ensures that the maximum fluctuation rate of the photovoltaic grid-connected power is lower than 10% of the rated power, but also fully considers the state of charge (SOC) of the battery, and ensures that the photovoltaic inverter power can be smoothly adjusted at a slope of 10% of the rated power when the photovoltaic power generation power drops or increases instantly, which not only improves the energy utilization efficiency, but also prolongs the service life of the battery. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0070] Figure 1 is a flow chart of the method of the present invention;

[0071] Figure 2 This is a flow chart of controlling the inverter power output using the dual-loop power control strategy of the present invention. DETAILED DESCRIPTION

[0072] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0073] Embodiment 1, as Figure 1As shown, the present invention provides an advanced power control method applicable to a marine photovoltaic power generation system, comprising the following steps:

[0074] Step 1: Collect multi-dimensional data during the ship's voyage, including speed, heading, light intensity, temperature, wind speed, ship position and swing angle, and use the multi-dimensional data as input information for subsequent photovoltaic power prediction. Ensure the accuracy and consistency of the data by cleaning, sorting and standardizing these data, deploy and calibrate sensors and monitoring equipment on the ship, and comprehensively collect multi-dimensional data on speed, heading, light intensity, temperature, wind speed, ship position and swing angle during the ship's voyage. Ensure that all sensors and monitoring equipment are calibrated to ensure the original accuracy of the data, clean the collected multi-dimensional data, check for missing values ​​in the data set, Outliers and erroneous data are detected and processed. Missing values ​​can be filled by interpolation methods. Outliers and erroneous data need to be deleted or corrected. Noise data is identified and removed. The cleaned data is sorted. The collected data is sorted according to time series to ensure the time series of the data. The data is classified and marked. The data of different speeds and headings are classified separately to ensure that each data point contains a timestamp and other necessary metadata for subsequent analysis and processing. The sorted data is standardized to unify the dimension and range of the data for subsequent data analysis and comparison. The multi-dimensional data after aggregation processing is obtained to obtain a multi-dimensional ship data sequence table and stored in the data warehouse;

[0075] Step 2. Based on the collected multi-dimensional data, an ultra-short-term photovoltaic power prediction model of CNN-LSTM-AM (convolutional neural network-long short-term memory network-attention mechanism) is constructed, in which CNN is used to extract the spatial features of the data, LSTM is used to capture the long-term dependencies of the time series, and AM is used to enhance the model's attention to key information. The multi-dimensional data sequence table after cleaning, sorting and standardization is used as the data set of input information, and the data set is divided into a training set and a test set to ensure effective training and verification of the model. The architecture of the CNN-LSTM-AM (convolutional neural network-long short-term memory network-attention mechanism) model is designed, in which CNN is used to extract the spatial features of the data, LSTM is used to capture the long-term dependencies of the time series, and AM is used to enhance the model's attention to key information. The input of multi-dimensional data is received, and multiple convolution kernels in the convolution layer of CNN (convolutional neural network) are used to extract the spatial features of the data, and the output of the convolution layer is downsampled in the pooling layer to reduce the dimension of the data while retaining important features. The features after convolution and pooling are mapped to a fixed-size vector in the fully connected layer to provide input for the subsequent LSTM layer. Through LSTM The input gate of the (long short-term memory network) controls the inflow of input information, and the forget gate is used to determine the information to be forgotten, that is, which information is discarded from the state of the previous time step. The output gate is used to determine the output information based on the input of the current time step and the hidden state of the previous time step, and the cell state is used to update the operation of the input gate, forget gate and output gate to pass information in the entire sequence. Multiple LSTM units are stacked in the LSTM layer to capture long-term dependencies in time series data, such as the potential impact of navigation states such as ship speed and heading on photovoltaic power generation. AM (attention mechanism) is introduced to dynamically adjust the contribution of different input features to the prediction results. According to the output of the LSTM layer, the attention score of each time step is calculated, and the output of the LSTM layer is weighted and summed according to the attention score to obtain the final context vector. The context vector is combined with the final output of the LSTM layer to enhance the model's attention to key information. The CNN-LSTM-AM model is trained using the relevant data of the training set, and the prediction error is minimized by continuously adjusting the model parameters. During the training process, the generalization ability of the model is evaluated based on the relevant data of the test set combined with the cross-validation method to avoid overfitting;

[0076] Step 3: Use the constructed ultra-short-term photovoltaic power prediction model to predict the power of the marine photovoltaic power generation system in the next second, and input the prediction results into the energy management system. Combined with the battery state of charge (SOC), the photovoltaic power is allocated to determine the optimal grid-connected power. And battery compensation power , after processing the multi-dimensional data at the current moment, input it into the trained CNN-LSTM-AM model to perform ultra-short-term photovoltaic power forecasting, output the photovoltaic power forecast value for the next second, and output the photovoltaic power forecast value for the next second predicted by the CNN-LSTM-AM model. Input into the energy management system as the predicted value of photovoltaic power generation, and check the current state of charge (SOC) of the battery. With the maximum fluctuation rate of photovoltaic grid-connected power lower than 10% of the rated power as the goal, combine the predicted value of photovoltaic power generation and the state of charge (SOC) of the battery to reasonably allocate photovoltaic power and calculate the optimal grid-connected power. And battery compensation power , where the optimal grid-connected power That is, the photovoltaic power part that meets the grid connection requirements and has a fluctuation rate of less than 10%, and the battery compensation power That is, the photovoltaic power part used to compensate for grid-connected power fluctuations and meet the battery charging and discharging requirements. The SOC value range is , where 1 indicates a fully charged state and 0 indicates a fully discharged state, clearly indicating the battery charging and discharging requirements. When the power drops instantly, the battery needs to discharge the photovoltaic inverter power at a slope of 10% of the rated power, so that the photovoltaic inverter power drops smoothly to 0. During instantaneous boost, the battery needs to charge the photovoltaic inverter power at a slope of 10% of the rated power, so that the photovoltaic inverter power rises smoothly from 0 to the maximum power value. And battery compensation power , the energy management system generates corresponding control instructions;

[0077] Furthermore, the optimal grid-connected power The calculation expression is:

[0078] ;

[0079] in, For the optimal grid-connected power, is the rated power of the photovoltaic power generation system, is the predicted value of photovoltaic power generation, is the maximum allowable fluctuation rate of photovoltaic grid-connected power, , The value range is ;

[0080] Battery compensation power The calculation expression is:

[0081] ;

[0082] in, is the battery compensation power, which means the photovoltaic power part used to compensate for grid power fluctuations and meet the battery charging and discharging needs. is the predicted value of photovoltaic power generation, For the optimal grid-connected power, The value range is ;

[0083] Step 4: According to the power allocation results of the energy management system, set the battery charging and discharging strategy. When the photovoltaic power is higher than the grid-connected demand, the battery is charged. When the photovoltaic power is lower than the grid-connected demand, the battery is discharged. At the same time, it is necessary to ensure that the battery charging and discharging process meets its physical properties and safety requirements. The energy management system obtains the grid-connected power demand at the current moment. , and the predicted value of photovoltaic power generation Grid-connected power demand Compare and set the charge and discharge strategy according to the comparison result. Greater than , the photovoltaic power is higher than the grid-connected power demand, the battery should be charged, and the charging rate should be controlled within the acceptable range of the battery to avoid overheating or damage. Constant current charging, constant voltage charging, or a combination of constant current and constant voltage charging can be used, depending on the type and characteristics of the battery. Less than , the photovoltaic power is lower than the grid-connected power demand, the battery should be discharged, and the discharge rate should be controlled within the acceptable range of the battery to ensure that it can continuously and stably provide power. During the discharge process, the battery SOC and temperature and other parameters should be closely monitored to prevent over-discharge or overheating. Combined with the charging and discharging strategies, the charging power is calculated separately. and discharge power , ensure that the charging and discharging power does not exceed the maximum charging and discharging current of the battery , , ,in, is the nominal voltage of the battery and ensures that the battery state of charge (SOC) is within a safe range, usually If the charging or discharging power exceeds the physical limit of the battery, the charging or discharging power needs to be adjusted to ensure that the maximum charging or discharging current is not exceeded. If the state of charge (SOC) of the battery is close to the limit of the safety range, the charging and discharging strategy needs to be adjusted to avoid overcharging or over-discharging. Based on the calculated charging and discharging power, a control instruction is generated and sent to the battery management system. The control instruction includes the charging or discharging power value and duration parameters;

[0084] Furthermore, the charging power The calculation expression is:

[0085] ;

[0086] in, is the charging power, is the predicted value of photovoltaic power generation, The grid-connected power requirement is is the maximum charging power of the battery, SOC is the state of charge;

[0087] Discharge power The calculation expression is:

[0088] ;

[0089] in, is the discharge power, is the maximum discharge power of the battery;

[0090] Step 5: Use a dual-loop power control strategy to accurately control the power output of the inverter. The inner loop control is used to achieve fast response and precise regulation, and the outer loop control is used to maintain the overall stability and performance of the system. Through the implementation of the dual-loop power control strategy, it is ensured that the marine photovoltaic power generation system can operate stably under different working conditions.

[0091] Step 6: Monitor the key parameters of photovoltaic power generation, battery SOC and DC bus voltage in real time, and dynamically adjust the power control strategy based on the monitoring data to cope with the real-time changes in photovoltaic power generation and system requirements.

[0092] Embodiment 2, as Figure 2 As shown, based on Example 1, the present invention provides a technical solution: Preferably, in step 5, the process of controlling the inverter power output by the dual-loop power control strategy is:

[0093] According to the power demand and voltage conditions of the system, set the reference value of the inverter output current , and measure the inverter output current in real time , calculate the current error signal , that is, the difference between the reference current and the actual current, , use the PI (proportional-integral) controller to output the modulation signal D, control the switching state of the inverter, and thus adjust the output current. According to the modulation signal D, adjust the switching state of the inverter to change the output current so that it tracks the reference current, and continuously feed back the actual current value and compare it with the reference value to achieve precise control, , where D is the modulating signal, is the proportional gain, is the integral gain, It is the error signal of the current. According to the voltage requirement of the system, it sets the reference value of the inverter output voltage. , real-time measurement of the inverter output voltage , and calculate the voltage error signal , that is, the difference between the reference voltage and the actual voltage, , using PI (proportional-integral) controller to output reference power , and then adjust the power, thereby adjusting the voltage, and adjust the output power of the inverter according to the reference power to change the energy balance of the system, and continuously feed back the actual power value and compare it with the reference value to achieve stable control. The reference current of the inner loop control and the reference power of the outer loop control are sent to the inverter as control signals. The inverter adjusts its working state according to the control signal to achieve accurate power output;

[0094] The formula for adjusting the output current is:

[0095] ;

[0096] in, is the reference current, is the proportional gain, is the integral gain, is the reference voltage, is the output voltage of the inverter, is the integral term of the error, which is used to eliminate the steady-state error;

[0097] The formula for calculating the regulated power is:

[0098] ;

[0099] in, is the reference power, used for outer loop control, is the proportional gain, is the integral gain, To set the power, i.e. the energy demand of the system, is the actual output power;

[0100] In step 6, the process of dynamically adjusting the power control strategy is as follows:

[0101] Through sensors and monitoring systems, key parameters of photovoltaic power generation, battery SOC (state of charge) and DC bus voltage are obtained in real time, and the collected data is transmitted to the energy management system for analysis and processing. The monitoring data is analyzed to identify the real-time change trend of photovoltaic power generation, the change of battery SOC and the stability of DC bus voltage, and predict the future change of photovoltaic power generation. It is evaluated whether the current power control strategy can meet the system requirements and whether there are potential unstable factors. According to the data analysis and prediction results, the parameters of the inner and outer loop control strategies are dynamically adjusted, and the control signals of the inverter and battery bidirectional rectifier are adjusted to adapt to the changes in photovoltaic power generation and system requirements. The adjusted control instructions are sent to the inverter and battery management system to execute the new power control strategy, and the performance of the system after the execution of the new control strategy is monitored, including the changes in photovoltaic power generation, battery SOC and DC bus voltage. It is evaluated whether the adjusted control strategy is effective and whether it meets the requirements of system stability and performance. The monitoring effect is fed back to the data analysis stage to form a closed-loop control process, continuously optimize the control strategy, and record all monitoring data and the history of control strategy adjustments for subsequent analysis and optimization.

[0102] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An advanced power control method applicable to a marine photovoltaic power generation system, characterized in that: The following steps are involved: Step 1: Collect multi-dimensional data during the ship's navigation process, and use the multi-dimensional data as input information. In step 1, the multi-dimensional data collection process is as follows: Deploy and calibrate sensors and monitoring equipment on ships to comprehensively collect multi-dimensional data on ship speed, heading, light intensity, temperature, wind speed, ship position and sway angle during navigation; Clean the collected multi-dimensional data, check for missing values, outliers, and erroneous data in the data set, and process them; Arrange the cleaned data, sort the collected data according to time series, classify and mark the data, and classify the data of different speeds and directions; Standardize the sorted data, unify the dimension and range of the data, aggregate the processed multi-dimensional data, obtain a multi-dimensional ship data sequence table, and store it in the data warehouse; Step 2: Based on the collected multi-dimensional data, an ultra-short-term photovoltaic power prediction model of CNN-LSTM-AM is constructed. In step 2, the construction process of the ultra-short-term photovoltaic power prediction model is as follows: The multi-dimensional data sequence table after cleaning, sorting and standardization is used as the data set of input information, and the data set is divided into a training set and a test set; Design the architecture of the CNN-LSTM-AM model, where CNN is used to extract spatial features of data, LSTM is used to capture long-term dependencies of time series, and AM is used to enhance the model's focus on key information; Receive multi-dimensional data input, use multiple convolution kernels in the convolution layer of CNN to extract the spatial features of the data, downsample the output of the convolution layer in the pooling layer, and map the convolution and pooling features to a fixed-size vector in the fully connected layer; The input gate of LSTM is used to control the inflow of input information, the forget gate is used to determine the information to be forgotten, and the output gate is used to determine the output information based on the input of the current time step and the hidden state of the previous time step. The cell state is updated through the operation of the input gate, forget gate, and output gate to pass information throughout the sequence. Multiple LSTM units are stacked in the LSTM layer to capture long-term dependencies in time series data. AM is introduced to dynamically adjust the contribution of different input features to the prediction results. According to the output of the LSTM layer, the attention score of each time step is calculated. According to the attention score, the output of the LSTM layer is weighted and summed to obtain the final context vector. The context vector is combined with the final output of the LSTM layer to enhance the model's attention to key information. The CNN-LSTM-AM model is trained using the relevant data of the training set. During the training process, the generalization ability of the model is evaluated based on the relevant data of the test set combined with the cross-validation method. Step 3: Use the constructed ultra-short-term photovoltaic power prediction model to predict the power of the marine photovoltaic power generation system in the next second, and input the prediction results into the energy management system. Combined with the battery charge state, the photovoltaic power is allocated to determine the optimal grid-connected power P. T And battery compensation power P B ; Step 4: Set the battery charging and discharging strategy according to the power allocation result of the energy management system; Step 5: Use a dual-loop power control strategy to accurately control the power output of the inverter; Step 6: Monitor the key parameters of photovoltaic power generation, battery SOC and DC bus voltage in real time, and dynamically adjust the power control strategy based on the monitoring data.

2. The advanced power control method applicable to a marine photovoltaic power generation system according to claim 1, characterized in that: In step 3, the process of determining the optimal grid-connected power and battery compensation power is as follows: After processing the multi-dimensional data at the current moment, input it into the trained CNN-LSTM-AM model to perform ultra-short-term photovoltaic power forecasting and output the photovoltaic power forecast value for the next second; The photovoltaic power prediction value P for the next second predicted by the CNN-LSTM-AM model P Input into the energy management system as the predicted value of photovoltaic power generation and check the current state of charge of the battery; With the goal of keeping the maximum fluctuation rate of photovoltaic grid-connected power lower than 10% of the rated power, the photovoltaic power is allocated based on the predicted value of photovoltaic power generation and the charge state of the battery, and the optimal grid-connected power P is calculated. T And battery compensation power P B ; Clearly define the battery charging and discharging requirements. P When the power drops instantly, the battery needs to discharge the photovoltaic inverter power at a slope of 10% of the rated power, so that the photovoltaic inverter power drops smoothly to 0. P During instantaneous increase, the battery needs to charge the photovoltaic inverter power at a slope of 10% of the rated power, so that the photovoltaic inverter power can be smoothly increased from 0 to the maximum power value; According to the calculated optimal grid-connected power P T And battery compensation power P B , the energy management system generates corresponding control instructions.

3. The advanced power control method applicable to a marine photovoltaic power generation system according to claim 2, characterized in that: The optimal grid-connected power P T The calculation expression is: P T =min(P P ,P rated ×(1-θ)); Among them, P T is the optimal grid-connected power, P rated is the rated power of the photovoltaic power generation system, P P is the predicted value of photovoltaic power generation, θ is the maximum allowable fluctuation rate of photovoltaic grid-connected power, θ<0.1, P T The value range is [0, P rated ×(1-θ]; The battery compensation power P B The calculation expression is: P B =P P -P T ; Among them, P B is the battery compensation power, which represents the photovoltaic power part used to compensate for grid power fluctuations and meet the battery charging and discharging requirements, P P is the predicted value of photovoltaic power generation, P T is the optimal grid-connected power, P P The value range is [0, P P ].

4. The advanced power control method applicable to a marine photovoltaic power generation system according to claim 3, characterized in that: In step 4, the process of setting the battery charging and discharging strategy is as follows: Obtain the grid-connected power demand P at the current moment through the energy management system grid , and the predicted value of photovoltaic power generation P P and the grid-connected power demand P grid Make comparisons; According to the comparison results, the charge and discharge strategy is set. P Greater than P grid , the photovoltaic power is higher than the grid-connected power demand, and the battery should be charged. P Less than P grid , the photovoltaic power is lower than the grid-connected power demand, and the battery should be discharged. The discharge rate is controlled within the acceptable range of the battery to ensure that it continues to provide power. During the discharge process, the SOC and temperature parameters of the battery are monitored to prevent over-discharge or overheating; Combined with the charging and discharging strategies, the charging power P is calculated respectively. charge and discharge power P discharge , ensure that the charging and discharging power does not exceed the maximum charging and discharging current I of the battery max , P charge ≤V battery ×I max , P discharge ≤V battery ×I max , where V battery is the nominal voltage of the battery; Based on the calculated charge and discharge power, a control instruction is generated and sent to the battery management system. The control instruction includes the power value and duration parameters of the charge or discharge.

5. The advanced power control method applicable to a marine photovoltaic power generation system according to claim 4, characterized in that: The charging power P charge The calculation expression is: P charge =min(P P -P grid ,SOC×C max -P grid ); Among them, P charge is the charging power, P P is the predicted value of photovoltaic power generation, P grid is the grid-connected power demand, C max is the maximum charging power of the battery, SOC is the state of charge; The discharge power P discharge The calculation expression is: Among them, P discharge is the discharge power, D max is the maximum discharge power of the battery.

6. The advanced power control method applicable to a marine photovoltaic power generation system according to claim 5, characterized in that: In step 5, the process of controlling the inverter power output by the dual-loop power control strategy is as follows: According to the system power demand and voltage conditions, set the reference value I of the inverter output current ref , and measure the inverter output current I in real time out , calculate the current error signal I error , that is, the difference between the reference current and the actual current, I error =I ref -I out ; The PI controller is used to output the modulation signal D to control the switching state of the inverter, thereby adjusting the output current. According to the modulation signal D, the switching state of the inverter is adjusted to change the output current so that it tracks the reference current. The actual current value is continuously fed back and compared with the reference value to achieve control. D = K p I error +K i ∫I error dt, where D is the modulation signal, K p is the proportional gain, K i is the integral gain, I error is the error signal of the current; According to the voltage requirements of the system, set the reference value V of the inverter output voltage ref , real-time measurement of the inverter output voltage V out , and calculate the voltage error signal V error , that is, the difference between the reference voltage and the actual voltage, V error =V ref -V out ; Use PI controller to output reference power P ref , and then adjust the power, thereby adjusting the voltage, and according to the reference power, adjust the output power of the inverter to change the energy balance of the system, and continuously feedback the actual power value and compare it with the reference value to achieve stable control; The reference current of the inner loop control and the reference power of the outer loop control are sent to the inverter as control signals. The inverter adjusts its working state according to the control signals to achieve accurate power output.

7. The advanced power control method applicable to a marine photovoltaic power generation system according to claim 6, characterized in that: The calculation formula for adjusting the output current is: I ref =K p (V ref -V out )+K i ∫(V ref -V out )dt; Among them, I ref is the reference current, K p is the proportional gain, K i is the integral gain, V ref is the reference voltage, V out is the output voltage of the inverter, ∫(V ref -V out )dt is the integral term of the error, which is used to eliminate the steady-state error; The calculation formula of the adjustment power is: Among them, P ref is the reference power, used for outer loop control, is the proportional gain, is the integral gain, P set To set the power, i.e. the energy demand of the system, P out is the actual output power.

8. The advanced power control method applicable to a marine photovoltaic power generation system according to claim 7, characterized in that: In step 6, the process of dynamically adjusting the power control strategy is: Through sensors and monitoring systems, key parameters of photovoltaic power generation, battery SOC and DC bus voltage are obtained in real time, and the collected data is transmitted to the energy management system; Analyze monitoring data, identify the real-time change trend of photovoltaic power generation, changes in battery SOC, and stability of DC bus voltage, and predict future changes in photovoltaic power generation, evaluate whether the current power control strategy can meet system requirements and whether there are potential unstable factors; According to the data analysis and prediction results, the parameters of the inner and outer loop control strategies are dynamically adjusted, and the control signals of the inverter and battery bidirectional rectifier are adjusted to adapt to the changes in photovoltaic power generation and system requirements; Send the adjusted control instructions to the inverter and battery management system to execute the new power control strategy, and monitor the performance of the system after the new control strategy is implemented, including the changes in photovoltaic power generation, battery SOC and DC bus voltage, to evaluate whether the adjusted control strategy is effective and whether it meets the requirements of system stability and performance; Feedback the monitoring results to the data analysis stage to form a closed-loop control process, continuously optimize the control strategy, and record the history of all monitoring data and control strategy adjustments.

Images