Intelligent energy-saving control method and system for ship propulsion system

Abstract

The invention discloses an intelligent energy-saving control method and system for a ship propulsion system. The method comprises the steps that S1, operation parameters of the ship propulsion system, ship navigation state parameters and marine environment parameters are collected; s2, preprocessing each parameter, and constructing a multi-dimensional state feature set; s3, evaluating the operation efficiency of the propulsion system based on the gradient lifting-neural network composite model, and determining an energy-saving optimization space; s4, generating an optimization instruction of the key control parameters of the propulsion system through a self-adaptive fuzzy PID algorithm in combination with the dynamic change of the sea condition and the load; s5, executing the optimization instruction, collecting response parameters of the propulsion system in real time, and dynamically correcting control parameters according to deviation to form a closed-loop control cycle; and S6, periodically updating adjustable parameters of the evaluation model and the adaptive fuzzy PID algorithm. According to the invention, through multi-parameter fusion perception and dynamic optimization control, accurate matching of the propulsion system, the navigation environment and the load demand is realized, and the economical efficiency and the environmental protection property of ship navigation are improved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent control technology for ship propulsion systems, specifically relating to an intelligent energy-saving control method and system for ship propulsion systems. Background Technology

[0002] The ship propulsion system is the core power unit of a ship, and its operating efficiency directly determines the ship's fuel consumption, sailing costs, and environmental performance. Most existing ship propulsion systems employ fixed control strategies, that is, setting parameters such as propeller speed and pitch according to preset speed requirements, resulting in a lack of control flexibility.

[0003] Ships face complex and ever-changing marine environments during navigation, and their loads also adjust dynamically. Traditional fixed control strategies cannot adapt to these dynamic changes in real time, resulting in a mismatch between the operating status of the propulsion system and actual needs. Maintaining high power output under harsh sea conditions will lead to fuel waste, and failure to optimize control parameters in time under light loads will result in low propulsion efficiency.

[0004] Existing control methods mostly rely on single parameter feedback and lack multi-dimensional information fusion analysis, which further limits the realization of energy-saving potential.

[0005] In summary, achieving precise matching between propulsion system control and dynamic environment and load requirements, and improving the intelligence and adaptability of control, has become a key breakthrough direction for energy-saving technologies in ship propulsion systems. Summary of the Invention

[0006] To address the technical problems of fixed energy-saving control strategies, poor adaptability, and high fuel consumption in existing ship propulsion systems, this invention provides an intelligent energy-saving control method and system for ship propulsion systems, which can achieve dynamic optimization control of the propulsion system and improve the economy and environmental friendliness of ship navigation.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention relates to an intelligent energy-saving control method for a ship propulsion system, characterized in that the control method includes the following steps: Step S1 involves collecting real-time data from multiple dimensions of the ship's propulsion system, hull, and deck using sensor arrays deployed on the ship's propulsion system, hull, and deck. The propulsion system operating parameters reflect the working status of core components, the ship's navigation status parameters reflect the overall operation of the ship, and the marine environment parameters capture changes in the external environment. These three types of parameters comprehensively cover the input information required for control. Step S2: Use the moving average filtering method to remove outliers of the above parameters, and preprocess the above parameters by min-max normalization to construct a multi-dimensional state feature set. Step S3: Input the multi-dimensional state feature set into the pre-trained gradient boosting-neural network composite evaluation model to evaluate the operating efficiency of the propulsion system and determine the energy-saving optimization space. Step S4: Based on the energy-saving optimization space, real-time sea state changes and ship load requirements, an adaptive fuzzy PID algorithm is used to generate optimized instructions for key control parameters of the propulsion system. Step S5: Execute the optimization command, collect the propulsion system response parameters in real time, and dynamically correct the control parameters according to the deviation to form a closed-loop control; Step S6: Periodically update the adjustable parameters of the evaluation model and the adaptive fuzzy PID algorithm.

[0008] Furthermore, in step S1, the propulsion system operating parameters include propeller speed. pitch Host load Fuel consumption rate and exhaust temperature The ship navigation status parameters include ship speed. ,course Draft and the ship's roll angle The marine environmental parameters include wave levels. Ocean current speed ,wind direction and wind speed .

[0009] Furthermore, the specific process of step S2 includes: Step S21: Perform moving average filtering preprocessing on the collected propulsion system operating parameters, ship navigation status parameters, and marine environmental parameters respectively. The processing formula is as follows: (1) in, For the size of the filter window, For the original parameter sequence, Here are the filtered parameters, and k is the index of the original parameter sequence; Step S22: Normalize the preprocessed data of the collected parameters using the following formula: (2) in, , These are the historical extreme values ​​of the parameters. To normalize the parameters and eliminate the dimensional differences between different parameters, all parameters are mapped to the [0,1] interval, and finally integrated to form a multi-dimensional state feature set.

[0010] Furthermore, the gradient boosting-neural network composite evaluation model in step S3 includes: Step S31, Gradient boosting sub-model output: (3) in, For the number of weak learners, For the first The weights of a weak learner For the first A decision tree weak learner, For the first The parameters of a decision tree weak learner For the state feature set; the expression for F is: (4) Step S32, output of the neural network sub-model: (5) in, For the hidden layer activation function, The output layer is a linear activation function. The number of neurons in the output layer. The number of neurons in the hidden layer. , For weight parameters, , For bias parameters, For the feature set One element; Step S33, Dynamic weighted fusion output: (6) in, ; Output for the neural network sub-model; Weight The expression is as follows: (7) Weight The expression is as follows (8) These represent the real-time prediction errors of the two sub-models.

[0011] Furthermore, the adaptive fuzzy PID algorithm in step S4 has the following control formula: (9) in, , , All of these are dynamically adjusted PID parameters. Forward bias, is the sampling period; k is the discrete sampling time number. Let k be the current deviation at time k. The historical deviation at time k-1; Furthermore, the initialization formula for the gradient boosting sub-model is as follows: (19) in, These are the predicted values ​​of the initial constants. Mean square loss function ; For the Round of iterations, calculating the negative gradient: (20) Will Use the target value to train the decision tree ; Calculate the weights of the weak learner: (twenty one) Update the model: (twenty two) Iteration Post-wheel output final result .

[0012] Neural network sub-model network structure: Input layer (8 neurons, corresponding to the feature set) → Hidden layer (32 neurons, activation function) → Output layer (1 neuron, linear activation function) ); Forward propagation formula: Hidden layer output Output layer output ,in , This is the weight matrix. , It is the bias vector; Perform dynamic weighted fusion and calculate prediction error in real time: (twenty three) (twenty four) This represents the number of samples within the sliding window at the current moment. Perform weight calculation: (7) in, To ensure that the sub-model with the smaller the prediction error has a larger weight; Final output: (6) Simultaneously, it outputs energy-saving optimization directions such as pitch adjustment, speed optimization, or power distribution adjustment.

[0013] The weighting coefficients are determined through cross-validation, and the output directions for pitch adjustment, speed optimization, or power distribution adjustment are also used for energy-saving optimization.

[0014] Furthermore, the specific process of step S5 includes: Step S51: After executing the optimization command, collect the fuel consumption rate in real time. , Promotion efficiency Ship speed ; Step S52, calculate the deviation: (10) in, To set the optimization target, This is the actual response value; Step S53: Dynamically adjust the PID parameters until the deviation is corrected. Permissible error range: (11) in, , For correction factor, These are the current control parameters.

[0015] Furthermore, the specific process of step S6 includes: Step S61: Based on the parameter data accumulated during long-term voyages, update the model parameters every 100 hours of voyage using the gradient descent method, with the following formula: (12) in, For learning rate, The gradient of the loss function. Step S62, Gradient boosting sub-model parameter update: (13) (14) in, The loss function; Step S63, update the parameters of the neural network sub-model: , , (15) Where Δw is the update increment of the neural network weight parameters, and μ is the learning rate. Let be the partial derivative of the loss function J with respect to the weight parameter w, the negative sign - indicates the direction correction for gradient descent, and Δb is the update increment of the neural network bias parameters. Let J be the partial derivative of the loss function J with respect to the bias parameter b, where J is the loss function. Let be the "actual true value" of the i-th sample. Let be the "neural network evaluation value" for the i-th sample.

[0016] Furthermore, by combining fuzzy rules, based on sea state levels... With load deviation Dynamically adjust the PID parameters: (16) (17) (18) in, These are the initial parameters. This is an increment for fuzzy inference.

[0017] Furthermore, the optimization instructions include propeller pitch adjustment. Main engine speed adjustment and propulsion system power allocation ratio .

[0018] The second aspect of the present invention relates to an intelligent energy-saving control system for a ship propulsion system, characterized in that it comprises: The data acquisition unit is used to collect multi-dimensional parameters through a group of sensors, including propulsion system operating parameters, ship navigation status parameters, and marine environmental parameters. The data preprocessing unit is used to preprocess multi-dimensional parameters through moving average filtering and normalization to construct a multi-dimensional state feature set; The efficiency evaluation unit is used to input the constructed multi-dimensional state feature set into the pre-trained gradient boosting-neural network composite evaluation model to evaluate the operating efficiency of the propulsion system and determine the energy-saving optimization space. The control optimization unit is used to generate optimized control commands through an adaptive fuzzy PID algorithm and dynamically correct them according to the deviation correction formula. The execution unit is used to receive control commands and drive the propulsion system to perform adjustments; The data acquisition unit, data preprocessing unit, efficiency evaluation unit, control optimization unit, and execution unit are sequentially connected by signals to form a closed-loop control.

[0019] The technical concept of this invention is as follows: Data reliability is ensured through moving average filtering and normalization formulas; the composite evaluation model improves efficiency evaluation accuracy through weighted fusion formulas, solving the problem of poor adaptability caused by the "single information" of traditional control; the adaptive fuzzy PID algorithm adapts to complex sea conditions and load changes through dynamic parameter adjustment formulas, avoiding the problems of lag and large overshoot in traditional PID control; dynamic closed-loop control is achieved through deviation calculation and parameter correction formulas, reducing fuel consumption and improving the economic efficiency of ship navigation; it can be adapted to the propulsion systems of different types of ships without large-scale modification of existing equipment, resulting in low implementation cost and strong applicability.

[0020] Compared with the prior art, the beneficial effects of the present invention are reflected in: Data reliability is ensured through moving average filtering and normalization formulas. The composite evaluation model improves efficiency evaluation accuracy through weighted fusion formulas, solving the problem of poor adaptability caused by the "single information" of traditional control. The adaptive fuzzy PID algorithm adapts to complex sea conditions and load changes through dynamic parameter adjustment formulas, avoiding the problems of response lag and large overshoot of traditional PID control. Dynamic closed-loop control is achieved through deviation calculation and parameter correction formulas, reducing fuel consumption and improving the economic efficiency of ship navigation. It can be adapted to the propulsion systems of different types of ships without large-scale modification of existing equipment, resulting in low implementation cost and strong applicability. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the system structure of Embodiment 1 of the present invention.

[0023] Explanation of key figure labels: 1 - Sensor group, 2 - Data acquisition unit, 3 - Data preprocessing unit, 4 - Efficiency evaluation unit, 5 - Control optimization unit, 6 - Execution unit, 7 - Ship propulsion system. Detailed Implementation

[0024] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, but it should be understood that the scope of protection of the present invention is not limited to the specific embodiments.

[0025] Example 1

[0026] refer to Figure 1 This invention discloses an intelligent energy-saving control method for a ship propulsion system, applied to the propulsion system of a 10,000-ton ocean-going cargo ship. The specific steps are as follows: Step S1: The sensor group collects parameters in multiple dimensions at a frequency of 10Hz. The collected parameters include: Propulsion system operating parameters: propeller speed pitch Host load Fuel consumption rate Exhaust temperature ; Ship navigation status parameters: Ship speed ,course Draft , hull rolling angle ; Marine environmental parameters: wave level Level, ocean current speed ,wind direction Wind speed .

[0027] Step S2: Use the moving average filtering method to remove outliers of the above parameters, and preprocess the above parameters by min-max normalization to construct a multi-dimensional state feature set. Specifically, using window size Using the moving average filter, taking fuel consumption rate as an example, the original sequence is: ; After filtering ; Perform normalization processing, such as pitch. Historical extreme values , After normalization ; Constructing a state feature set: .

[0028] Step S3: Input the multi-dimensional state feature set into the pre-trained gradient boosting-neural network composite evaluation model to evaluate the operating efficiency of the propulsion system and determine the energy-saving optimization space. Gradient boosting sub-model output: After 100 iterations, Points, real-time prediction error ; Neural network sub-model output: calculated via forward propagation Points, real-time prediction error ; Dynamic weighted fusion: , After fusion The optimization direction is "pitch adjustment + speed optimization".

[0029] Step S4: Based on the energy-saving optimization space, real-time sea state changes and ship load requirements, an adaptive fuzzy PID algorithm is used to generate optimized instructions for key control parameters of the propulsion system. Preset optimization goals Fuel consumption rate reduced to 209g / kWh; Initial PID parameters , , ; According to fuzzy rules, the deviation under sea state 3 is... hour, , , Dynamic parameters , , ; Generate optimized instructions, pitch Adjusted from 1.1 to 1.05 ( ), host speed Adjust from 110 r / min to 108 r / min ), power allocation ratio .

[0030] Step S5: Execute the optimization command, collect the propulsion system response parameters in real time, and dynamically correct the control parameters according to the deviation to form a closed-loop control; Fuel consumption rate after executing the command Actual deviation ,satisfy No further corrections are needed. If the fuel consumption rate only drops to 218g / kWh, the deviation is: Then the parameters are updated by correcting the formula; ; Reissue the instruction until the deviation meets the requirements.

[0031] Step S6: Regularly update the evaluation model and intelligent control algorithm parameters; After 100 hours of sailing, the weight parameters of the composite model and the rule parameters of the fuzzy PID are updated using the gradient descent method, based on the accumulated effective data, to further improve the energy-saving effect during subsequent voyages.

[0032] Example 2

[0033] refer to Figure 2The present invention provides an intelligent energy-saving control system for a ship propulsion system, comprising: Data acquisition unit 2 is used to collect multi-dimensional parameters through sensor group 1. These multi-dimensional parameters include propulsion system operating parameters, ship navigation status parameters, and marine environment parameters. Data preprocessing unit 3 is used to preprocess multi-dimensional parameters through moving average filtering and normalization to construct a multi-dimensional state feature set; Efficiency evaluation unit 4 is used to input the constructed multi-dimensional state feature set into the pre-trained gradient boosting-neural network composite evaluation model to evaluate the operating efficiency of the propulsion system and determine the energy-saving optimization space. The control optimization unit 5 is used to generate optimized control commands through an adaptive fuzzy PID algorithm and dynamically correct them according to the deviation correction formula. Execution unit 6 is used to receive control commands and drive propulsion system 7 to perform adjustments; The data acquisition unit 2, data preprocessing unit 3, efficiency evaluation unit 4, control optimization unit 5, and execution unit 6 are sequentially connected to form a closed-loop control.

[0034] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent energy-saving control of a ship propulsion system, characterized in that, The control method includes the following steps: Step S1: Real-time collection of ship propulsion system operating parameters, ship navigation status parameters, and marine environmental parameters; Step S2: Use the moving average filtering method to remove outliers of the above parameters, and preprocess the above parameters by min-max normalization to construct a multi-dimensional state feature set. Step S3: Input the multi-dimensional state feature set into the pre-trained gradient boosting-neural network composite evaluation model to evaluate the operating efficiency of the propulsion system and determine the energy-saving optimization space. Step S4: Based on the energy-saving optimization space, real-time sea state changes and ship load requirements, an adaptive fuzzy PID algorithm is used to generate optimized instructions for key control parameters of the propulsion system. Step S5: Execute the optimization command, collect the propulsion system response parameters in real time, and dynamically correct the control parameters according to the deviation to form a closed-loop control; Step S6: Periodically update the adjustable parameters of the evaluation model and the adaptive fuzzy PID algorithm.

2. The intelligent energy-saving control method for a ship propulsion system according to claim 1, characterized in that, In step S1, the propulsion system operating parameters include propeller speed. Pitch Host load Fuel consumption rate and exhaust temperature The ship navigation status parameters include ship speed. ,course Draft and the ship's roll angle The marine environmental parameters include wave levels. Ocean current speed ,wind direction and wind speed .

3. The intelligent energy-saving control method for a ship propulsion system according to claim 1, characterized in that, The specific process of step S2 includes: Step S21: Perform moving average filtering preprocessing on the collected propulsion system operating parameters, ship navigation status parameters, and marine environmental parameters respectively. The processing formula is as follows: （1） in, For the size of the filter window, For the original parameter sequence, Here are the filtered parameters, and k is the index of the original parameter sequence; Step S22: Normalize the preprocessed data of the collected parameters using the following formula: （2） in, , These are the historical extreme values ​​of the parameters. To normalize the parameters and eliminate the dimensional differences between different parameters, all parameters are mapped to the [0,1] interval, and finally integrated to form a multi-dimensional state feature set.

4. The intelligent energy-saving control method for a ship propulsion system according to claim 1, characterized in that, The gradient boosting-neural network composite evaluation model in step S3 includes: Step S31, Gradient boosting sub-model output: （3） in, For the number of weak learners, For the first The weights of a weak learner For the first A decision tree weak learner, For the first The parameters of a decision tree weak learner For the state feature set; the expression for F is: （4） Step S32, output of the neural network sub-model: （5） in, For the hidden layer activation function, The output layer is a linear activation function. The number of neurons in the output layer. The number of neurons in the hidden layer. , For weight parameters, , For bias parameters, For the feature set One element; Step S33, Dynamic weighted fusion output: （6） in, ; Output for the neural network sub-model; Weight The expression is as follows: （7） Weight The expression is as follows （8） These represent the real-time prediction errors of the two sub-models.

5. The intelligent energy-saving control method for a ship propulsion system according to claim 1, characterized in that, The adaptive fuzzy PID algorithm in step S4 has the following control formula: （9） in, , , All of these are dynamically adjusted PID parameters. Forward bias, The sampling period is k is the discrete sampling time number. Let k be the current deviation at time k. This represents the historical deviation at time k-1.

6. The intelligent energy-saving control method for a ship propulsion system according to claim 1, characterized in that, The specific process of step S5 includes: Step S51: After executing the optimization command, collect the fuel consumption rate in real time. , Promotion efficiency Ship speed ; Step S52, calculate the deviation: （10） in, To set the optimization target, This is the actual response value; Step S53: Dynamically adjust the PID parameters until the deviation is corrected. Permissible error range: （11） in, , For correction factor, These are the current control parameters.

7. The intelligent energy-saving control method for a ship propulsion system according to claim 1, characterized in that, The specific process of step S6 includes: Step S61: Based on the parameter data accumulated during long-term voyages, update the model parameters every 100 hours of voyage using the gradient descent method, with the following formula: （12） in, For learning rate, The gradient of the loss function. Step S62, Gradient boosting sub-model parameter update: （13） （14） in, The loss function; To boost the weights of the weak learner in the sub-model using gradient gradation; For decision tree parameters; Step S63, update the parameters of the neural network sub-model: ， ， （15） Where Δw is the update increment of the neural network weight parameters, and μ is the learning rate. Let be the partial derivative of the loss function J with respect to the weight parameter w, the negative sign - indicates the direction correction for gradient descent, and Δb is the update increment of the neural network bias parameters. Let J be the partial derivative of the loss function J with respect to the bias parameter b, where J is the loss function. Let be the "actual true value" of the i-th sample. Let be the "neural network evaluation value" for the i-th sample.

8. The intelligent energy-saving control method for a ship propulsion system according to claim 5, characterized in that, Combining fuzzy rules, based on sea state levels With load deviation The PID parameters are dynamically adjusted. （16） （17） （18） in, These are the initial parameters. This is an increment for fuzzy inference.

9. The intelligent energy-saving control method for a ship propulsion system according to claim 1, characterized in that, The optimization instructions include propeller pitch adjustment. Main engine speed adjustment and propulsion system power allocation ratio .

10. An intelligent energy-saving control system for a ship propulsion system, characterized in that, include: The data acquisition unit is used to collect multi-dimensional parameters through a group of sensors, including propulsion system operating parameters, ship navigation status parameters, and marine environmental parameters. The data preprocessing unit is used to preprocess multi-dimensional parameters through moving average filtering and normalization to construct a multi-dimensional state feature set; The efficiency evaluation unit is used to input the constructed multi-dimensional state feature set into the pre-trained gradient boosting-neural network composite evaluation model to evaluate the operating efficiency of the propulsion system and determine the energy-saving optimization space. The control optimization unit is used to generate optimized control commands through an adaptive fuzzy PID algorithm and dynamically correct them according to the deviation correction formula. The execution unit is used to receive control commands and drive the propulsion system to perform adjustments; The data acquisition unit, data preprocessing unit, efficiency evaluation unit, control optimization unit, and execution unit are sequentially connected by signals to form a closed-loop control.

Images