Methods, devices and electronic equipment for optimizing the minimum fuel consumption rate speed of diesel generator sets

Abstract

This invention relates to a method, apparatus, and electronic equipment for optimizing the minimum fuel consumption rate speed of a diesel generator set. The method includes: real-time acquisition of the generator set's operating parameter sequence and online steady-state detection and screening to obtain effective steady-state operating point data characterizing the current engine state; comparing the effective steady-state operating point data with existing data in a steady-state database, updating the steady-state database by dynamically replacing old data with new data; periodically, or when the data update volume of the steady-state database reaches a threshold, retraining the fuel consumption rate prediction model using all data in the updated steady-state database to calibrate the model parameters to match the engine's actual fuel consumption characteristics; calculating the minimum fuel consumption rate speed corresponding to the current power generation, and continuously updating a minimum fuel consumption rate speed reference curve for closed-loop control of the generator set speed. This invention overcomes the optimization lag and failure problems caused by relying on fixed offline data.

Description

Technical Field

[0001] This invention relates to the field of engine control technology, and in particular to a method, apparatus, and electronic equipment for optimizing the minimum fuel consumption rate of a diesel generator set. Background Technology

[0002] Marine variable-speed diesel generator sets dynamically adjust the diesel engine speed to match changing electrical loads, aiming to ensure the diesel engine operates at high efficiency and low fuel consumption, a key technology for achieving energy conservation and emission reduction in ships. The core of this technology lies in accurately finding and tracking the lowest fuel consumption rate speed of the diesel engine for any given electrical load. Traditional methods for obtaining the lowest fuel consumption rate speed mainly fall into two categories: one is the stepless adjustment optimization method based on mathematical models, such as fitting fuel consumption rate curves to experimental data and using optimization algorithms, or solving for the optimal speed by calculating partial derivatives based on empirical formulas. This type of method can theoretically find the optimal speed, but under ship operating conditions with frequent load fluctuations, it can easily lead to continuous fluctuations in diesel engine speed, affecting operational stability and reliability. The other type is the graded step adjustment method based on preset rules, such as setting a fixed optimal speed according to power ranges and introducing hysteresis control to reduce switching frequency. While this method improves operational stability, due to the randomness and non-uniformity of actual load distribution, its fuel-saving effect often does not reach the theoretical optimum, and near the range boundary points, load fluctuations may still cause frequent speed switching. In addition, there are now improvement schemes that attempt to balance fuel-saving effects and operational stability, such as using sliding window technology to process historical data to generate continuous optimized speed curves.

[0003] However, regardless of the model optimization, step adjustment, or curve generation methods based on historical data mentioned above, their common fundamental limitation lies in the fact that the fuel consumption rate characteristic model or optimal speed curve they rely on are all pre-obtained and fixed based on offline test data or quasi-static models of the diesel engine at a specific point in time (such as after factory testing or major overhaul). During actual long-term operation, the performance of a diesel engine will gradually degrade due to unavoidable factors such as wear and aging, causing its true fuel consumption rate characteristics and minimum fuel consumption speed to change accordingly. Since existing methods cannot respond online to this inherent performance change, the optimization models or curves they rely on will gradually become disconnected from actual operating conditions, thus limiting the long-term energy-saving effect that the system can achieve throughout its service life. Summary of the Invention

[0004] Based on this, it is necessary to provide a method, device, and electronic equipment for optimizing the minimum fuel consumption rate speed of diesel generator sets, which enables the optimization of the minimum fuel consumption rate speed to track the performance status changes of the diesel engine over time in an online and adaptive manner, and overcomes the optimization lag and failure caused by relying on fixed offline data.

[0005] This invention provides a method for optimizing the minimum fuel consumption rate speed of a diesel generator set, the method comprising:

[0006] The generator set's operating parameter sequence is collected in real time, and the operating parameter sequence is subjected to online steady-state detection and screening to obtain effective steady-state operating point data that characterizes the current engine state.

[0007] The effective steady-state operating point data is compared with the existing data in the steady-state database. Based on the comparison results, the steady-state database is updated by dynamically replacing the old data with the new data, so that the data distribution in the database can reflect the latest performance status of the engine.

[0008] Periodically, or when the amount of data updated in the steady-state database reaches a threshold, the fuel consumption rate prediction model is retrained using all the data in the updated steady-state database to calibrate the model parameters so that they match the actual fuel consumption characteristics of the engine.

[0009] Based on the newly trained fuel consumption rate prediction model, the minimum fuel consumption rate speed corresponding to the current power generation is calculated, and a minimum fuel consumption rate speed reference curve for unit speed closed-loop control is continuously updated based on this result.

[0010] In one embodiment, the online steady-state detection and screening of the operating parameter sequence includes:

[0011] A preset detection algorithm is used to determine the steady state of the real-time collected power and speed parameter sequences, and to identify whether the overall fluctuation of the parameter vector is in a steady state.

[0012] When a steady state is determined, a timer is started to record the duration of the steady state. Only when the duration of the steady state exceeds a preset time threshold is the data within the current time period marked as the valid steady-state operating point data.

[0013] In one embodiment, the preset detection algorithm includes, but is not limited to, any one of sliding window combined with linear regression fitting and residual analysis, Kalman filtering, threshold determination, and machine learning classification; the preset detection algorithm can screen out steady-state data with parameter fluctuation amplitude within a preset range, and reduce the false judgment rate through a significance level adjustment mechanism when multiple parameters are jointly determined.

[0014] When the preset detection algorithm employs linear regression fitting and residual analysis based on a sliding window, it specifically includes:

[0015] For the parameter sequence within the window The slope was obtained by linear regression fitting. and intercept And calculate the residual sequence and the estimated value of white noise fluctuation amplitude. ,in, For time;

[0016] At a given significance level, if the window satisfies If the sample proportion exceeds a preset proportion, the parameter is determined to be in a steady state. This is the critical value;

[0017] For joint determination of multiple parameters, the significance level is adjusted using the Sidak inequality: ,in, To monitor the number of parameters, The overall significance level is denoted as .

[0018] In one embodiment, the step of comparing the effective steady-state operating point data with existing data in the steady-state database specifically involves:

[0019] The new effective steady-state data vector is calculated using a pre-defined similarity metric method. With existing data vectors in the database Similarity indicators between them;

[0020] If the similarity index meets the preset judgment conditions, it is judged as similar data; otherwise, it is judged as new working condition data.

[0021] When the preset similarity measurement method uses the Euclidean distance method, the calculation formula is as follows:

[0022]

[0023] in, For both new and old power, For the old and new speeds;

[0024] If the Euclidean distance is less than or equal to a preset dynamic distance threshold, it is determined to be similar data.

[0025] In one embodiment, the fuel consumption rate prediction model is a machine learning prediction model that can characterize the mapping relationship between engine speed, power and fuel consumption rate. Its input parameters include engine speed and real-time power, and its output parameter is the predicted fuel consumption rate.

[0026] When the fuel consumption rate prediction model adopts a feedforward neural network model, its input layer parameter is the engine speed. With real-time power The output layer parameters are the predicted fuel consumption rate. The mapping function relationship established by the fuel consumption rate prediction model is expressed as follows:

[0027]

[0028] in, This represents a feedforward neural network that has completed training. It is the set of all weights and bias parameters in a neural network.

[0029] In one embodiment, the retraining of the fuel consumption rate prediction model further includes:

[0030] The speed, power and fuel consumption data extracted from the steady-state database are preprocessed, including standardization and outlier removal; the preprocessed data are then divided into training set, validation set and test set according to a preset ratio.

[0031] In one embodiment, the step of calculating the minimum fuel consumption rate speed corresponding to the current power generation includes:

[0032] Optimize the width of the load slip range with the goal of minimizing average fuel consumption. The objective function is: The constraints are ， in, This represents the average fuel consumption rate. Given the set of natural numbers, the average fuel consumption rate Through formula Calculation, where This refers to the instantaneous fuel consumption of the diesel engine. This refers to the instantaneous power of the diesel engine. The integration time;

[0033] Based on the optimized interval width, a sliding interval is constructed. Within each sliding interval, the fuel consumption rate prediction model is used to find the lowest fuel consumption rate speed corresponding to the center load of that interval through an optimization algorithm.

[0034] In one embodiment, the optimization algorithm is used to find the optimal solution. The initialization strategy of the preset optimization algorithm is selected from local disturbance generation, global random generation, or a combination of both, depending on the actual working conditions.

[0035] When the gray wolf optimization algorithm is used as the preset optimization algorithm, during the initialization of the population, some individuals are generated by small-scale perturbation around the historical best speed in the same load range in the database, while others are randomly generated within the preset speed range.

[0036] In one embodiment, the method further includes:

[0037] The actual operating parameters generated by running based on the updated minimum fuel consumption rate speed reference curve are used as input again. After online steady-state detection and screening, new effective steady-state operating point data are generated for subsequent steady-state database updates and model training, thus forming a closed-loop adaptive optimization process of running-optimizing-updating-running again.

[0038] The present invention also provides a diesel generator set minimum fuel consumption rate speed optimization device, comprising:

[0039] The effective steady-state operating point data acquisition module is used to collect the operating parameter sequence of the generator set in real time, and to perform online steady-state detection and screening on the operating parameter sequence to obtain effective steady-state operating point data that characterizes the current engine state.

[0040] The steady-state database update module is used to compare the effective steady-state operating point data with the existing data in the steady-state database, and update the steady-state database by dynamically replacing the old data with new data based on the comparison results, so that the data distribution in the database can reflect the latest performance status of the engine.

[0041] The model parameter calibration module is used to periodically or when the data update volume of the steady-state database reaches a threshold, to retrain the fuel consumption rate prediction model using all the updated data in the steady-state database, so as to calibrate the model parameters to match the actual fuel consumption characteristics of the engine.

[0042] The fuel consumption rate speed reference curve update module is used to calculate the minimum fuel consumption rate speed corresponding to the current power generation based on the latest trained fuel consumption rate prediction model, and continuously update a minimum fuel consumption rate speed reference curve for unit speed closed-loop control based on this result.

[0043] The present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the diesel generator set minimum fuel consumption rate speed optimization method as described above.

[0044] The aforementioned method, device, and electronic equipment for optimizing the minimum fuel consumption rate speed of diesel generator sets directly and effectively solve the technical problem of the failure of a fixed offline data model due to engine performance degradation by constructing a closed-loop adaptive process that includes dynamic data updates and online model calibration. Specifically, firstly, online steady-state detection and screening ensure that the collected effective steady-state operating point data can truly reflect the current actual operating state of the engine; secondly, by comparing the new data with the steady-state database and implementing a dynamic update strategy of replacing the old with the new, the steady-state database can continuously track and characterize the gradual changes in engine performance; based on this, by periodically triggering the retraining of the fuel consumption rate prediction model or based on the amount of data updates, the model parameters can be continuously calibrated, thereby maintaining an accurate mapping of the current real fuel consumption characteristics of the engine; finally, using this online-updated model, the minimum fuel consumption rate speed reference curve is calculated in real time and continuously updated, so that the speed optimization control of the generator set can adaptively follow the time-varying degradation of engine performance, overcoming the optimization lag problem caused by model fixation, and achieving a long-term stable operation effect of maintaining near-optimal fuel consumption throughout the entire life cycle. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in this invention 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0046] Figure 1 A flowchart of a method for optimizing the minimum fuel consumption rate speed of a diesel generator set, as shown in one embodiment;

[0047] Figure 2 Invention content framework diagram optimized for online use;

[0048] Figure 3 The power detection results are shown in the figure.

[0049] Figure 4 The result of the rotational speed detection is shown in the figure.

[0050] Figure 5 This is a timing diagram of the neural network training process.

[0051] Figure 6 This is a schematic diagram of the load sliding range;

[0052] Figure 7 To optimize the curve comparison diagram;

[0053] Figure 8 A schematic diagram of a diesel generator set minimum fuel consumption rate speed optimization device according to one embodiment;

[0054] Figure 9 This is an internal structural diagram of an electronic device according to one embodiment. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] The following is combined Figures 1-9 The present invention describes a method, apparatus, and electronic equipment for optimizing the minimum fuel consumption rate of a diesel generator set.

[0057] like Figure 1 As shown, in one embodiment, a method for optimizing the minimum fuel consumption rate speed of a diesel generator set includes the following steps:

[0058] Step S110: Real-time acquisition of the generator set's operating parameter sequence, and online steady-state detection and screening of the operating parameter sequence to obtain effective steady-state operating point data characterizing the current engine state.

[0059] A preset detection algorithm is used to determine the steady-state condition of the real-time acquired power and speed parameter sequences, identifying whether the overall fluctuation of the parameter vector is in a steady state. When a steady state is determined, a timer is started to record the steady-state duration. Only when the steady-state duration exceeds a preset threshold is the data within the current time period marked as valid steady-state operating point data. The preset detection algorithm includes, but is not limited to, any one of the following: sliding window combined with linear regression fitting and residual analysis, Kalman filtering, threshold determination method, and machine learning classification method. The preset detection algorithm can filter out steady-state data with parameter fluctuation amplitudes within a preset range, and reduces the false positive rate through a significance level adjustment mechanism when making joint determinations of multiple parameters. When the preset detection algorithm uses linear regression fitting and residual analysis based on a sliding window, the system continuously acquires real-time sequences of parameters such as power and speed through sensors. See [link to relevant documentation]. Figure 2 The algorithm processes the data stream using a steady-state detection algorithm based on sliding window and residual analysis. Specifically, the algorithm establishes a "random walk with drift" model for each parameter. ,in, For deterministic drift components ( The slope (for time), For the mean of a stationary process, It is zero-mean white noise. Non-stationary characteristics can be separated using first-order difference. If the signal shows a significant trend (drift components cannot be ignored), it is determined to be non-steady-state. Data is extracted by sliding window and linear regression is performed to calculate the slope, intercept, and residuals. The white noise fluctuation amplitude and statistical confidence interval are used as criteria. When most of the residuals of the data points within the window fall within the confidence interval, it is initially determined to be steady-state.

[0060] In practical applications, the system may monitor multiple parameters simultaneously. See also Figure 3 , Figure 4 The key parameters are treated as a whole parameter vector and kept relatively constant within a certain time window, while the joint fluctuation amplitude of each component in the parameter vector is controlled within a preset threshold range. By comprehensively analyzing the dynamic trend and stability of this vector, it is determined whether the system has reached a steady state. This multi-parameter joint detection method can effectively avoid misjudgments caused by anomalies in a single parameter. A sliding window is used to update data online, and the mean, slope, and residual are dynamically calculated for each sampling interval, supporting multi-window parallel processing to improve reliability.

[0061] To further eliminate transient fluctuations, the system introduces a duration criterion. Only when the state initially determined to be steady-state continues for more than a preset time threshold (e.g., 60 seconds) is the data within that time period finally marked as valid steady-state operating point data. This process accurately filters out transient and noise data from the highly fluctuating ship load conditions, ensuring that the data used for subsequent optimization all come from the system's equilibrium state. This avoids optimization errors introduced by non-steady-state data and provides a high-quality and reliable data foundation for the entire closed-loop optimization process.

[0062] When the preset detection algorithm uses linear regression fitting and residual analysis based on a sliding window, it specifically includes: for the parameter sequence within the window... The slope was obtained by linear regression fitting. and intercept And calculate the residual sequence and the estimated value of white noise fluctuation amplitude. At a given significance level, if the window satisfies If the sample proportion exceeds a preset proportion, the parameter is determined to be in a steady state; for joint determination of multiple parameters, the significance level is adjusted using the Sidak inequality: ,in, To monitor the number of parameters, The overall significance level is determined by the window length, which should be 3-5 times greater than the process time constant to balance detection sensitivity and timeliness.

[0063] Specifically, the steady-state detection logic is implemented through three progressively advancing judgments: the first layer is the trend stability judgment, which is based on linear regression on the parameter sequence within the sliding window. The slope was obtained by fitting. With intercept , If the slope If the absolute value is less than a preset minimum threshold, the parameter is considered to have no obvious trend change within the window; the second layer is the fluctuation stability judgment, which calculates the residual sequence of the data within the window based on the trend judgment. and its white noise fluctuation amplitude estimate And check if there are data points that meet the preset proportion (e.g., 95%). ,in The first layer is the critical value obtained by looking up the t-distribution table based on the significance level (e.g., α=0.05) and degrees of freedom; the third layer is a multi-parameter joint steady-state judgment. Based on the single-parameter steady-state judgment, to avoid the increased misjudgment rate caused by monitoring multiple parameters simultaneously, the Sidak correction method is used to adjust the significance level of each parameter. ,in To monitor the number of parameters (usually power, speed, and fuel consumption rate), and then use the adjusted... Recalculate each parameter A second layer of judgment is then performed. Only when all monitored parameters pass this second layer of judgment is the entire system considered to be in a steady state, thus ensuring the overall stability and representativeness of the selected data. It should be noted that in the "random walk with drift" model, the data sequence within the sliding window... Perform linear regression fitting The parameters obtained Mathematically, this is the intercept of a linear equation. In the specific application context of this algorithm, when determining the stationarity of data within a window, this intercept... The physical meaning is to remove the estimated linear trend (determined by the slope). After characterization, the data sequence revolves around the center value of the fluctuation. Therefore, in calculating the residual... and fluctuation criteria At that time, This means that the local steady-state mean after detrending within this window is used for calculation, and the "intercept" and "stationary process mean" refer to the same calculated value here.

[0064] Step S120: The effective steady-state operating point data is compared with the existing data in the steady-state database. Based on the comparison results, the steady-state database is updated by dynamically replacing the old data with the new data, so that the data distribution in the database can reflect the latest performance status of the engine.

[0065] The new effective steady-state data vector is calculated using a pre-defined similarity metric method. With existing data vectors in the database The similarity index between the data is used to determine whether the data is similar or not. If the similarity index meets the preset judgment conditions, the data is judged as similar; otherwise, it is judged as new working condition data. When the preset similarity measurement method adopts the Euclidean distance method, the calculation formula is as follows:

[0066]

[0067] in, For both new and old power, The new and old rotation speeds are used as references; if the Euclidean distance is less than or equal to the preset dynamic distance threshold, it is determined to be similar data.

[0068] The newly acquired effective steady-state data vector (denoted as) ,in For power, (for rotational speed) and each historical data vector in the database (denoted as) Calculate Euclidean distance This distance is then compared to a dynamically set threshold (dynamic distance threshold), which is derived based on typical operating conditions where the speed difference is no greater than 5 rpm and the power difference is no greater than 2%, with its upper limit being [value missing]. If the distance is less than or equal to the dynamic distance threshold, the system determines that the old and new data represent similar operating conditions, and overwrites the old records with the latest information such as fuel consumption rate from the new data. If the distance is greater than the threshold, the new data is inserted into the database as a completely new operating condition point. Furthermore, to prevent the database from expanding indefinitely, the system sets a maximum capacity. When the data volume reaches the limit, the oldest data entry is removed according to the first-in, first-out (FIFO) principle to make room for the new data entry. This mechanism achieves intelligent dynamic management of the steady-state database. It avoids the accumulation of redundant data through similarity comparison and keeps the fuel consumption rate parameters in the database up-to-date through overwriting updates. This allows the database as a whole to track and reflect the engine performance state migration caused by wear and aging in real time, providing a sample set that fits the current reality for subsequent model training.

[0069] Step S130: Periodically or when the data update volume of the steady-state database reaches a threshold, use all the data in the updated steady-state database to retrain the fuel consumption rate prediction model in order to calibrate the model parameters to match the actual fuel consumption characteristics of the engine.

[0070] Before training, the speed, power, and fuel consumption rate data extracted from the steady-state database are preprocessed to eliminate the influence of dimensions. Preprocessing includes, but is not limited to, standardization and outlier removal. The preprocessed data is then divided into training, validation, and test sets according to a preset ratio. The timing of this process is detailed in [link to relevant documentation]. Figure 5The fuel consumption rate prediction model is a machine learning prediction model that can characterize the mapping relationship between engine speed, power, and fuel consumption rate. Its input parameters include engine speed and real-time power, and its output parameter is the predicted fuel consumption rate. When the fuel consumption rate prediction model adopts a feedforward neural network model, its input layer parameter is engine speed. With real-time power The output layer parameters are the predicted fuel consumption rate. The mapping function relationship established by the fuel consumption rate prediction model is expressed as:

[0071]

[0072] in, This represents a feedforward neural network that has completed training. This is the set of all weights and bias parameters in the neural network. During training, a validation set is used to monitor the error, and early stopping may be triggered to prevent overfitting. After training, it needs to be evaluated on the test set. It is only updated and saved when the performance indicators (such as RMSE, R²) meet the standards. The function and effect of this step is that it establishes a "digital twin" model of fuel consumption rate characteristics that can be adapted online. Through periodic or triggered retraining, the parameters of the neural network model are recalibrated based on the latest steady-state database, so that the fuel consumption rate characteristic curve predicted by the model can synchronously track the real characteristic changes of the engine caused by performance degradation. This overcomes the fatal defect of the fixed model gradually becoming disconnected from the real engine state in traditional methods, and provides an accurate and real-time updated objective function basis for online optimization.

[0073] Step S140: Based on the latest trained fuel consumption rate prediction model, calculate the minimum fuel consumption rate speed corresponding to the current power generation, and continuously update a minimum fuel consumption rate speed reference curve for unit speed closed-loop control based on this result.

[0074] Calculate the minimum fuel consumption rate speed corresponding to the current power generation capacity, including: First, optimizing the width of the load slip range with the goal of minimizing the average fuel consumption rate. The objective function is: The constraints are ， in, This represents the average fuel consumption rate. Given the set of natural numbers, the average fuel consumption rate Through formula Calculation, where This refers to the instantaneous fuel consumption of the diesel engine. This refers to the instantaneous power of the diesel engine. The integral time is used. Secondly, a sliding interval is constructed based on the optimized interval width. Within each sliding interval, the fuel consumption rate prediction model is used to find the lowest fuel consumption rate speed corresponding to the center load of that interval through an optimization algorithm. The optimization algorithm employs a preset optimization algorithm. The initialization strategy of the preset optimization algorithm selects local perturbation generation, global random generation, or a combination of both based on actual operating conditions. When the preset optimization algorithm uses the Grey Wolf optimization algorithm, during initialization, some individuals are generated with small-scale perturbations around the historical best speed of the same load interval in the database, while others are randomly generated within a preset speed range.

[0075] After determining the optimal interval width, a sliding interval is constructed using this width, allowing it to move continuously within the full load range of the generator set with fixed step sizes. See [link to documentation]. Figure 6 Within each sliding interval, the system calls the latest fuel consumption rate prediction model as the objective function and uses intelligent algorithms such as Grey Wolf optimization to find the lowest fuel consumption rate speed corresponding to the center load of that interval. The set of optimization results for all intervals forms an updated baseline curve of the lowest fuel consumption rate speed covering the entire load range (the comparison of its effects before and after optimization can be found in [reference]). Figure 7 Therefore, by utilizing a real-time updated, high-precision fuel consumption rate model, online optimization calculations are dynamically performed, and the results are transformed into setpoint curves that can be directly used for closed-loop speed control. This allows the generator set's speed setting to respond instantly to the latest engine performance state, regardless of whether it is in its initial state or has experienced performance degradation. The system can always generate the optimal speed command for the current state online, thereby achieving long-term, adaptive, and lowest fuel consumption operation. At the same time, through sliding interval optimization and weighted average logic, this method achieves both fuel-saving effects and operational stability.

[0076] Furthermore, this embodiment also includes a closed-loop feedback process: the actual operating parameters generated based on the updated minimum fuel consumption rate speed reference curve are used as input again. After online steady-state detection and screening, new effective steady-state operating point data are generated for subsequent steady-state database updates and model training, thus forming a closed-loop adaptive optimization process of "run-optimize-update-run". This closed-loop feedback mechanism gives the system the ability to self-improve and continuously track in the long term. As long as the unit is running, new steady-state data will be continuously collected, evaluated, and stored, and the model will be periodically updated and re-optimized, making the entire system an intelligent optimization system that can evolve synchronously with the performance degradation of the engine throughout its entire life cycle, ultimately achieving the goal of maintaining the best energy-saving effect throughout the entire service life.

[0077] The following describes the diesel generator set minimum fuel consumption rate speed optimization device provided by the present invention. The diesel generator set minimum fuel consumption rate speed optimization device described below and the diesel generator set minimum fuel consumption rate speed optimization method described above can be referred to in correspondence.

[0078] like Figure 8 As shown, in one embodiment, a diesel generator set minimum fuel consumption rate speed optimization device includes an effective steady-state operating point data acquisition module 810, a steady-state database update module 820, a model parameter calibration module 830, and a fuel consumption rate speed reference curve update module 840.

[0079] The effective steady-state operating point data acquisition module 810 is used to collect the operating parameter sequence of the generator set in real time, and to perform online steady-state detection and screening of the operating parameter sequence in order to obtain effective steady-state operating point data that characterizes the current engine state.

[0080] The steady-state database update module 820 is used to compare the effective steady-state operating point data with the existing data in the steady-state database, and update the steady-state database by dynamically replacing the old data with the new data based on the comparison results, so that the data distribution in the database can reflect the latest performance status of the engine.

[0081] The model parameter calibration module 830 is used to periodically or when the data update amount of the steady-state database reaches a threshold, to retrain the fuel consumption rate prediction model using all the data in the updated steady-state database, so as to calibrate the model parameters to match the actual fuel consumption characteristics of the engine.

[0082] The fuel consumption rate speed reference curve update module 840 is used to calculate the minimum fuel consumption rate speed corresponding to the current power generation based on the latest trained fuel consumption rate prediction model, and continuously update a minimum fuel consumption rate speed reference curve for unit speed closed-loop control based on this result.

[0083] Figure 9 This example illustrates a schematic diagram of the physical structure of an electronic device, which can be a smart terminal. Its internal structure diagram can be as follows: Figure 9 As shown. The electronic device includes a processor, memory, and network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for optimizing the speed of a diesel generator set to achieve the lowest fuel consumption rate. This method includes:

[0084] Real-time acquisition of generator set operating parameter sequences, and online steady-state detection and screening of operating parameter sequences to obtain effective steady-state operating point data characterizing the current engine state;

[0085] The effective steady-state operating point data is compared with the existing data in the steady-state database. Based on the comparison results, the steady-state database is updated by dynamically replacing the old data with the new data, so that the data distribution in the database can reflect the latest performance status of the engine.

[0086] Periodically, or when the amount of data updated in the steady-state database reaches a threshold, the fuel consumption rate prediction model is retrained using all the data in the updated steady-state database to calibrate the model parameters so that they match the actual fuel consumption characteristics of the engine.

[0087] Based on the latest trained fuel consumption rate prediction model, the minimum fuel consumption rate speed corresponding to the current power generation is calculated, and a minimum fuel consumption rate speed reference curve for unit speed closed-loop control is continuously updated based on this result.

[0088] Those skilled in the art will understand that Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the electronic device to which the present invention is applied. A specific electronic device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0089] On the other hand, the present invention also provides a computer storage medium storing a computer program, which, when executed by a processor, implements a method for optimizing the speed of a diesel generator set to achieve the lowest fuel consumption rate. This method includes:

[0090] Real-time acquisition of generator set operating parameter sequences, and online steady-state detection and screening of operating parameter sequences to obtain effective steady-state operating point data characterizing the current engine state;

[0091] The effective steady-state operating point data is compared with the existing data in the steady-state database. Based on the comparison results, the steady-state database is updated by dynamically replacing the old data with the new data, so that the data distribution in the database can reflect the latest performance status of the engine.

[0092] Periodically, or when the amount of data updated in the steady-state database reaches a threshold, the fuel consumption rate prediction model is retrained using all the data in the updated steady-state database to calibrate the model parameters so that they match the actual fuel consumption characteristics of the engine.

[0093] Based on the latest trained fuel consumption rate prediction model, the minimum fuel consumption rate speed corresponding to the current power generation is calculated, and a minimum fuel consumption rate speed reference curve for unit speed closed-loop control is continuously updated based on this result.

[0094] In another aspect, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium, and when the processor executes the computer instructions, it implements a method for optimizing the minimum fuel consumption rate speed of a diesel generator set, the method comprising:

[0095] Real-time acquisition of generator set operating parameter sequences, and online steady-state detection and screening of operating parameter sequences to obtain effective steady-state operating point data characterizing the current engine state;

[0096] The effective steady-state operating point data is compared with the existing data in the steady-state database. Based on the comparison results, the steady-state database is updated by dynamically replacing the old data with the new data, so that the data distribution in the database can reflect the latest performance status of the engine.

[0097] Periodically, or when the amount of data updated in the steady-state database reaches a threshold, the fuel consumption rate prediction model is retrained using all the data in the updated steady-state database to calibrate the model parameters so that they match the actual fuel consumption characteristics of the engine.

[0098] Based on the latest trained fuel consumption rate prediction model, the minimum fuel consumption rate speed corresponding to the current power generation is calculated, and a minimum fuel consumption rate speed reference curve for unit speed closed-loop control is continuously updated based on this result.

[0099] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory.

[0100] By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0101] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0102] The above-described embodiments are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A method for optimizing the minimum fuel consumption rate speed of a diesel generator set, characterized in that, The method includes: The generator set's operating parameter sequence is collected in real time, and the operating parameter sequence is subjected to online steady-state detection and screening to obtain effective steady-state operating point data that characterizes the current engine state. The effective steady-state operating point data is compared with the existing data in the steady-state database. Based on the comparison results, the steady-state database is updated by dynamically replacing the old data with the new data, so that the data distribution in the database can reflect the latest performance status of the engine. Periodically, or when the amount of data updated in the steady-state database reaches a threshold, the fuel consumption rate prediction model is retrained using all the data in the updated steady-state database to calibrate the model parameters so that they match the actual fuel consumption characteristics of the engine. Based on the latest trained fuel consumption rate prediction model, the minimum fuel consumption rate speed corresponding to the current power generation is calculated, and a minimum fuel consumption rate speed reference curve for unit speed closed-loop control is continuously updated based on this result. The step of comparing the effective steady-state operating point data with existing data in the steady-state database specifically involves: The new effective steady-state data vector is calculated using a pre-defined similarity metric method. With existing data vectors in the database Similarity indicators between them; If the similarity index meets the preset judgment conditions, it is judged as similar data; otherwise, it is judged as new working condition data. When the preset similarity measurement method uses the Euclidean distance method, the calculation formula is as follows: in, For both new and old power, For the old and new speeds; If the Euclidean distance is less than or equal to a preset dynamic distance threshold, it is determined to be similar data.

2. The method for optimizing the minimum fuel consumption rate speed of a diesel generator set according to claim 1, characterized in that, The online steady-state detection and screening of the operating parameter sequence includes: A preset detection algorithm is used to determine the steady state of the real-time collected power and speed parameter sequences, and to identify whether the overall fluctuation of the parameter vector is in a steady state. When a steady state is determined, a timer is started to record the duration of the steady state. Only when the duration of the steady state exceeds a preset time threshold is the data within the current time period marked as the valid steady-state operating point data.

3. The method for optimizing the minimum fuel consumption rate speed of a diesel generator set according to claim 2, characterized in that, The preset detection algorithm includes, but is not limited to, any one of the following: sliding window combined with linear regression fitting and residual analysis, Kalman filtering, threshold determination, and machine learning classification; the preset detection algorithm can screen out steady-state data with parameter fluctuations within a preset range, and reduces the false judgment rate through a significance level adjustment mechanism when making joint determinations of multiple parameters; When the preset detection algorithm employs linear regression fitting and residual analysis based on a sliding window, it specifically includes: For the parameter sequence within the window The slope was obtained by linear regression fitting. and intercept And calculate the residual sequence and the estimated value of white noise fluctuation amplitude. ,in, For time; At a given significance level, if the window satisfies If the sample proportion exceeds a preset proportion, the parameter is determined to be in a steady state. This is the critical value; For joint determination of multiple parameters, the significance level is adjusted using the Sidak inequality: ,in, To monitor the number of parameters, The overall significance level is denoted as .

4. The method for optimizing the minimum fuel consumption rate speed of a diesel generator set according to claim 1, characterized in that, The fuel consumption rate prediction model is a machine learning prediction model that can characterize the mapping relationship between engine speed, power and fuel consumption rate. Its input parameters include engine speed and real-time power, and its output parameter is the predicted fuel consumption rate. When the fuel consumption rate prediction model adopts a feedforward neural network model, its input layer parameter is the engine speed. With real-time power The output layer parameters are the predicted fuel consumption rate. The mapping function relationship established by the fuel consumption rate prediction model is expressed as follows: in, This represents a feedforward neural network that has completed training. It is the set of all weights and bias parameters in a neural network.

5. The method for optimizing the minimum fuel consumption rate speed of a diesel generator set according to claim 1, characterized in that, The retrained fuel consumption rate prediction model previously included: The speed, power and fuel consumption rate data extracted from the steady-state database are preprocessed, including standardization and outlier removal. The preprocessed data is divided into training set, validation set and test set according to a preset ratio.

6. The method for optimizing the minimum fuel consumption rate speed of a diesel generator set according to claim 1, characterized in that, The lowest fuel consumption rate speed corresponding to the current power generation is calculated, including: Optimize the width of the load slip range with the goal of minimizing average fuel consumption. The objective function is: The constraints are ， in, This represents the average fuel consumption rate. Given the set of natural numbers, the average fuel consumption rate Through formula Calculation, where This refers to the instantaneous fuel consumption of the diesel engine. This refers to the instantaneous power of the diesel engine. The integration time; Based on the optimized interval width, a sliding interval is constructed. Within each sliding interval, the fuel consumption rate prediction model is used to find the lowest fuel consumption rate speed corresponding to the center load of that interval through an optimization algorithm.

7. The method for optimizing the minimum fuel consumption rate speed of a diesel generator set according to claim 6, characterized in that, The optimization algorithm is used to find the best candidate. The initialization strategy of the optimization algorithm is selected based on the actual working conditions, which may be local disturbance generation, global random generation, or a combination of both. When the gray wolf optimization algorithm is used as the preset optimization algorithm, during the initialization of the population, some individuals are generated by small-scale perturbation around the historical best speed in the same load range in the database, while others are randomly generated within the preset speed range.

8. The method for optimizing the minimum fuel consumption rate speed of a diesel generator set according to any one of claims 1 to 7, characterized in that, The method further includes: The actual operating parameters generated by running based on the updated minimum fuel consumption rate speed reference curve are used as input again. After online steady-state detection and screening, new effective steady-state operating point data are generated for subsequent steady-state database updates and model training, thus forming a closed-loop adaptive optimization process of running-optimizing-updating-running again.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method for optimizing the minimum fuel consumption rate speed of a diesel generator set as described in any one of claims 1 to 8.

Images