A shipborne multi-source flexible direct-current looped network test bench and an energy management method thereof

Abstract

The application discloses a shipborne multi-source flexible DC loop network test bench, which comprises a DC loop network main system, a first hydrogen fuel cell power supply system, a second hydrogen fuel cell power supply system, a first composite energy storage system, a second composite energy storage system, a first load system, a second load system, a compact reconfigurable input / output (CRIO) controller, a lower computer and an upper computer, wherein the DC loop network main system comprises a first DC bus, a second DC bus, a third DC bus and a fourth DC bus which are electrically connected in sequence through four solid-state switches; the first hydrogen fuel cell power supply system is electrically connected with the second DC bus, the second hydrogen fuel cell power supply system is electrically connected with the fourth DC bus, the first composite energy storage system is electrically connected with the first DC bus, and the second composite energy storage system is electrically connected with the third DC bus. The application can solve the technical problem that the existing rule type control method excessively depends on engineering experience and preset rules and is difficult to adapt to complex and changeable working conditions.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of ship power systems, and more particularly relates to a shipborne multi-source flexible DC loop network test bench and an energy management method thereof. BACKGROUND

[0002] With the continuous improvement of ship electrification and intelligence, shipborne flexible DC loop network systems have gradually become an important development direction of ship power systems. In order to verify and optimize the performance of the system, it is particularly important to develop a professional shipborne multi-source flexible DC loop network test bench. The shipborne multi-source flexible DC loop network test bench not only can simulate actual sailing conditions, but also can verify and optimize the energy management strategy of the system to achieve efficient coordination and control between multiple energy sources. Among them, energy management, as the core of the system, is mainly responsible for coordinating the power distribution of various energy equipment to ensure the economy and reliability of the system.

[0003] Currently, the energy management of the shipborne multi-source flexible DC loop network test bench mainly includes three types, rule-based control method, dynamic programming (Dynamic Programming, DP) method, and model predictive control (Model Predictive Control, MPC) method: the rule-based control method performs energy distribution through pre-set logical rules, which has the characteristics of simple implementation; the dynamic programming method realizes optimal energy distribution through global optimization, but requires complete condition information to be known in advance; the model predictive control method realizes optimal control of the system within the prediction time domain through rolling optimization, which can effectively balance the calculation efficiency and control performance.

[0004] However, the above-mentioned energy management methods of the shipborne multi-source flexible DC loop network test bench have some defects that cannot be ignored: first, the rule-based control method relies too much on engineering experience and pre-set rules, and is difficult to adapt to complex and variable conditions; second, although the DP method can achieve global optimization, it has large calculation amount and requires complete condition information to be known in advance, making it difficult to realize real-time control; third, the rule-based control method, the DP method and the model predictive control method generally lack consideration of the degradation characteristics of hydrogen fuel cells and lithium batteries, and cannot adapt to the characteristics of device performance changing with time, thus leading to a decline in control performance and exacerbating device degradation. SUMMARY

[0005] In view of the above defects or improvement needs of the prior art, the present application provides a shipborne multi-source flexible DC loop network test bench and an energy management method thereof, which aims to solve the technical problems that the existing rule-based control method excessively relies on engineering experience and preset rules and is difficult to adapt to complex and variable working conditions, the existing DP method has a large amount of calculation and needs to know the complete working condition information in advance, and is difficult to realize real-time control, and the existing rule-based control method, the DP method and the model predictive control method generally lack consideration of the degradation characteristics of hydrogen fuel cells and lithium batteries, and cannot adapt to the characteristics that the performance of equipment changes over time, thus leading to the technical problems of control performance degradation and aggravation of equipment degradation.

[0006] To achieve the above-mentioned purpose, according to one aspect of the present application, a shipborne multi-source flexible DC loop network test bench is provided, comprising a DC loop network main system, a first hydrogen fuel cell power supply system, a second hydrogen fuel cell power supply system, a first composite energy storage system, a second composite energy storage system, a first load system, a second load system, a compact reconfigurable input / output (CRIO) controller, a lower computer, and an upper computer, wherein the DC loop network main system comprises a first DC bus, a second DC bus, a third DC bus, and a fourth DC bus electrically connected in series through four solid-state switches;

[0007] The first hydrogen fuel cell power supply system is electrically connected to the second DC bus, the second hydrogen fuel cell power supply system is electrically connected to the fourth DC bus, the first composite energy storage system is electrically connected to the first DC bus, and the second composite energy storage system is electrically connected to the third DC bus;

[0008] The first load system is electrically connected to the first DC bus, and the second load system is electrically connected to the third DC bus.

[0009] The CRIO controller is electrically connected to the first hydrogen fuel cell power supply system, the second hydrogen fuel cell power supply system, the first composite energy storage system, the second composite energy storage system, the first load system, and the second load system through a weak electrical signal line, and is also electrically connected to the first DC bus, the second DC bus, the third DC bus, and the fourth DC bus, so as to monitor the working state thereof in real time.

[0010] The lower computer is electrically connected to the second load system through a weak point signal line.

[0011] The upper computer is electrically connected to the CRIO controller and the lower computer through a weak point signal line.

[0012] Preferably, the first hydrogen fuel cell power supply system and the second hydrogen fuel cell power supply system are completely identical.

[0013] The first hydrogen fuel cell power supply system comprises a hydrogen fuel cell, a first DC circuit breaker, a first DC fuse, a first unidirectional boost DC converter, a first current-voltage sensor, a second DC fuse, and a second DC circuit breaker.

[0014] The first hydrogen fuel cell is electrically connected to the second DC bus in sequence through the first DC circuit breaker, the first DC fuse, the first unidirectional boost DC converter, the first current-voltage sensor, the second DC fuse, and the second DC circuit breaker.

[0015] The rated power of the first hydrogen fuel cell is 40 kW, and the output voltage range is 450-700 V.

[0016] The rated voltage of the first DC circuit breaker and the second DC circuit breaker is 1000 V, and the rated current is 63 A.

[0017] The rated voltage of the first DC fuse and the second DC fuse is 1000 V, and the rated current is 63 A.

[0018] The first unidirectional boost DC converter is a unidirectional isolation type converter, the input side voltage range is 400-850 V, the output side voltage range is 400-850 V, and the rated power is 40 kW.

[0019] The current measurement range of the first voltage current sensor is 0-100 A, and the voltage measurement range is 0-1000 V.

[0020] Preferably, the first composite energy storage system and the second composite energy storage system are completely identical.

[0021] The first composite energy storage system comprises a first lithium battery, a first super capacitor, a third DC circuit breaker, a third DC fuse, a first bidirectional boost DC converter, a second current-voltage sensor, a fourth DC fuse, a fourth DC circuit breaker, a fifth DC circuit breaker, a fifth DC fuse, a second bidirectional boost DC converter, a third current-voltage sensor, a sixth DC fuse, and a sixth DC circuit breaker.

[0022] The first lithium battery is electrically connected to the first DC bus in sequence through the third DC circuit breaker, the third DC fuse, the first bidirectional boost DC converter, the second current-voltage sensor, the fourth DC fuse, and the fourth DC circuit breaker.

[0023] The first super capacitor is electrically connected to the first DC bus in sequence through the fifth DC circuit breaker, the fifth DC fuse, the second bidirectional boost DC converter, the third current-voltage sensor, the sixth DC fuse, and the sixth DC circuit breaker.

[0024] The capacity of the first lithium battery is 65 kWh, the nominal voltage is 633 V, the working voltage range is 574-693 V, and the maximum charge and discharge rate is 1 C.

[0025] The capacity of the first super capacitor is 5.2F, the rated power is 80kW, and the working voltage range is 200-800V;

[0026] The rated voltage of the third DC circuit breaker, the fourth DC circuit breaker, the fifth DC circuit breaker and the sixth DC circuit breaker is 1000V, and the rated current is 63A;

[0027] The rated voltage of the third DC fuse, the fourth DC fuse, the fifth DC fuse and the sixth DC fuse is 1000V, and the rated current is 63A;

[0028] The first bidirectional boost DC converter and the second bidirectional boost DC converter are bidirectional non-isolated converters, the A-side voltage range is 400-850V, the B-side voltage range is 300-850V, and the rated power is 60kW.

[0029] The current measurement range of the second voltage and current sensor and the third voltage and current sensor is 0-100A, and the voltage measurement range is 0-1000V.

[0030] Preferably, the first load system comprises a first resistance load cabinet, a seventh DC fuse, a seventh DC circuit breaker, and a fourth voltage and current sensor;

[0031] The first resistance load cabinet is electrically connected to the first DC bus through the seventh DC fuse, the seventh DC circuit breaker and the fourth voltage and current sensor in sequence.

[0032] The rated power of the first resistance load cabinet is 10kW;

[0033] The rated voltage of the seventh DC fuse is 1000V, and the rated current is 63A;

[0034] The rated voltage of the seventh DC circuit breaker is 1000V, and the rated current is 63A;

[0035] The current measurement range of the fourth voltage and current sensor is 0-100A, and the voltage measurement range is 0-1000V.

[0036] The second load system comprises a first dynamometer, a first reversible motor, a first DC / AC inverter, an eighth DC fuse, an eighth DC circuit breaker, and a fifth voltage and current sensor;

[0037] The first dynamometer is electrically connected to the third DC bus through the first reversible motor, the first DC / AC inverter, the eighth DC fuse, the eighth DC circuit breaker and the fifth voltage and current sensor in sequence.

[0038] The first dynamometer is an eddy current dynamometer, and the rated absorption power is 320kW;

[0039] The rated power of the first reversible motor is 75kW, the rated voltage is 380V, and the rated rotating speed is 1500r / min;

[0040] The first DC / AC inverter outputs 380V AC three-phase electricity;

[0041] The rated voltage of the eighth DC fuse is 1000V, and the rated current is 63A;

[0042] The rated voltage of the eighth DC circuit breaker is 1000V, and the rated current is 63A;

[0043] The current measurement range of the fifth voltage and current sensor is 0-100A, and the voltage measurement range is 0-1000V.

[0044] Overall, compared with the prior art, the above technical scheme conceived by the shipborne multi-source flexible DC loop network test bench of the present application can achieve the following beneficial effects:

[0045] (1) The first hydrogen fuel cell power supply system, the second hydrogen fuel cell power supply system, the first composite energy storage system, the second composite energy storage system, the first load system and the second load system are integrated, power flexible flow and reliable distribution are realized through the four-section loop network structure, and a complete test platform is provided for multi-source flexible DC loop network research;

[0046] (2) The design of connecting the four-section DC buses end to end by using solid-state switches supports multiple topology structure transformation and operation mode switching, and can simulate various sailing conditions and fault scenarios;

[0047] (3) The first resistance load cabinet and the first dynamometer system accurately simulate the ship propulsion and service load, and combined with high-precision voltage and current sensors and a layered control architecture, comprehensive verification and optimization of energy management strategies can be realized, which is particularly suitable for evaluating multi-energy coordinated control strategies.

[0048] According to another aspect of the present application, an energy management method for a shipborne multi-source flexible DC loop network test bench is provided, characterized by comprising the following steps:

[0049] (1) The first hydrogen fuel cell power supply system and the second hydrogen fuel cell power supply system generate electric energy and modulate the voltage to 690V, and are connected to the second DC bus and the fourth DC bus respectively, and the first composite energy storage system and the second composite energy storage system are connected to the first DC bus and the third DC bus respectively;

[0050] (2) The host computer loads the power data during the historical sailing of the ship, and transmits the power data to the first load system and the second load system.

[0051] (3) The host computer obtains the voltage signals and current signals of the first hydrogen fuel cell power supply system, the second hydrogen fuel cell power supply system, the first lithium battery and the first super capacitor in the first composite energy storage system, the second lithium battery and the second super capacitor in the second composite energy storage system, and the voltage signals and current signals of the first load system collected by the CRIO controller, generates corresponding power signals according to the respective voltage and current signals, obtains the first hydrogen fuel cell internal resistance signal in the first hydrogen fuel cell power supply system and the second hydrogen fuel cell internal resistance signal in the second hydrogen fuel cell power supply system collected by the CRIO controller, and obtains the speed signal and torque signal of the first dynamometer in the second load system collected by the lower computer, and generates the power signal of the second load system according to the speed signal and torque signal.

[0052] (4) The host computer adds the power signals of the first load system and the power signals of the second load system obtained in step (3) to obtain a total power signal, and inputs the total power signal into a pre-trained working condition recognition model to obtain the typical running working condition of the ship at present.

[0053] (5) The host computer selects a corresponding power prediction model according to the typical running working condition of the ship at present obtained in step (4), and performs power prediction on the total power signal obtained in step (4) to obtain load prediction data (specifically, load prediction data for 20s after the current time).

[0054] (6) The host computer obtains the reference power of the first hydrogen fuel cell, the reference power of the second hydrogen fuel cell, the reference power of the first lithium battery, the reference power of the second lithium battery, the reference power of the second super capacitor, and the reference power of the second super capacitor according to the total power signal obtained in step (4), the load prediction data obtained in step (5), the first hydrogen fuel cell internal resistance signal and the second hydrogen fuel cell internal resistance signal obtained in step (3), the current signal of the first lithium battery in the first composite energy storage system, and the current signal of the second lithium battery in the second composite energy storage system.

[0055] (7) The CRIO controller obtains the reference power of the first hydrogen fuel cell, the reference power of the second hydrogen fuel cell, the reference power of the first lithium battery, the reference power of the second lithium battery, the reference power of the first super capacitor, and the reference power of the second super capacitor from the host computer, and transmits the reference power of the first hydrogen fuel cell to the first one-way boost DC converter in the first hydrogen fuel cell power supply system, so that the first hydrogen fuel cell is kept near its reference power by power tracking control, transmits the reference power of the second hydrogen fuel cell to the second one-way boost DC converter in the second hydrogen fuel cell power supply system, so that the second hydrogen fuel cell is kept near its reference power by power tracking control, transmits the reference power of the first lithium battery to the first bidirectional boost DC converter in the first composite energy storage system, so that the first lithium battery is kept near its reference power by power tracking control, transmits the reference power of the first super capacitor to the second bidirectional boost DC converter in the first composite energy storage system, so that the first super capacitor is kept near its reference power by power tracking control, transmits the reference power of the second lithium battery to the third bidirectional boost DC converter in the second composite energy storage system, so that the second lithium battery is kept near its reference power by power tracking control, and transmits the reference power of the second super capacitor to the fourth bidirectional boost DC converter in the second composite energy storage system, so that the second super capacitor is kept near its reference power by power tracking control.

[0056] (8) The shipborne multi-source flexible DC loop network test bench keeps closed-loop operation, and the process ends.

[0057] Preferably, step (3) is specifically that the host computer multiplies the voltage signal and the current signal of the first load system to obtain the power signal of the first load system, multiplies the voltage signal and the current signal of the first hydrogen fuel cell power supply system to obtain the output power signal of the first hydrogen fuel cell, multiplies the voltage signal and the current signal of the second hydrogen fuel cell power supply system to obtain the output power signal of the second hydrogen fuel cell, multiplies the voltage signal and the current signal of the first lithium battery in the first composite energy storage system to obtain the output power signal of the first lithium battery, multiplies the voltage signal and the current signal of the first super capacitor in the first composite energy storage system to obtain the output power signal of the first super capacitor, multiplies the voltage signal and the current signal of the second lithium battery in the second composite energy storage system to obtain the output power signal of the second lithium battery, multiplies the voltage signal and the current signal of the second super capacitor in the second composite energy storage system to obtain the output power signal of the second super capacitor, and obtains the power signal of the second load system according to the speed signal and the torque signal of the first dynamometer in the second load system and by using the following formula.

[0058]

[0059] wherein T is a torque signal of the first dynamometer in the second load system, and n is a rotational speed signal of the first dynamometer in the second load system.

[0060] The working condition recognition model is obtained by using a support vector machine (SVM) in combination with a rolling time window.

[0061] In the working condition recognition, the key features of 10 data points at the current time and in the past 18 seconds are calculated to determine the working condition category in which the ship is located. The rolling window at time t contains data in the time range of [t-18s, t]. The window is updated every 2 seconds by removing the earliest data point and adding the latest data point, so that there are always 10 data points in the window.

[0062] Preferably, the working condition recognition model is obtained by the following steps:

[0063] (4-1) Obtain an original ship historical load power data set, and preprocess the original ship historical load power data set to obtain a preprocessed ship historical load power data set.

[0064] Specifically, for missing data in the original ship historical load power data set, a linear interpolation method is used to fill in the missing data, that is, a linear estimation is performed based on the effective data points before and after the missing value to obtain a reasonable filling value. For abnormal data in the original ship historical load power data, first, the Z-score method is used to detect abnormal values, that is, the Z-score of each data point in the original ship historical load power data is calculated, and the data points with an absolute value of the Z-score greater than 3 are marked as abnormal values. Then, all abnormal values are corrected by using a moving median based on a 30-second window.

[0065] (4-2) Extract a plurality of key features from the preprocessed ship historical load power data set obtained in step (4-1), which are used to characterize the running state of the ship.

[0066] Specifically, in order to accurately represent the running state of the ship, a rolling time window is used to extract features from the preprocessed ship historical load power data set obtained in step (4-1), and the window length is 20s and the step length is 2s. In each rolling sliding time window, the mean f1, the variance f2 and the ratio of the variance to the mean f3 of the power fluctuation are selected as the key features, wherein the mean f1 reflects the steady-state power level of the ship in the rolling time window, the variance f2 represents the fluctuation degree of the power data deviating from the mean value, which characterizes the intensity of the power fluctuation, and the ratio of the variance to the mean f3 represents the power fluctuation intensity at the unit power level, which reflects the dynamic change characteristics of the power. The calculation formulas of the three are as follows:

[0067]

[0068]

[0069]

[0070] and also:

[0071]

[0072] ΔP i = P i - P i-1 , (i ≥ 2)

[0073]

[0074] where N is the number of sampling points in the time window, P i is the i-th pre-processed ship historical load power data in the pre-processed ship historical load power data set (where i ∈ [1, total number of data in the pre-processed ship historical load power data set]), is the variance of the change amount of the i-th pre-processed ship historical load power data, ΔP i represents the difference of the pre-processed ship historical load data at the i-th time relative to the previous time, is the average value of the change amount of the i-th pre-processed ship historical load power data.

[0075] (4-3) Perform K-Means clustering analysis on all the key features obtained in step (4-2) to obtain multiple typical operating conditions of the ship and their corresponding multiple pre-processed ship historical load power data.

[0076] This step specifically involves inputting the key features obtained in step (4-2) into the K-Means clustering algorithm for automatic classification. By the principle of minimizing intra-class distance and maximizing inter-class distance, the K-Means clustering algorithm can automatically cluster data points with similar features into the same category, thereby obtaining 5 typical operating conditions of the ship, namely complex condition, power gradually increasing condition, power gradually decreasing condition, high power steady state condition and low power steady state condition. After obtaining these typical operating conditions, the pre-processed ship historical load power data corresponding to each condition are integrated to obtain the pre-processed ship historical load power data of the 5 typical operating conditions.

[0077] (4-4) Use the multiple pre-processed ship historical load power data corresponding to each typical operating condition obtained in step (4-3) and their key features to train the operating condition recognition model offline to obtain a trained operating condition recognition model;

[0078] Preferably, the power prediction model corresponding to each typical operating condition is a convolutional neural network-long short-term memory network (CNN-LSTM) model based on an attention mechanism, which is trained by the following steps:

[0079] (5-1) Obtain the plurality of preprocessed ship historical load power data corresponding to the typical operating condition obtained in step (4-3), and construct a training set of the power prediction model corresponding to the typical operating condition according to the ship historical load power data.

[0080] In this step, first, the volatility of each preprocessed ship historical load power data corresponding to the typical operating condition is obtained, and the calculation formula is:

[0081]

[0082] In the formula, P i is the i-th ship historical load power data in all preprocessed ship historical load power data corresponding to the typical operating condition, wherein i∈[2, the total number of all preprocessed ship historical load power data corresponding to the typical operating condition], then, all preprocessed ship historical load power data corresponding to the typical operating condition and its volatility are selected to construct a training set of the power prediction model corresponding to the typical operating condition.

[0083] (5-2) Train the power prediction model corresponding to the typical operating condition obtained in step (5-1) using the training set of the power prediction model to obtain the trained power prediction model corresponding to the typical operating condition.

[0084] Preferably, step (6) specifically includes the following sub-steps:

[0085] (6-1) Decompose the load data using a low-pass filter controller to obtain the reference power of the first super capacitor, the reference power of the second super capacitor, and the low-frequency power.

[0086] (6-2) According to the first hydrogen fuel cell internal resistance signal and the second hydrogen fuel cell internal resistance signal obtained in step (3), obtain the state of health (SOH) of the first hydrogen fuel cell and the SOH of the second hydrogen fuel cell.

[0087] (6-3) According to the SOH of the first hydrogen fuel cell and the SOH of the second hydrogen fuel cell obtained in step (6-2), obtain the hydrogen consumption rate curve of the first hydrogen fuel cell and the hydrogen consumption rate curve of the second hydrogen fuel cell, respectively.

[0088] (6-4) Based on the hydrogen consumption rate data of the first hydrogen fuel cell and the hydrogen consumption rate data of the second hydrogen fuel cell obtained in step (6-3), fitting equations of the hydrogen consumption rate and power of the first hydrogen fuel cell and the second hydrogen fuel cell are respectively obtained.

[0089] (6-5) The fitting equations of the hydrogen consumption rate and power of the first hydrogen fuel cell and the second hydrogen fuel cell obtained in step (6-4), the low frequency power obtained in step (6-1), and the current signals of the first lithium battery and the second lithium battery obtained in step (3) are input into the MPC controller to obtain the reference power of the first hydrogen fuel cell, the second hydrogen fuel cell, the first lithium battery, and the second lithium battery, respectively.

[0090] Preferably, the expressions of the reference power of the supercapacitor and the low frequency power in step (6-1) are respectively:

[0091]

[0092]

[0093] wherein P load is the load data, s is a complex variable, T1 is a filter time constant, P SC is the reference power of the first supercapacitor, which is equal to the reference power of the second supercapacitor, P low is the low frequency power.

[0094] The SOH of the hydrogen fuel cell in step (6-2) is equal to:

[0095]

[0096] wherein R end represents the ohmic internal resistance value of the first / second hydrogen fuel cell at the end of life / Ω; R new represents the ohmic internal resistance value of the first / second hydrogen fuel cell when not in use / Ω; R represents the ohmic internal resistance value of the first / second hydrogen fuel cell in the current state / Ω.

[0097] The hydrogen consumption rate C fc of the hydrogen fuel cell in step (6-3) is calculated by the following formula:

[0098]

[0099] wherein P fc is the output power of the hydrogen fuel cell after passing through the DC / DC converter, is the low heat value of hydrogen. η fc is the system efficiency of the hydrogen fuel cell, η fc is defined as follows:

[0100] ηfc = ε elec * ε conv * η init

[0101] wherein η init is the hydrogen fuel cell system efficiency at the initial state of the hydrogen fuel cell (SOH = 1). The coefficient of variation ε elec represents the degree of reduction of the electric efficiency η elec , and has:

[0102]

[0103] wherein η DC / DC is the conversion efficiency of the DC / DC converter; E auz is the parasitic power of the auxiliary equipment; E stack is the stack output power; the coefficient of variation ε conv represents the degree of reduction of the conversion efficiency η conv , and has:

[0104] ε conv = 0.9 + 0.1 * SOH

[0105] In step (6-4), the hydrogen fuel cell hydrogen consumption rate data calculated according to step (6-3) is fitted by using a quadratic polynomial fitting method to obtain a fitting equation of the hydrogen consumption rate C fc .

[0106]

[0107] wherein a, b, c are fitting coefficients of the hydrogen fuel cell.

[0108] In step (6-5), the objective function and the constraint conditions of the MPC controller are as follows:

[0109]

[0110]

[0111] wherein C fc1 and C fc2 are the hydrogen consumption rate of the first hydrogen fuel cell and the hydrogen consumption rate of the first hydrogen fuel cell, respectively; C batt is the equivalent hydrogen consumption rate of the first / second lithium battery; k fc_deg and k Batt_deg are degradation weights, used to represent the optimization bias of the degradation of the hydrogen fuel cell and the lithium battery in the overall performance of the power system, and the value range is 0 to 1, k fc_deg is preferably 0.6, and k Batt_deg is preferably 0.4; C fc_deg is the equivalent hydrogen consumption amount of the degradation of the hydrogen fuel cell; and C batt_degis the degradation cost of the first / second lithium battery; P fc1 is the reference power of the first hydrogen fuel cell, P fc1 is the reference power of the second hydrogen fuel cell, P batt is the reference power of the first lithium battery, which is equal to the reference power of the second lithium battery; SOC batt is the state of charge of the first / second lithium battery, d is the degradation factor, and the calculation formula is: d = a * SOH, wherein a is the energy efficiency optimization factor of the hydrogen fuel cell, and the value determines the sensitivity of the adjustment range of the output power of the hydrogen fuel cell to the influence of its degradation.

[0112] Overall, the above technical solutions conceived by the energy management method of the shipborne multi-source flexible DC loop network test bench of the present application can achieve the following beneficial effects compared with the prior art:

[0113] (1) The present application adopts steps (4) and (5), which recognize the working condition by combining support vector machine with rolling time window, and select the corresponding CNN-LSTM prediction model based on attention mechanism according to the recognized working condition to predict the load power, so as to solve the technical problem that the existing rule type control method excessively relies on engineering experience and preset rules, and is difficult to adapt to complex and variable working conditions;

[0114] (2) The present application adopts step (6), which decomposes the load data into high and low frequency parts, and uses model predictive control to solve the rolling optimization, so as to solve the technical problem that the existing dynamic programming method needs to know the complete working condition information, has large calculation amount and is difficult to realize real-time control;

[0115] (3) The present application adopts steps (6-2), (6-3) and (6-4), which obtain the health state by measuring the internal resistance of the hydrogen fuel cell, calculate the hydrogen consumption rate under different health states, and include the degradation cost of the hydrogen fuel cell and the degradation cost of the lithium battery into the optimization objective function, so as to solve the technical problem that the existing energy management method generally lacks consideration of the degradation characteristics of the hydrogen fuel cell and the lithium battery. BRIEF DESCRIPTION OF DRAWINGS

[0116] Figure 1 is a structural schematic diagram of the shipborne multi-source flexible DC loop network test bench of the present application;

[0117] Figure 2 is a structural schematic diagram of the first hydrogen fuel cell power supply system in the present application;

[0118] Figure 3 is a structural schematic diagram of the first composite energy storage system in the present application;

[0119] Figure 4 is a structural schematic diagram of the first load system in the present application;

[0120] Figure 5 is a structural schematic diagram of a second load system in the application;

[0121] Figure 6 is a single-line diagram of a shipborne multi-source flexible DC loop network test bench power supply system of the application;

[0122] Figure 7 is a schematic diagram of an energy management method of a shipborne multi-source flexible DC loop network test bench power supply system of the application;

[0123] Figure 8 is a working condition recognition window diagram provided by an embodiment of the application;

[0124] Figure 9 is an efficiency and hydrogen consumption rate curve diagram under different SOH provided by an embodiment of the application. DETAILED DESCRIPTION

[0125] In order to make the objectives, technical solutions and advantages of the application clearer, the application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the application and do not limit the application. In addition, the technical features involved in each embodiment of the application described below can be combined with each other as long as they do not conflict with each other.

[0126] As shown in Figure 1 , the application provides a shipborne multi-source flexible DC loop network test bench, which comprises a DC loop network main system, a first hydrogen fuel cell power supply system 1, a second hydrogen fuel cell power supply system 2, a first composite energy storage system 3, a second composite energy storage system 4, a first load system 5, a second load system 6, a compact reconfigurable input / output (Compact RIO, referred to as CRIO) controller 12, a lower computer 13, and an upper computer 14.

[0127] The DC loop network main system comprises a first DC bus 8, a second DC bus 9, a third DC bus 10, and a fourth DC bus 11 which are electrically connected in series through four solid-state switches 7.

[0128] The first hydrogen fuel cell power supply system 1 is electrically connected with the second DC bus 9, the second hydrogen fuel cell power supply system 2 is electrically connected with the fourth DC bus 11, the first composite energy storage system 3 is electrically connected with the first DC bus 8, and the second composite energy storage system 4 is electrically connected with the third DC bus 10.

[0129] The first load system 5 is electrically connected with the first DC bus 8, and the second load system 6 is electrically connected with the third DC bus 10.

[0130] The CRIO controller 12 is electrically connected to the first hydrogen fuel cell power supply system 1, the second hydrogen fuel cell power supply system 2, the first composite energy storage system 3, the second composite energy storage system 4, the first load system 5, and the second load system 6 via low-voltage signal lines (shown by dashed lines in the figure), and is also electrically connected to the first DC bus 8, the second DC bus 9, the third DC bus 10, and the fourth DC bus 11 to monitor their operating status in real time.

[0131] The lower-level machine 13 is electrically connected to the second load system 6 through a weak signal line (shown by the dotted line in the figure).

[0132] The host computer 14 is electrically connected to the CRIO controller 12 and the slave computer 13 through a weak signal line (shown by the dotted line in the figure).

[0133] The first hydrogen fuel cell power supply system 1 and the second hydrogen fuel cell power supply system 2 are completely identical. For example... Figure 2 As shown, the first hydrogen fuel cell power supply system 1 includes a hydrogen fuel cell 1-1, a first DC circuit breaker 1-2, a first DC fuse 1-3, a first unidirectional boost DC converter 1-4, a first current and voltage sensor 1-5, a second DC fuse 1-6, and a second DC circuit breaker 1-7. The first hydrogen fuel cell 1-1 is electrically connected to the second DC bus 9 via the first DC circuit breaker 1-2, the first DC fuse 1-3, the first unidirectional boost DC converter 1-4, the first current and voltage sensor 1-5, the second DC fuse 1-6, and the second DC circuit breaker 1-7 in sequence.

[0134] Specifically, the first hydrogen fuel cell 1-1 has a rated power of 40kW and an output voltage range of 450-700V; the first DC circuit breaker 1-2 and the second DC circuit breaker 1-7 have a rated voltage of 1000V and a rated current of 63A; the first DC fuse 1-3 and the second DC fuse 1-6 have a rated voltage of 1000V and a rated current of 63A; the first unidirectional boost DC converter 1-4 is a unidirectional isolated converter with an input voltage range of 400-850V, an output voltage range of 400-850V, and a rated power of 40kW; the first voltage and current sensor 1-5 has a current measurement range of 0-100A and a voltage measurement range of 0-1000V.

[0135] The first composite energy storage system 3 and the second composite energy storage system 4 are completely identical. For example... Figure 3As shown, the first composite energy storage system 3 includes a first lithium battery 3-1, a first super capacitor 3-2, a third DC circuit breaker 3-3, a third DC fuse 3-4, a first bidirectional boost DC converter 3-5, a second current-voltage sensor 3-6, a fourth DC fuse 3-7, a fourth DC circuit breaker 3-8, a fifth DC circuit breaker 3-9, a fifth DC fuse 3-10, a second bidirectional boost DC converter 3-11, a third current-voltage sensor 3-12, a sixth DC fuse 3-13, and a sixth DC circuit breaker 3-14. The first lithium battery 3-3 is electrically connected to the first DC bus 8 in sequence through the third DC circuit breaker 3-3, the third DC fuse 3-4, the first bidirectional boost DC converter 3-5, the second current-voltage sensor 3-6, the fourth DC fuse 3-7, and the fourth DC circuit breaker 3-8. The first super capacitor 3-2 is electrically connected to the first DC bus 8 in sequence through the fifth DC circuit breaker 3-9, the fifth DC fuse 3-10, the second bidirectional boost DC converter 3-11, the third current-voltage sensor 3-12, the sixth DC fuse 3-13, and the sixth DC circuit breaker 3-14.

[0136] Specifically, the capacity of the first lithium battery 3-3 is 65 kWh, the nominal voltage is 633 V, the working voltage range is 574-693 V, and the maximum charge-discharge rate is 1C. The capacity of the first super capacitor 3-4 is 5.2 F, the rated power is 80 kW, and the working voltage range is 200-800 V. The rated voltage of the third DC circuit breaker 3-3, the fourth DC circuit breaker 3-8, the fifth DC circuit breaker 3-9, and the sixth DC circuit breaker 3-14 is 1000 V, and the rated current is 63 A. The rated voltage of the third DC fuse 3-4, the fourth DC fuse 3-7, the fifth DC fuse 3-10, and the sixth DC fuse 3-13 is 1000 V, and the rated current is 63 A. The first bidirectional boost DC converter 3-5 and the second bidirectional boost DC converter 3-11 are bidirectional non-isolated converters, the A-side voltage range is 400-850 V, the B-side voltage range is 300-850 V, and the rated power is 60 kW. The current measurement range of the second voltage-current sensor 3-6 and the third voltage-current sensor 3-12 is 0-100 A, and the voltage measurement range is 0-1000 V.

[0137] As shown in Figure 4 The first load system 5 includes a first resistance load cabinet 5-1, a seventh DC fuse 5-2, a seventh DC circuit breaker 5-3, and a fourth voltage-current sensor 5-4. The first resistance load cabinet 5-1 is electrically connected to the first DC bus 8 in sequence through the seventh DC fuse 5-2, the seventh DC circuit breaker 5-3, and the fourth voltage-current sensor 5-4.

[0138] Specifically, the rated power of the first resistance load cabinet 5-1 is 10kW, the rated voltage of the seventh DC fuse 5-2 is 1000V, and the rated current is 63A; the rated voltage of the seventh DC circuit breaker 5-3 is 1000V, and the rated current is 63A; the current measurement range of the fourth voltage and current sensor 5-4 is 0-100A, and the voltage measurement range is 0-1000V.

[0139] As shown in Figure 5 The second load system 6 includes a first dynamometer 6-1, a first reversible motor 6-2, a first DC / AC inverter 6-3, an eighth DC fuse 6-4, an eighth DC circuit breaker 6-5, and a fifth voltage and current sensor 6-6.

[0140] Specifically, the first dynamometer 6-1 is an eddy current dynamometer, and the rated absorption power is 320kW; the rated power of the first reversible motor 6-2 is 75kW, the rated voltage is 380V, and the rated speed is 1500r / min; the first DC / AC inverter 6-3 outputs 380V AC three-phase power; the rated voltage of the eighth DC fuse 6-4 is 1000V, and the rated current is 63A; the rated voltage of the eighth DC circuit breaker 6-5 is 1000V, and the rated current is 63A; the current measurement range of the fifth voltage and current sensor 6-6 is 0-100A, and the voltage measurement range is 0-1000V.

[0141] As shown in Figure 6 It is a single-line diagram of the power supply system of the shipborne multi-source flexible DC loop network test bench of the application.

[0142] The working principle of the application is as follows:

[0143] First, the first hydrogen fuel cell power supply system, the second hydrogen fuel cell power supply system 2 generates electric energy and modulates the boost to 690V, respectively access the second DC bus 9 and the fourth DC bus 11. The first composite energy storage system 3, the second composite energy storage system 4 are respectively connected to the first DC bus 8, the third DC bus 10. The host computer 14 according to the output voltage and output current of each hydrogen fuel cell power supply system, each composite energy storage system, the voltage and current of each DC bus, the voltage and current of the first load system 5, the actual load of the second load system 6 collected by the lower computer 13, implement the corresponding control strategy. At the same time, the host computer 14 transmits the actual load and daily load power command to the CRIO controller 12, and the CRIO controller 12 converts the actual power into speed and torque control command and transmits it to the second load system 6; the CRIO controller 12 controls the daily load control command to the first load system 5.

[0144] As shown in Figure 7 The present application also provides an energy management method for the above-mentioned shipborne multi-source flexible DC loop network test bench, comprising the following steps:

[0145] (1) The first hydrogen fuel cell power supply system 1 and the second hydrogen fuel cell power supply system 2 generate electric energy and modulate the boost to 690V, and are respectively connected to the second DC bus 9 and the fourth DC bus 11. The first composite energy storage system 3 and the second composite energy storage system 4 are respectively connected to the first DC bus 8 and the third DC bus 10.

[0146] (2) The host computer 14 loads the power data during the historical navigation of the ship, and transmits the power data to the first load system 5 and the second load system 6.

[0147] Specifically, the power data during the historical navigation of the ship includes driving load power and service load power. The host computer 14 transmits the loaded driving load power and service load power to the CRIO controller 12, which converts the driving load power into torque control instructions and the service load power into power control instructions, and then transmits the torque control instructions to the first reversible motor in the second load system 6 and the power control instructions to the first resistance load cabinet in the first load system 5.

[0148] More specifically, the first reversible motor adjusts the motor output torque according to the torque control instructions of the CRIO controller 12, simulates the propeller load change when the ship is sailing, and realizes the simulation of the ship propulsion load; the first resistance load cabinet automatically adjusts the internal resistance value according to the power control instructions transmitted by the CRIO controller 12, controls the power consumption by changing the resistance value, and realizes the simulation of the ship lighting, communication and other service power load.

[0149] (3) The host computer 14 obtains the voltage signal and the current signal of the first hydrogen fuel cell power supply system 1, the second hydrogen fuel cell power supply system 2, the first lithium battery and the first super capacitor in the first composite energy storage system 3, the voltage signal and the current signal of the second lithium battery and the second super capacitor in the second composite energy storage system 4, and the voltage signal and the current signal of the first load system 5 collected by the CRIO controller 12, generates the corresponding power signal according to the respective voltage and current signal, obtains the first hydrogen fuel cell internal resistance signal in the first hydrogen fuel cell power supply system 1 and the second hydrogen fuel cell internal resistance signal in the second hydrogen fuel cell power supply system 2 collected by the CRIO controller 12, and obtains the speed signal and the torque signal of the first dynamometer in the second load system 6 collected by the lower computer 13, and generates the power signal of the second load system 6 according to the speed signal and the torque signal.

[0150] Specifically, the host computer 14 multiplies the voltage signal and the current signal of the first load system 5 to obtain the power signal of the first load system 5, multiplies the voltage signal and the current signal of the first hydrogen fuel cell power supply system 1 to obtain the output power signal of the first hydrogen fuel cell, multiplies the voltage signal and the current signal of the second hydrogen fuel cell power supply system 2 to obtain the output power signal of the second hydrogen fuel cell, multiplies the voltage signal and the current signal of the first lithium battery in the first composite energy storage system 3 to obtain the output power signal of the first lithium battery, multiplies the voltage signal and the current signal of the first super capacitor in the first composite energy storage system 3 to obtain the output power signal of the first super capacitor, multiplies the voltage signal and the current signal of the second lithium battery in the second composite energy storage system 4 to obtain the output power signal of the second lithium battery, multiplies the voltage signal and the current signal of the second super capacitor in the second composite energy storage system 4 to obtain the output power signal of the second super capacitor, and obtains the power signal of the second load system 6 according to the speed signal and the torque signal of the first dynamometer in the second load system 6 and using the following formula.

[0151]

[0152] Wherein, T is the torque signal of the first dynamometer in the second load system 6, and n is the speed signal of the first dynamometer in the second load system 6.

[0153] More specifically, the first dynamometer generates a braking torque by electromagnetic induction principle and absorbs the mechanical energy output by the first reversible motor. When the first reversible motor drives the first dynamometer rotor to rotate, the excitation coil in the stator generates a magnetic field, and the rotor cuts the magnetic induction line to induce eddy current, thereby generating a braking torque related to the speed. The first dynamometer obtains the output torque in real time through the built-in torque sensor and obtains the rotor speed in real time through the magneto-electric sensor.

[0154] (4) The host computer 14 adds the power signal of the first load system 5 and the power signal of the second load system 6 obtained in step (3) to obtain a total power signal, and inputs the total power signal into the pre-trained working condition recognition model to obtain the typical running working condition of the ship at present.

[0155] Specifically, the working condition recognition model of the present application is a support vector machine (SVM) combined with a rolling time window for working condition recognition. The rolling time window dynamically analyzes the power data during the historical voyage of the ship within a set time range by segmenting the power data, and continuously updates the power data in the time window to improve the accuracy and reliability of working condition recognition.

[0156] More specifically, Figure 8 For the working condition recognition rolling time window, the time window T l = 20s, t is the discrete time step, which is set to 2s in the present application, that is, when recognizing the working condition, the key features of 10 data points at the current time and the previous 18s are calculated to determine the working condition category of the current ship. The rolling window at time t contains data within the [t-18s, t] time range, and the window is updated every 2s by removing the earliest data point (at time t-18s) and adding the latest data point (at time t+2), so that there are always 10 data points in the window, ensuring data continuity and real-time performance.

[0157] The working condition recognition model in this step is trained by the following steps:

[0158] (4-1) Obtain the original ship historical load power data set, and pre-process the original ship historical load power data set (the purpose is to remove abnormal and missing values caused by sensor failure, communication interference or other factors) to obtain the pre-processed ship historical load power data set;

[0159] Specifically, for the missing data in the original ship historical load power data set, a linear interpolation method is used to fill in the missing data, that is, a linear estimation is made based on the valid data points before and after the missing value to obtain a reasonable filling value. For abnormal data in the original ship historical load power data, first, use the Z-score method to detect abnormal values, that is, calculate the Z-score (the difference between the current value and the mean value divided by the standard deviation) of each data point in the original ship historical load power data, and mark the data points with an absolute value greater than 3 as abnormal values. Then, all abnormal values are corrected based on a 30-second window moving median (specifically, the median of the data 15 seconds before and after the abnormal point is used to replace the abnormal value).

[0160] (4-2) Extracting a plurality of key features from the pre-processed ship historical load power data set obtained in step (4-1), which are used to characterize the operating state of the ship.

[0161] Specifically, in order to accurately characterize the operating state of the ship, the pre-processed ship historical load power data set obtained in step (4-1) is subjected to feature extraction using a rolling time window, with a window length of 20s and a step length of 2s. Within each rolling sliding time window, the mean f1, variance f2 and variance-to-mean ratio f3 of power fluctuations are selected as key features, where the mean f1 reflects the steady-state power level of the ship within the rolling time window, the variance f2 characterizes the degree of fluctuation of the power data from the mean value, which characterizes the intensity of the power fluctuations, and the variance-to-mean ratio f3 characterizes the power fluctuation intensity at unit power level, which reflects the dynamic change characteristics of the power. The calculation formulae of the three are:

[0162]

[0163]

[0164]

[0165] and

[0166]

[0167] ΔP i = P i -P i-1 , (i ≥ 2)

[0168]

[0169] where N is the number of sampling points in the time window, P i is the i-th pre-processed ship historical load power data in the pre-processed ship historical load power data set (where i ∈ [1, total number of data in the pre-processed ship historical load power data set]), is the variance of the change amount of the i-th pre-processed ship historical load power data, ΔP i represents the difference of the pre-processed ship historical load data at the i-th time relative to the previous time, is the mean value of the change amount of the i-th pre-processed ship historical load power data.

[0170] (4-3) K-Means clustering analysis is performed on all the key features obtained in step (4-2) to obtain a plurality of typical operating conditions of the ship and a plurality of pre-processed ship historical load power data corresponding thereto.

[0171] Specifically, the key features obtained in step (4-2) are input into a K-Means clustering algorithm for automatic classification. According to the principle of minimizing intra-class distance and maximizing inter-class distance, the K-Means clustering algorithm can automatically cluster data points with similar features into the same category, thereby obtaining five typical operating conditions of the ship, i.e., a complex condition, a power gradually increasing condition, a power gradually decreasing condition, a high-power steady-state condition, and a low-power steady-state condition. After obtaining these typical operating conditions, the preprocessed historical load power data of the ship corresponding to each operating condition are integrated to obtain the preprocessed historical load power data of the five typical operating conditions.

[0172] (4-4) The multiple preprocessed historical load power data of the ship corresponding to each typical operating condition obtained in step (4-3) and the key features thereof are used to train the operating condition recognition model offline to obtain a trained operating condition recognition model.

[0173] Specifically, the preprocessed historical load power data of the ship corresponding to the five typical operating conditions and the key features thereof are constructed into a training set, and the constructed training set is input into an SVM. Each sample in the training set is mapped to a high-dimensional feature space through a kernel function, and an optimal classification hyperplane is constructed in the high-dimensional feature space to realize the training of the operating condition recognition model, and finally a trained operating condition recognition model is obtained.

[0174] (5) The host computer 14 selects the power prediction model corresponding to the current typical operating condition of the ship obtained in step (4) to perform power prediction on the total power signal obtained in step (4) to obtain load prediction data (specifically, load prediction data for 20 seconds after the current time).

[0175] Specifically, the power prediction model of the present application is a Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) model based on an attention mechanism, which can realize prediction of load power in the next 20 seconds.

[0176] The power prediction model corresponding to each typical operating condition of the present application is trained by the following steps:

[0177] (5-1) Obtain the multiple preprocessed historical load power data of the ship corresponding to the typical operating condition obtained in step (4-3), and construct a training set of the power prediction model corresponding to the typical operating condition according to the historical load power data of the ship.

[0178] Specifically, first, the volatility of each preprocessed historical load power data of the ship corresponding to the typical operating condition is obtained, and the calculation formula is:

[0179]

[0180] P i is the i-th ship historical load power data in all pre-processed ship historical load power data corresponding to the typical operating condition, wherein i∈[2, the total number of all pre-processed ship historical load power data corresponding to the typical operating condition], then, all pre-processed ship historical load power data corresponding to the typical operating condition and its fluctuation rate are selected to construct a training set of the power prediction model corresponding to the typical operating condition.

[0181] (5-2) training the power prediction model corresponding to the typical operating condition obtained in step (5-1) using the training set of the power prediction model to obtain the trained power prediction model corresponding to the typical operating condition.

[0182] The steps (4) to (5) have the advantage that an innovative operating condition identification and power prediction system is constructed: the mean, variance and other multi-dimensional features are used to automatically classify the five typical operating conditions such as complex operating conditions and power gradually changing operating conditions by K-Means clustering, the ship operating condition is identified in real time by the support vector machine combined with the rolling time window method, the CNN-LSTM prediction model based on the attention mechanism is used for special power prediction for different operating conditions, and the accurate prediction of the future 20-second load power is realized. The classification prediction method based on operating condition identification not only improves the adaptability of the system to the complex navigation environment, but also improves the prediction accuracy by considering the power fluctuation and other features, and provides reliable data support for energy management decision-making.

[0183] (6) The host computer 14 obtains the reference power of the first hydrogen fuel cell, the reference power of the second hydrogen fuel cell, the reference power of the first lithium battery, the reference power of the second lithium battery, the reference power of the second super capacitor, and the reference power of the second super capacitor according to the total power signal obtained in step (4), the load prediction data obtained in step (5), the first hydrogen fuel cell internal resistance signal and the second hydrogen fuel cell internal resistance signal obtained in step (3), the current signal of the first lithium battery in the first composite energy storage system 3, and the current signal of the second lithium battery in the second composite energy storage system 4.

[0184] Specifically, this step first combines the total power signal obtained in step (4) and the load prediction data obtained in step (5) to obtain complete load data, then decomposes the load data into high-frequency power and low-frequency power using a low-pass filter controller, wherein the high-frequency power is borne by the super capacitor, and finally inputs the hydrogen consumption rate data of the hydrogen fuel cell, the internal resistance data of the hydrogen fuel cell, and the low-frequency power into a model predictive control (MPC) controller to obtain the reference power of the hydrogen fuel cell and the lithium battery by setting a multi-objective function including hydrogen consumption rate minimization, hydrogen fuel cell life maximization, and lithium battery service life maximization.

[0185] This step (6) specifically includes the following sub-steps:

[0186] (6-1) decompose the load data using a low-pass filter controller to obtain the reference power of the first super capacitor, the reference power of the second super capacitor, and the low-frequency power.

[0187] Specifically, the expressions of the reference power of the super capacitor and the low-frequency power are respectively:

[0188]

[0189]

[0190] wherein P load is the load data, s is a complex variable, T1 is a filter time constant, P SC is the reference power of the first super capacitor, which is equal to the reference power of the second super capacitor, and P low is the low-frequency power.

[0191] (6-2) obtain the State of Health (SOH) of the first hydrogen fuel cell and the SOH of the second hydrogen fuel cell according to the first hydrogen fuel cell internal resistance signal and the second hydrogen fuel cell internal resistance signal obtained in step (3).

[0192] Specifically, since the output power of the hydrogen fuel cell under different operating conditions shows dynamic variation characteristics, the hydrogen fuel cell has different system efficiencies under different SOHs, and the hydrogen fuel cell often uses ohmic resistance as a representation value of its health state, and its calculation formula is as follows:

[0193]

[0194] wherein R end represents the ohmic resistance value of the first / second hydrogen fuel cell at the end of its life / Ω; R newR represents the ohmic internal resistance of the first / second hydrogen fuel cell when it is not in use (Ω); R represents the ohmic internal resistance of the first / second hydrogen fuel cell in its current state (Ω).

[0195] (6-3) Based on the SOH of the first hydrogen fuel cell and the SOH of the second hydrogen fuel cell obtained in step (6-2), the hydrogen consumption rate curve of the first hydrogen fuel cell and the hydrogen consumption rate curve of the second hydrogen fuel cell are obtained respectively.

[0196] Specifically, the hydrogen consumption rate C of a hydrogen fuel cell fc It is calculated using the following formula:

[0197]

[0198] Where P fc The output power of the hydrogen fuel cell after passing through the DC / DC converter, η is the lower heating value of hydrogen. fc For the system efficiency of hydrogen fuel cells, η fc The definition is as follows:

[0199] η fc =ε elec *ε conv *η init

[0200] Where η init The efficiency of the hydrogen fuel cell system is given by the initial state (SOH = 1). The coefficient of variation is ε. elec Indicates electrical efficiency η elec The degree of reduction (determined by the efficiency of the bidirectional boost DC-DC converter and the parasitic power of the auxiliary equipment) is as follows:

[0201]

[0202] Where η DC / DC E represents the conversion efficiency of a DC / DC converter. auz E represents the parasitic power of auxiliary equipment. stack The output power of the fuel cell stack; coefficient of variation ε conv Indicates the conversion efficiency η conv (It is defined as the ratio between the electrical energy generated by a hydrogen fuel cell and the chemical energy of the hydrogen consumed by the hydrogen fuel cell, which decreases as the performance of the hydrogen fuel cell degrades) the degree of reduction, and includes:

[0203] ε conv =0.9 + 0.1 * SOH

[0204] More specifically, such as Figure 9 As shown, by calculating the hydrogen consumption rate of hydrogen fuel cells under different SOH conditions, all the hydrogen consumption rates were fitted to obtain hydrogen consumption rate curves.

[0205] (6-4) According to the hydrogen consumption rate data of the first hydrogen fuel cell and the hydrogen consumption rate data of the second hydrogen fuel cell obtained in step (6-3), a fitting equation of the hydrogen consumption rate and the power of the first hydrogen fuel cell and the second hydrogen fuel cell is obtained respectively.

[0206] Specifically, according to the hydrogen fuel cell hydrogen consumption rate data calculated in step (6-3), a fitting equation of the hydrogen consumption rate C fc is obtained by using a quadratic polynomial fitting method, and the fitting equation is as follows:

[0207]

[0208] Wherein a, b, and c are fitting coefficients of the hydrogen fuel cell.

[0209] The advantages of the above sub-steps (6-2) to (6-3) are that an innovative hydrogen fuel cell performance monitoring and hydrogen consumption rate calculation system is constructed: the health state is calculated by monitoring the ohmic resistance of the hydrogen fuel cell in real time, a system efficiency model considering the efficiency of the bidirectional direct current converter, the parasitic power of the auxiliary equipment and the performance degradation of the stack is established, and the quantitative relationship between the hydrogen consumption rate and the power is obtained by using a quadratic polynomial fitting method. The complete evaluation chain from health state monitoring to hydrogen consumption rate fitting not only realizes the accurate quantification of the performance degradation of the hydrogen fuel cell, but also provides a reliable mathematical model for the subsequent energy management method, effectively supporting the economic scheduling decision based on performance degradation.

[0210] (6-5) The fitting equation of the hydrogen consumption rate and the power of the first hydrogen fuel cell and the second hydrogen fuel cell obtained in step (6-4), the low-frequency power obtained in step (6-1), and the current signal of the first lithium battery and the current signal of the second lithium battery obtained in step (3) are input into the MPC controller to obtain the reference power of the first hydrogen fuel cell, the second hydrogen fuel cell, the first lithium battery and the second lithium battery respectively.

[0211] Specifically, the optimization objective function of the MPC controller is to minimize the overall hydrogen consumption of the system while taking into account the performance degradation of the hydrogen fuel cell and the lithium battery.

[0212] The objective function and the constraint condition of the MPC controller are as follows:

[0213]

[0214]

[0215] Wherein C fc1 and C fc2 are the hydrogen consumption rate of the first hydrogen fuel cell and the hydrogen consumption rate of the first hydrogen fuel cell respectively; C battis the equivalent hydrogen consumption rate of the first / second lithium battery; k fc_deg and k Batt_deg is the degradation weight, representing the optimization bias of hydrogen fuel cell and lithium battery degradation in the overall performance of the power system, with a value range of 0 to 1, k fc_deg preferably 0.6, k Batt_deg preferably 0.4; C fc_deg is the equivalent hydrogen consumption of hydrogen fuel cell degradation; C batt_deg is the degradation cost of the first / second lithium battery; P fc1 is the reference power of the first hydrogen fuel cell, P fc1 is the reference power of the second hydrogen fuel cell, P batt / 2 is the reference power of the first lithium battery, which is equal to the reference power of the second lithium battery. SOC batt is the state of charge of the first / second lithium battery, d is the degradation factor, and its calculation formula is: d = a * SOH, where a is the energy efficiency optimization factor of the hydrogen fuel cell, which determines the sensitivity of the output power adjustment range of the hydrogen fuel cell to the influence of its degradation.

[0216] The advantage of this sub-step (6-5) is that a multi-objective optimization energy management strategy based on the MPC controller is proposed: by designing a target function containing system hydrogen consumption minimization, hydrogen fuel cell life maximization and lithium battery service life maximization, and introducing a degradation weight coefficient to balance the performance degradation of different power sources, the coordinated optimization of hydrogen fuel cell and lithium battery reference power is realized. This strategy not only considers the economy of system operation, but also realizes the active management of power source performance decline through the introduction of degradation factor, achieving the dual goals of improving the overall efficiency of the system and prolonging the service life of the equipment.

[0217] (7) The CRIO controller 12 obtains the reference power of the first hydrogen fuel cell, the reference power of the second hydrogen fuel cell, the reference power of the first lithium battery, the reference power of the second lithium battery, the reference power of the first super capacitor, and the reference power of the second super capacitor obtained by the host computer 14, and transmits the reference power of the first hydrogen fuel cell to the first one-way boost DC converter in the first hydrogen fuel cell power supply system, so that the first hydrogen fuel cell is kept near its reference power by power tracking control, transmits the reference power of the second hydrogen fuel cell to the second one-way boost DC converter in the second hydrogen fuel cell power supply system, so that the second hydrogen fuel cell is kept near its reference power by power tracking control, transmits the reference power of the first lithium battery to the first bidirectional boost DC converter in the first composite energy storage system, so that the first lithium battery is kept near its reference power by power tracking control, transmits the reference power of the first super capacitor to the second bidirectional boost DC converter in the first composite energy storage system, so that the first super capacitor is kept near its reference power by power tracking control, transmits the reference power of the second lithium battery to the third bidirectional boost DC converter in the second composite energy storage system, so that the second lithium battery is kept near its reference power by power tracking control, and transmits the reference power of the second super capacitor to the fourth bidirectional boost DC converter in the second composite energy storage system, so that the second super capacitor is kept near its reference power by power tracking control.

[0218] (8) The shipborne multi-source flexible DC loop network test bench keeps closed-loop operation, and the process ends.

[0219] Those skilled in the art will readily understand that the above description is only the preferred embodiment of the present application, and is not intended to limit the present application, and any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An energy management method for a shipborne multi-source flexible DC ring network test bench, the shipborne multi-source flexible DC ring network test bench comprising a DC ring network main system, a first hydrogen fuel cell power supply system, a second hydrogen fuel cell power supply system, a first composite energy storage system, a second composite energy storage system, a first load system, a second load system, a compact reconfigurable input / output (CRIO) controller, a lower computer, and an upper computer, the DC ring network main system comprising a first DC bus, a second DC bus, a third DC bus, and a fourth DC bus electrically connected in a head-tail manner through four solid-state switches, the first hydrogen fuel cell power supply system being electrically connected to the second DC bus, the second hydrogen fuel cell power supply system being electrically connected to the fourth DC bus, the first composite energy storage system being electrically connected to the first DC bus, the second composite energy storage system being electrically connected to the third DC bus, the first load system being electrically connected to the first DC bus, the second load system being electrically connected to the third DC bus, the CRIO controller being electrically connected to the first hydrogen fuel cell power supply system, the second hydrogen fuel cell power supply system, the first composite energy storage system, the second composite energy storage system, the first load system, and the second load system through weak electrical signal lines, and being electrically connected to the first DC bus, the second DC bus, the third DC bus, and the fourth DC bus at the same time to monitor the working state in real time, the lower computer being electrically connected to the second load system through a weak electrical signal line, and the upper computer being electrically connected to the CRIO controller and the lower computer through a weak electrical signal line; characterized in that, The energy management method includes the following steps: (1) The first hydrogen fuel cell power supply system and the second hydrogen fuel cell power supply system generate electrical energy and modulate and boost it to 690V, and connect to the second DC bus and the fourth DC bus respectively. The first composite energy storage system and the second composite energy storage system are connected to the first DC bus and the third DC bus respectively. (2) The host computer loads the power data of the ship during its historical voyage and transmits the power data to the first load system and the second load system; (3) The host computer acquires the voltage and current signals of the first hydrogen fuel cell power supply system, the second hydrogen fuel cell power supply system, the first lithium battery and the first supercapacitor in the first composite energy storage system, the second lithium battery and the second supercapacitor in the second composite energy storage system, and the voltage and current signals of the first load system collected by the CRIO controller. Based on their respective voltage and current signals, the host computer generates corresponding power signals. It also acquires the internal resistance signal of the first hydrogen fuel cell in the first hydrogen fuel cell power supply system and the internal resistance signal of the second hydrogen fuel cell in the second hydrogen fuel cell power supply system collected by the CRIO controller. Furthermore, it acquires the speed and torque signals of the first dynamometer in the second load system collected by the host computer. Based on these speed and torque signals, the host computer generates the power signal of the second load system. (4) The host computer adds the power signal of the first load system and the power signal of the second load system obtained in step (3) to obtain the total power signal, and inputs the total power signal into the pre-trained working condition recognition model to obtain the ship's current typical operating conditions. The working condition recognition model is trained through the following steps: (4-1) Obtain the original ship historical load power dataset, and preprocess the original ship historical load power dataset to obtain the preprocessed ship historical load power dataset. (4-2) Extract several key features from the preprocessed ship historical load power dataset obtained in step (4-1), which are used to characterize the ship's operating status. (5) The host computer selects the corresponding power prediction model to perform power prediction on the total power signal obtained in step (4) based on the typical operating conditions of the ship obtained in step (4) in order to obtain load prediction data, specifically the load prediction data 20 seconds after the current time. (6) The host computer obtains the reference power of the first hydrogen fuel cell, the reference power of the second hydrogen fuel cell, the reference power of the first lithium battery, the reference power of the second lithium battery, the reference power of the second supercapacitor, and the reference power of the second supercapacitor based on the total power signal obtained in step (4), the load prediction data obtained in step (5), the internal resistance signal of the first hydrogen fuel cell and the internal resistance signal of the second hydrogen fuel cell obtained in step (3), the current signal of the first lithium battery in the first composite energy storage system, and the current signal of the second lithium battery in the second composite energy storage system. (7) The CRIO controller obtains the reference power of the first hydrogen fuel cell, the reference power of the second hydrogen fuel cell, the reference power of the first lithium battery, the reference power of the second lithium battery, the reference power of the first supercapacitor, and the reference power of the second supercapacitor from the host computer. It transmits the reference power of the first hydrogen fuel cell to the first unidirectional boost DC converter in the first hydrogen fuel cell power supply system, which uses power point tracking control to keep the first hydrogen fuel cell near its reference power. It transmits the reference power of the second hydrogen fuel cell to the second unidirectional boost DC converter in the second hydrogen fuel cell power supply system, which uses power point tracking control to keep the second hydrogen fuel cell near its reference power. It transmits the reference power of the first lithium battery to the first composite storage... The first bidirectional boost DC-DC converter in the energy storage system uses power point tracking (PPT) to keep the first lithium battery near its reference power. It transfers the reference power of the first supercapacitor to the second bidirectional boost DC-DC converter in the first composite energy storage system, which also uses PPT to keep the first supercapacitor near its reference power. The third bidirectional boost DC-DC converter in the second composite energy storage system uses PPT to keep the second lithium battery near its reference power. The fourth bidirectional boost DC-DC converter in the second composite energy storage system uses PPT to keep the second supercapacitor near its reference power. (8) The shipborne multi-source flexible DC ring network test bench maintains closed-loop operation until the process ends.

2. The energy management method for the shipborne multi-source flexible DC ring network test bench according to claim 1, characterized in that, Step (3) specifically involves the host computer multiplying the voltage signal and current signal of the first load system to obtain the power signal of the first load system, multiplying the voltage signal and current signal of the first hydrogen fuel cell power supply system to obtain the output power signal of the first hydrogen fuel cell, multiplying the voltage signal and current signal of the second hydrogen fuel cell power supply system to obtain the output power signal of the second hydrogen fuel cell, multiplying the voltage signal and current signal of the first lithium battery in the first composite energy storage system to obtain the output power signal of the first lithium battery, multiplying the voltage signal and current signal of the first supercapacitor in the first composite energy storage system to obtain the output power signal of the first supercapacitor, multiplying the voltage signal and current signal of the second lithium battery in the second composite energy storage system to obtain the output power signal of the second lithium battery, multiplying the voltage signal and current signal of the second supercapacitor in the second composite energy storage system to obtain the output power signal of the second supercapacitor, and obtaining the power signal of the second load system based on the speed and torque signals of the first dynamometer in the second load system and using the following formula. ， wherein is a torque signal of the first dynamometer in the second load system, is a rotational speed signal of the first dynamometer in the second load system. The working condition identification model uses a support vector machine (SVM) combined with a rolling time window for working condition identification. During the operation condition identification, the key features of 10 data points at the current time and the previous 18 seconds are calculated to determine the current operation condition category of the ship. The rolling window at time t contains data within the time range of [t-18s, t]. The window is updated every 2 seconds by sliding. During the sliding update, the oldest data point is removed and the latest data point is added, always keeping 10 data points in the window.

3. The energy management method of the ship-based multi-source flexible DC loop network test bench according to claim 2, characterized in that, The training process of the working condition recognition model further includes the following steps after step (4-2): (4-3) Perform K-Means clustering analysis on all key features obtained in step (4-2) to obtain multiple typical operating conditions of the ship and their corresponding multiple preprocessed historical load power data of the ship. (4-4) The operating condition identification model is trained offline using the preprocessed historical load power data of ships and their key features obtained in step (4-3) for each typical operating condition, so as to obtain a trained operating condition identification model.

4. The energy management method of the ship-based multi-source flexible DC loop network test bench according to claim 3, characterized in that, The power prediction model corresponding to each typical operating condition is a Convolutional Neural Network-Long Short-Term Memory Network (CNN-LSTM) model based on an attention mechanism, which is trained through the following steps: (5-1) Obtain multiple preprocessed historical load power data of ships corresponding to the typical operating condition obtained in step (4-3), and construct a training set of the power prediction model corresponding to the typical operating condition based on the historical load power data of ships. Specifically, this step involves first obtaining the volatility of the preprocessed historical load power data of the ship corresponding to this typical operating condition. The calculation formula is as follows: ， In the formula, is the i-th ship historical load power data in all preprocessed ship historical load power data corresponding to the typical operating condition, wherein i∈[2, the total number of all preprocessed ship historical load power data corresponding to the typical operating condition], then, all preprocessed ship historical load power data and its fluctuation rate corresponding to the typical operating condition are selected to construct a training set of the power prediction model corresponding to the typical operating condition. (5-2) Use the training set of the power prediction model corresponding to the typical operating condition obtained in step (5-1) to train the power prediction model to obtain the trained power prediction model corresponding to the typical operating condition.

5. The energy management method of the ship-based multi-source flexible DC loop network test bench according to claim 4, characterized in that, Step (6) specifically includes the following sub-steps: (6-1) The load data is decomposed using a low-pass filter controller to obtain the reference power of the first supercapacitor, the reference power of the second supercapacitor, and the low-frequency power; (6-2) Based on the internal resistance signal of the first hydrogen fuel cell and the internal resistance signal of the second hydrogen fuel cell obtained in step (3), obtain the health status (SOH) of the first hydrogen fuel cell and the SOH of the second hydrogen fuel cell. (6-3) Based on the SOH of the first hydrogen fuel cell and the SOH of the second hydrogen fuel cell obtained in step (6-2), the hydrogen consumption rate curve of the first hydrogen fuel cell and the hydrogen consumption rate curve of the second hydrogen fuel cell are obtained respectively. (6-4) Based on the hydrogen consumption rate data of the first hydrogen fuel cell and the hydrogen consumption rate data of the second hydrogen fuel cell obtained in step (6-3), respectively obtain the fitting equations of hydrogen consumption rate and power of the first hydrogen fuel cell and the second hydrogen fuel cell. (6-5) Input the fitting equations of hydrogen consumption rate and power of the first hydrogen fuel cell and the second hydrogen fuel cell obtained in step (6-4), the low-frequency power obtained in step (6-1), and the current signals of the first lithium battery and the second lithium battery obtained in step (3) into the MPC controller to obtain the reference power of the first hydrogen fuel cell, the second hydrogen fuel cell, the first lithium battery and the second lithium battery respectively.

6. The energy management method for the shipborne multi-source flexible DC ring network test bench according to claim 5, characterized in that, The expressions for the reference power and low-frequency power of the supercapacitor in step (6-1) are as follows: ， wherein, is the load data, is the complex variable, is the filter time constant, is the reference power of the first super capacitor, which is equal to the reference power of the second super capacitor, is the low frequency power; In step (6-2), the SOH of the hydrogen fuel cell is equal to: ， wherein R1 / 2end represents the ohmic internal resistance value at the end of life of the first / second hydrogen fuel cell / Ω; R1 / 2unused represents the ohmic internal resistance value of the first / second hydrogen fuel cell when not in use / Ω; R1 / 2current represents the ohmic internal resistance value of the first / second hydrogen fuel cell in its current state / Ω; Hydrogen consumption rate of the hydrogen fuel cell in step (6-3) is calculated by the following equation: ， wherein is the output power of the hydrogen fuel cell after passing through the DC / DC converter, is the lower heating value of hydrogen gas; is the system efficiency of the hydrogen fuel cell, is defined as follows: ， wherein is the hydrogen fuel cell system efficiency at the initial state of the hydrogen fuel cell, i.e. SOH = 1 ; coefficient of variation represents the degree of reduction of the electrical efficiency and has: ， wherein is the conversion efficiency of the DC / DC converter; is the parasitic power of the auxiliary equipment; is the stack output power; coefficient of variation represents the degree of reduction of the conversion efficiency and has: ， In step (6-4), the hydrogen fuel cell hydrogen consumption rate is obtained from the hydrogen fuel cell hydrogen consumption rate data calculated in step (6-3) using a quadratic polynomial fitting method. The fitting equation is: ， Where a, b, and c are the fitting coefficients for the hydrogen fuel cell; The objective function and constraints of the MPC controller in step (6-5) are as follows: ， ， in and These are the hydrogen consumption rate of the first hydrogen fuel cell and the hydrogen consumption rate of the first hydrogen fuel cell, respectively. The equivalent hydrogen consumption rate of the first / second lithium battery; and The degradation weight represents the degree of optimization bias in the overall performance of the power system due to the degradation of hydrogen fuel cells and lithium batteries; its value ranges from 0 to 1. It is 0.

6. It is 0.4; This represents the equivalent hydrogen consumption due to the degradation of a hydrogen fuel cell. The degradation cost of the first / second lithium battery; This is the reference power for the first hydrogen fuel cell. This is the reference power for the second hydrogen fuel cell. This is the reference power of the first lithium battery, which is equal to the reference power of the second lithium battery; The state of charge of the first / second lithium battery. The degradation factor is calculated as follows: ,in This is the energy efficiency optimization factor for hydrogen fuel cells, and its value determines the sensitivity of the adjustment range of the hydrogen fuel cell output power to its degradation.

7. The energy management method for the shipborne multi-source flexible DC ring network test bench according to claim 1, characterized in that, The first hydrogen fuel cell power supply system and the second hydrogen fuel cell power supply system are exactly the same; The first hydrogen fuel cell power supply system includes a hydrogen fuel cell, a first DC circuit breaker, a first DC fuse, a first unidirectional boost DC converter, a first current and voltage sensor, a second DC fuse, and a second DC circuit breaker. The first hydrogen fuel cell is connected to the second DC bus via a first DC circuit breaker, a first DC fuse, a first unidirectional boost DC converter, a first current and voltage sensor, a second DC fuse, and a second DC circuit breaker in sequence. The first hydrogen fuel cell has a rated power of 40kW and an output voltage range of 450-700V. The rated voltage of the first DC circuit breaker and the rated current of the second DC circuit breaker are 1000V and 63A respectively. The rated voltage of the first DC fuse and the rated current of the second DC fuse are 1000V and 63A respectively. The first unidirectional boost DC-DC converter is a unidirectional isolated converter with an input voltage range of 400-850V, an output voltage range of 400-850V, and a rated power of 40kW. The first voltage and current sensor has a current measurement range of 0-100A and a voltage measurement range of 0-1000V.

8. The energy management method for the shipborne multi-source flexible DC ring network test bench according to claim 7, characterized in that, The first composite energy storage system and the second composite energy storage system are exactly the same; The first composite energy storage system includes a first lithium battery, a first supercapacitor, a third DC circuit breaker, a third DC fuse, a first bidirectional boost DC converter, a second current and voltage sensor, a fourth DC fuse, a fourth DC circuit breaker, a fifth DC circuit breaker, a fifth DC fuse, a second bidirectional boost DC converter, a third current and voltage sensor, a sixth DC fuse, and a sixth DC circuit breaker. The first lithium battery is connected to the first DC bus in sequence via the third DC circuit breaker, the third DC fuse, the first bidirectional boost DC converter, the second current and voltage sensor, the fourth DC fuse, and the fourth DC circuit breaker. The first supercapacitor is connected to the first DC bus via the fifth DC circuit breaker, the fifth DC fuse, the second bidirectional boost DC converter, the third current and voltage sensor, the sixth DC fuse, and the sixth DC circuit breaker in sequence. The first lithium battery has a capacity of 65kWh, a nominal voltage of 633V, an operating voltage range of 574-693V, and a maximum charge / discharge rate of 1C. The first supercapacitor has a capacitance of 5.2F, a rated power of 80kW, and an operating voltage range of 200-800V; The rated voltage of the third, fourth, fifth, and sixth DC circuit breakers is 1000V, and the rated current is 63A. The rated voltage of the third, fourth, fifth, and sixth DC fuses is 1000V, and the rated current is 63A. The first and second bidirectional boost DC-DC converters are bidirectional non-isolated converters with an A-side voltage range of 400-850V and a B-side voltage range of 300-850V, and a rated power of 60kW. The second and third voltage and current sensors have a current measurement range of 0-100A and a voltage measurement range of 0-1000V.

9. The energy management method for the shipborne multi-source flexible DC ring network test bench according to claim 8, characterized in that, The first load system includes a first resistive load cabinet, a seventh DC fuse, a seventh DC circuit breaker, and a fourth voltage and current sensor; The first resistive load cabinet is electrically connected to the first DC bus via the seventh DC fuse, the seventh DC circuit breaker, and the fourth voltage and current sensor in sequence. The rated power of the first resistive load cabinet is 10kW; The rated voltage of the seventh DC fuse is 1000V and the rated current is 63A; The rated voltage of the seventh DC circuit breaker is 1000V and the rated current is 63A; The fourth voltage and current sensor has a current measurement range of 0-100A and a voltage measurement range of 0-1000V. The second load system includes a first dynamometer, a first reversible motor, a first DC / AC inverter, an eighth DC fuse, an eighth DC circuit breaker, and a fifth voltage and current sensor. The first dynamometer is connected to the third DC bus in sequence via the first reversible motor, the first DC / AC inverter, the eighth DC fuse, the eighth DC circuit breaker, the fifth voltage and current sensor; The first dynamometer is an eddy current dynamometer with a rated absorption power of 320kW; The first reversible motor has a rated power of 75kW, a rated voltage of 380V, and a rated speed of 1500r / min. The first DC / AC inverter outputs 380V AC three-phase power; The rated voltage of the eighth DC fuse is 1000V and the rated current is 63A; The rated voltage of the eighth DC circuit breaker is 1000V and the rated current is 63A; The fifth voltage and current sensor has a current measurement range of 0-100A and a voltage measurement range of 0-1000V.

Images