Power prediction and dynamic setting method of urban rail hybrid energy storage three-level DC / DC converter

Abstract

The invention provides a power prediction and dynamic setting method of an urban rail hybrid energy storage three-level DC / DC converter, aiming at the problems of dynamic adjustment and dynamic response speed of an urban rail hybrid energy storage system. On the basis of a non-isolated direct-current three-level converter, a fuzzy reasoning system and a PI-free power MPC control method are combined, a fuzzy rule is established, and a charging and discharging threshold value of a lithium battery is dynamically generated according to the real-time state of charge (SOCsc) of a super capacitor which is monitored in real time; and in combination with a PI-free power MPC control method, the problems of slow dynamic response of the system and low efficiency of the two-level DC / DC converter under traditional PI control are solved. Compared with the traditional PI control scheme, the power prediction control method is introduced, so that the use of a PI control ring is avoided, and the parameter setting process is simplified. Through fuzzy reasoning and MPC collaborative optimization, the charging and discharging states of the energy storage system are dynamically adjusted, the service life of a lithium battery and a super capacitor is prolonged, the power supply stability and the energy recovery capacity of the system are enhanced, meanwhile, the energy conversion rate and the dynamic response performance of the system are remarkably improved, and the operation loss and the manufacturing cost are reduced. And an efficient, reliable and low-maintenance-cost energy storage solution is provided for urban rail transit.

Description

Technical Field

[0001] The present invention relates to the technical field of urban rail transit energy storage and energy saving, specifically a method for dynamically setting the charging and discharging control and charging and discharging threshold fuzzy inference of a three-level DC converter based on energy storage. Background Art

[0002] With the rapid development of urban rail transit, the hybrid energy storage system (HESS) has played an important role in improving energy utilization efficiency and power supply stability. Traditional rail transit systems usually adopt two-level DC / DC converters. Although the structure is simple, the loss is large under high-power conditions and the conversion efficiency is low. In contrast, the three-level DC / DC converter has lower switching losses and voltage stress, can effectively improve the energy conversion efficiency, reduce electromagnetic interference, support a higher power density, and is especially suitable for the rail transit application scenarios with high-dynamic loads.

[0003] Under the conditions of frequent acceleration and deceleration and high-frequency load fluctuations in rail transit, due to the limitations of the regulation speed and dynamic performance, the traditional PI control method is difficult to meet the rapid response requirements. The model predictive control (MPC) does not need to rely on the PI loop and can achieve fast and accurate power regulation, thus significantly improving the system response speed and dynamic performance. However, when facing the common charging and discharging bidirectional control with dynamic thresholds or fixed thresholds, the system may not be able to achieve efficient and flexible energy management. Therefore, the present invention proposes a power prediction control method without a PI loop, which is applied to the charging and discharging control process of the converter, aiming to achieve efficient management of the energy storage system, improve the system response speed, reduce energy loss, extend the equipment life, and at the same time improve the flexibility and efficiency of the energy management of the energy storage system. Summary of the Invention

[0004] Technical Problem: In high-power energy storage converters, a two-level topology structure is generally adopted, which has a simple structure, convenient maintenance, and high reliability. However, its power module cost is high, the volume of the filter inductor is large, the operating noise is large, and the loss of the switching tube is large during high-frequency switching, which limits the efficiency and reliability of the system. The traditional control with a PI loop has cumbersome parameter adjustment, slow response speed, and is not sensitive to the dynamic changes of the system in a complex system. At the same time, for trains with frequent starts and stops, the HESS not only needs to take into account multiple indicators such as efficiency, equipment life, and system stability, but also requires an efficient energy management scheme.

[0005] Technical solution: To solve the above problems, the present invention applies fuzzy inference to the calculation of charge and discharge voltage thresholds, and combines the high dynamic response speed of power MPC, so as to better adjust the setting of voltage thresholds according to the operation state of urban rail transit, better play the coordinated control between lithium batteries and supercapacitors, enable the supercapacitor to reach the optimal charge and discharge depth within a safe range, further reduce the battery output, effectively extend its life, and thus improve the comprehensive performance of HESS.

[0006] In the first aspect, since the dynamic response speed of power MPC is extremely excellent, to give full play to its advantages. The present invention introduces a dynamic setting method for charge and discharge threshold fuzzy inference, and according to the supercapacitor SOC, fuzzy rules are summarized based on experience and data to perform dynamic control of voltage thresholds, so as to better adjust the charge and discharge voltage thresholds of lithium batteries in real time according to the operation state.

[0007] The dynamic setting and PI-free power MPC control method of the urban rail hybrid energy storage system proposed by the present invention is characterized in that the control method is based on the PI-free power MPC control strategy of a three-level DC converter and combines a fuzzy inference system. By establishing fuzzy rules, according to the real-time state of charge (SOC sc ) of the supercapacitor monitored in real time, the charge and discharge thresholds of the lithium battery are dynamically generated, and the voltage U of the traction network is detected in real time dc changes to determine the working mode of the hybrid energy storage system, generate corresponding real-time signal input quantities, and finally combine the PI-free power MPC control method to dynamically adjust the charge and discharge state of the energy storage system, so as to realize the dual optimization of energy management and lithium battery protection. The present invention is divided into a design part and an operation part, and is realized as follows:

[0008] Design a charge and discharge threshold fuzzy inference system, including membership function selection, membership function expression, fuzzy domain determination, determination of fuzzy subsets, and establishment of fuzzy inference rules.

[0009] The present invention relates to a coordinated control method for a rail transit hybrid energy storage system. Aiming at the problem of insufficient coordination efficiency between power-type supercapacitors and energy-type lithium batteries under the existing fixed voltage threshold strategy, a dynamic threshold adaptive method based on fuzzy control is proposed. In this design, the voltage fluctuation characteristics of the DC traction network are used as constraints to construct a collaborative architecture in which the supercapacitor responds first and the lithium battery compensates dynamically. In the design, the supercapacitor discharge voltage threshold U sc_dis is initially set to 1480V, and the charging voltage threshold U sc_char is 1520V; by collecting the real-time feedback value of the supercapacitor SOC sc , the charge and discharge voltage thresholds (U bat_dis / U bat_charThe optimized parameter set enables the hybrid energy storage system (HESS) to achieve coordinated power - energy control. Compared with the traditional fixed - threshold method, it realizes the dynamic coordination of the operating modes of energy storage components: at the initial stage of train acceleration / braking, the supercapacitor quickly absorbs / releases high - power components; when the SOC value of the supercapacitor deviates from the set range, the voltage action boundary of the lithium - battery is autonomously adjusted through fuzzy rules to trigger its energy compensation function. This design realizes the efficient distribution of the energy of the energy storage system and the coordinated optimization of energy management by optimizing the output levels of the supercapacitor and the lithium - battery.

[0010] The fuzzy inference engine selects the SOC of the supercapacitor sc as the input, and the charging threshold (U bat_char ) and the discharging threshold (U bat_dis ) of the lithium - battery as the outputs; their settings are as follows:

[0011] 1) Selection of membership functions for fuzzy inference, determination of fuzzy universes and fuzzy subsets:

[0012] SOC sc Seven fuzzy subsets are selected: NL (Negative Large), NM (Negative Medium), NS (Negative Small), ZO (Zero), PS (Positive Small), PM (Positive Medium), PL (Positive Large)

[0013] For the charging voltage threshold U bat_char of the lithium - battery, seven fuzzy subsets are selected: NL (Negative Large), NM (Negative Medium), NS (Negative Small), ZO (Zero), PS (Positive Small), PM (Positive Medium), PL (Positive Large)

[0014] For the discharging voltage threshold U bat_dis of the lithium - battery, seven fuzzy subsets are selected: NL (Negative Large), NM (Negative Medium), NS (Negative Small), ZO (Zero), PS (Positive Small), PM (Positive Medium), PL (Positive Large)

[0015] SOC sc The fuzzy universe is [0.2 0.8]

[0016] U bat_char The fuzzy universe is [1510 1530]

[0017] U bat_dis The fuzzy universe is [1470 1490]

[0018] SOC sc 、U bat_char 、U bat_dis The membership functions adopt the trigonometric function formula or the Gaussian - type formula;

[0019] Trigonometric function formula:

[0020] It is required that a ≤ b ≤ c, where a, b, and c are the left, middle, and right coordinates of the membership degree respectively, and x is the input value;

[0021] Gaussian type:

[0022] Among them, c is the position of the center of the membership function, σ is the width of the membership function curve, and x is the input value;

[0023] 2) Fuzzy inference rules:

[0024] Discharge state:

[0025] If (SOC sc is PL) then (U bat_dis is NL)

[0026] If (SOC sc is PM) then (U bat_dis is NM)

[0027] If (SOC sc is PS) then (U bat_dis is NS)

[0028] If (SOC sc is ZO) then (U bat_dis is ZO)

[0029] If (SOC sc is NS) then (U bat_dis is PS)

[0030] If (SOC sc is NM) then (U bat_dis is PM)

[0031] If (SOC sc is NL) then (U bat_dis is PL)

[0032] Charging state:

[0033] If (SOC sc is NL) then (U bat_char is PL)

[0034] If (SOC sc is NM) then (U bat_char is PM)

[0035] If (SOC sc is NS) then (Ubat_char is PS)

[0036] If(SOC sc is ZO)then(U bat_char is ZO)

[0037] If(SOC sc is PS)then(U bat_char is NS)

[0038] If(SOC sc is PM)then(U bat_char is NM)

[0039] If(SOC sc is PL)then(U bat_char is NL)

[0040] Fuzzy dynamic setting principle: During the discharging process, when the SOC of the supercapacitor sc has a larger value, it indicates that the remaining charge of the supercapacitor is relatively large, and it can still bear most of the peak power. When the SOC of the supercapacitor sc becomes smaller and smaller, the remaining charge of the supercapacitor energy storage device becomes less and less. At this time, the discharge threshold of the lithium battery increases, causing the lithium battery energy storage device to start assuming the insufficient power output. During the charging process, when the SOC of the supercapacitor sc has a larger value, it indicates that the remaining charge of the supercapacitor is large. At this time, the charging of the supercapacitor should be reduced, and the charging voltage threshold of the lithium battery should be lowered to make the lithium battery charge as early as possible. On the contrary, when the SOC sc becomes smaller and smaller, the remaining charge of the supercapacitor becomes less and less. At this time, the charging voltage threshold of the lithium battery is increased, reducing the power absorbed by the lithium battery energy storage device, thereby increasing the output of the supercapacitor.

[0041] Taking the supercapacitor SOC sc as the fuzzy input, through the fuzzy inference system, the charging and discharging thresholds of the lithium battery voltage are obtained, and the method of combining the non-PI power MPC control is used to solve the disadvantage of difficult setting of the PI controller control parameters in the PI double closed-loop control and the current prediction control, making the energy storage system have a better dynamic response speed, giving full play to the fast charge and discharge characteristics of the supercapacitor, and further protecting the lithium battery. The utilization degree of the energy storage system tends to be balanced and reasonable, avoiding overcharge and overdischarge of the supercapacitor. At the same time, a three-level DC converter is introduced to provide an additional voltage level selection for the converter, reducing the voltage stress of the switching tubes and diodes, reducing the switching loss, reducing the voltage fluctuation amplitude, and ensuring the stability and reliability of the output of the energy storage system;

[0042] For the PI-free power MPC control method of a DC three-level converter, it is characterized in that the three-level converter adopts a non-isolated three-level DC / DC topology structure. The converter topology consists of four IGBT tubes T1, T2, T3, and T4, two equalizing capacitors C1, C2, and inductors L1, L2. The low-voltage side is connected to the hybrid energy storage system, and the high-voltage side is connected to the urban rail traction network;

[0043] This method obtains the optimal switching state through power prediction. Without a PI link, it simplifies the control structure, omits the adjustment of multiple parameters, and can significantly improve the dynamic response speed of the system; for different operating modes of the energy storage system, corresponding prediction equations can be established;

[0044] For the switching state of the three-level DC / DC converter, the switching state can be summarized into 4 modes in the Boost / Buck mode;

[0045] The switching modes of the converter during operation are shown in Table (1):

[0046]

[0047] When the switching tube T works, T1 and T2 conduct complementarily, and T3 and T4 conduct complementarily. 0 indicates that the switching tube T is off, and 1 indicates that the switching tube T is on, U ab which represents the three voltage levels of the three-level converter; U dc is the DC traction network voltage;

[0048] The HESS adopts a dual-active parallel topology structure composed of a supercapacitor and a lithium battery. Their control structures are completely symmetrical, only the parameters are different. Taking the supercapacitor (SC) as an example, the control method is introduced in detail, and then the control method of the HESS is summarized.

[0049] The available regenerative energy storage devices in the urban rail HESS mainly include three working conditions: when the train is in the traction acceleration stage, the energy storage system is in the discharge state; when the train is in the coasting operation stage, the energy storage system is in the holding state; when the train is in the braking deceleration stage, the energy storage system is in the charging state; the energy management of the energy storage system mainly refers to the standard of the DC traction network voltage as a condition to specifically design the charge and discharge voltage thresholds of the energy storage system as whether the hybrid energy storage system is put into use;

[0050] The HESS energy management strategy is studied based on the characteristics of the hybrid energy storage system. Since the lithium battery and the supercapacitor have complementary characteristics, in terms of capacity ratio, the capacity of the lithium battery is greater than that of the supercapacitor; in terms of energy management, the response priority of the supercapacitor is higher than that of the lithium battery. The energy storage device must work within the discharge rate limit during operation, so the SOC constraint charge and discharge limit values are introduced.

[0051]

[0052] Where: SOC bat , SOC sc are the state of charge of the lithium battery and the supercapacitor respectively; E bat and E sc are the currently stored energies of the lithium battery and the supercapacitor respectively; E bat_max and E sc_max are the maximum stored energies of the lithium battery and the supercapacitor respectively; C bat , C sc are the capacities of the lithium battery and the supercapacitor respectively, U bat , U sc are the operating voltages of the lithium battery and the supercapacitor respectively, U bat_max , U sc_max are the rated voltages of the lithium battery and the supercapacitor respectively.

[0053] During the entire process of charging and discharging, it is also necessary to note that the state of charge of the energy storage device must be within the allowable operating range.

[0054] SOC bat_min (t) ≤ SOC bat (t) ≤ SOC bat_max (t) (5)

[0055] SOC sc_min (t) ≤ SOC sc (t) ≤ SOC sc_max (t) (6)

[0056] SOC bat_min , SOC bat_max represent the minimum and maximum safety values of the lithium battery, SOC sc_min , SOC sc_max represent the minimum and maximum safety values of the supercapacitor. Usually, the safety values of the state of charge of the lithium battery and the supercapacitor are between [0.2, 0.8].

[0057] The specific design process is as follows:

[0058] First, to avoid overcharging and over-discharging of the energy storage system, the prohibited and normal operating modes are set according to the state of charge SOC of the energy storage system. Assume that the state of charge SOC of the energy storage system is lower than the set minimum safety value SOC min or higher than the set maximum safety value SOC max , that is, when SOC ≤ SOC min or when SOC ≥ SOC max , it enters the prohibited state; conversely, when SOC max ≤ SOC ≤ SOC minWhen the energy storage operates normally. Furthermore, the charge-discharge state of the energy storage system is judged by voltage comparison. Under the normal operating state of the energy storage system, the specific steps for establishing the prediction model are as follows:

[0059] a) If the voltage U of the supercapacitor SC is higher than the set supercapacitor voltage threshold U sc_dis , at this time, the converter Boost charges / the supercapacitor discharges, with the SC discharge current as the reference direction. At this time, T2 and T3 are the main control tubes, and T1 and T4 are turned off. The four switching states in this mode are: State 1 (T2 = 1, T3 = 1), State 2 (T2 = 1, T3 = 0), State 3 (T2 = 0, T3 = 0), State 4 (T2 = 0, T3 = 1). Among them, the state of the switching tube is represented by the value of the binary variable T, where 1 represents the switching tube is on and 0 represents the switching tube is off. The mathematical model of the three-level converter in different switching modes is established as follows:

[0060] Mode 1: T2 is on, T3 is on, and the switching state of [T1, T2, T3, T4] is [0, 1, 1, 0] at this time;

[0061] The corresponding mathematical model is:

[0062] Mode 2: T2 is on, T3 is off, and the switching state of [T1, T2, T3, T4] is [0, 1, 0, 1] at this time;

[0063] The corresponding mathematical model is:

[0064] Mode 3: T2 is off, T3 is off, and the switching state of [T1, T2, T3, T4] is [1, 0, 0, 1] at this time;

[0065] The corresponding mathematical model is:

[0066] Mode 4: T2 is off, T3 is on, and the switching state of [T1, T2, T3, T4] is [1, 0, 1, 0] at this time;

[0067] The corresponding mathematical model is:

[0068] By performing Euler discretization on the inductor current mathematical models of equations (7) to (10), the values of the inductor current at times k and k + 1 in the corresponding operating modes can be obtained. The specific steps are as follows:

[0069] The inductor current derivative di / dt is approximately replaced by the backward Euler formula, and the formula is as shown in equation (11):

[0070]

[0071] Further obtain the inductor current \(i\). L That is, the capacitor current \(i\). sc The discrete form is as shown in Equation (12):

[0072]

[0073] For the derivative of the capacitor voltage \(dU / dt\), the backward Euler approximation is also adopted, and the expression is as shown in Equation (13):

[0074]

[0075] Further obtain the discrete form of the capacitor voltage as shown in Equation (14):

[0076]

[0077] Where: \(U\). cj (k + 1) and \(U\). cj (k) are the sampled values of the capacitor voltage at the \((k + 1)\) - th and \(k\) - th moments respectively, \(i\). cj (k) is the sampled value of the capacitor current at the \(k\) - th moment, \(j = 1,2,\cdots,T\). s is the sampling period;

[0078] Substitute Equations (7) - (10) into Equation (11), and the inductor current \(i\). L (k + 1) at the \((k + 1)\) - th moment under the Boost charging / SC discharging condition of the converter can be obtained, that is, the capacitor current \(i\). sc (k + 1) of the discrete form is as shown in Equations (15) - (18):

[0079] Mode 1:

[0080] Mode 2:

[0081] Mode 3:

[0082] Mode 4: It can be uniformly expressed as: For HESS:

[0083] b) If the super - capacitor voltage \(U\). SC is lower than the voltage threshold \(U\). sc_char set for the super - capacitor, the energy - storage starts the charging process. At this time, the converter is in Buck / super - capacitor charging mode, and T1 and T4 are the main control transistors, and T2, T3 are turned off. In this mode, the four switching states are: State 1 (T1 = 1, T4 = 1), State 2 (T1 = 1, T4 = 0), State 3 (T1 = 0, T4 = 0), State 4 (T1 = 0, T4 = 1). The mathematical model is established as follows:

[0084] Mode 5: T1 is turned on, T4 is turned on. At this time, the switch states of [T1, T2, T3, T4] are [1, 0, 0, 1];

[0085] The corresponding mathematical model is:

[0086] Mode 6: T1 is turned on, T4 is turned off. At this time, the switch states of [T1, T2, T3, T4] are [1, 0, 1, 0];

[0087] The corresponding mathematical model is:

[0088] Mode 7: T1 is turned off, T4 is turned off. At this time, the switch states of [T1, T2, T3, T4] are [0, 1, 1, 0];

[0089] The corresponding mathematical model is:

[0090] Mode 8: T1 is turned off, T4 is turned on. At this time, the switch states of [T1, T2, T3, T4] are [0, 1, 0, 1];

[0091] The corresponding mathematical model is:

[0092] Substituting Eqs. (21) to (24) into Eq. (11), the discrete form of the capacitor current i sc (k + 1) can be obtained as shown in Eqs. (25) to (28):

[0093] Mode 5:

[0094] Mode 6:

[0095] Mode 7:

[0096] Mode 8:

[0097] It can be unified as:

[0098] For HESS:

[0099] c) If the supercapacitor voltage U SC is within the set working range, the converter / supercapacitor enters the standby state. That is, when U sc_dis ≤U sc ≤U sc_char the converter / supercapacitor is in the standby state.

[0100] Where: L1 and L2 are the upper and lower inductance values in the converter, Usc (k), U b (k) is the voltage of the supercapacitor and the lithium battery on the energy storage side at time k, U Hess (k) represents the output voltage U of the energy storage side sc (k) or U b (k); T s is the sampling period; U c1 (k) and U c2 (k) are the voltage sampling values of C1 and C2 of the voltage-dividing capacitor at time k respectively; U dc is the DC bus voltage; the inductor current i at time k + 1 L (k + 1) and the supercapacitor current i sc (k + 1);

[0101] Input the sampling values into the discrete prediction model in the corresponding charge / discharge state to predict the inductor current i under all possible switching states at the next moment L (k + 1) the voltage of the voltage-dividing capacitor U c1 (k + 1), U c2 (k + 1);

[0102] Set the prediction objective function J of the supercapacitor as shown in Equation (31):

[0103]

[0104] In the formula, P sc (k + 1) is the predicted power value of the supercapacitor at time k + 1. Substitute the discretized equations (19) and (29) into P sc (k + 1)=|i sc (k + 1)×U sc (k + 1)|, and the power prediction model corresponding to all switching states of the supercapacitor can be obtained. i sc (k + 1) is the predicted value of the capacitor current at time k + 1, U c1 (k + 1), U c2 (k + 1) is the predicted value of the voltage-dividing capacitor obtained according to Equation (14) at time k + 1, U dc (k + 1)=U c1 (k + 1)+U c2 (k + 1); is the power reference value of the supercapacitor, i dc (k) is the current value flowing through the traction network at time k, is the bus reference voltage;

[0105] Then for HESS, the objective function can be expressed as Equation (32):

[0106]

[0107] Where P Hess (k + 1) is the predicted power value of the energy storage system at the (k + 1)-th moment. Similarly, substituting the discretized equations (20) and (30) into P Hess (k + 1) = |i L (k + 1) × U dc (k + 1)|, the power prediction model corresponding to all switch states of the system can be obtained; U c1 (k + 1), U c2 (k + 1) is the predicted value of the voltage-dividing capacitor obtained according to Equation (14) at the (k + 1)-th moment; U dc (k + 1) = U c1 (k + 1) + U c2 (k + 1); is the power reference value of the energy storage system, is the bus reference voltage, set to 1500V;

[0108] In each cycle when the MPC algorithm works, when the objective function value J corresponding to the switch state is less than the current optimal value J op , J op is replaced by the new objective function value. Conversely, the optimal value J op remains unchanged. The optimization of the control objective is achieved by minimizing the objective function. J1, J2, J3, and J4 are the objective functions corresponding to the 4 prediction models respectively. The switch state J corresponding to min{J1, J2, J3, J4} is selected min as the switch state at the next moment. It can be seen that after comparing all possible switch states one by one, the switch state S i acting on the three-level bidirectional DC converter is unique and also optimal. The optimal switch state S i obtained by this algorithm is updated in real time in each working cycle of the algorithm, and is selected according to the objective function value from the following 4 switch states: when the state j = 1, the switch tubes T2 and T3 are both turned on; when the state j = 2, T2 is turned on and T3 is turned off; when the state j = 3, T2 and T3 are both turned off; when the state j = 4, T2 is turned off and T3 is turned on. Since at each sampling moment, the optimization performance index only involves a finite future time starting from this moment, and by the next sampling moment, this optimization time period moves forward, the optimization process can be repeated online, thus realizing cyclic optimization;

[0109] For the system operation part, the specific operation steps are as follows:

[0110] Step1: Real-time collect the traction network terminal voltage U dc and the supercapacitor SOC sc , and use the SOC sc as the input quantity of the fuzzy system;

[0111] Step 2: Run the fuzzy inference system to obtain the corresponding dynamic setting value U bat_dis and U bat_char ;

[0112] Step 3: Compare U dc with the charge and discharge thresholds of the energy storage system to determine the charge and discharge status of the system;

[0113] Step 4: At the corresponding state, sample the input voltage U dc (k) and input current i L (k) of the converter at the current moment k by the sensor, as well as the terminal voltage U b (k), U sc (k) and terminal current i b (k), i sc (k) of the hybrid energy storage system;

[0114] Step 5: The computer calculates the power prediction value and reference value of the HESS at the k + 1 moment respectively, and then calculates the objective function values J corresponding to all switch states according to the objective function, and selects the smaller value J min corresponding to the switch state S i as the switch state at the k + 1 moment, and S i is the switch signal composed of 0 and 1 for the 4 IGBT tubes T1, T2, T3, T4;

[0115] Step 6: Apply the new switch state to the DC three-level converter;

[0116] Step 7: Return to Step 1 to start the optimization of the next switch state; if a fault occurs in the energy storage system or the DC three-level converter, the program ends.

[0117] Compared with the prior art, the beneficial effects brought by the technical solution of the present invention are:

[0118] 1. The present invention combines a fuzzy inference dynamic setting method for the charge and discharge thresholds of an urban rail hybrid energy storage, and the setting of the threshold can be changed in real time according to fuzzy rules, with strong robustness, especially suitable for the control of non-linear, time-varying and pure lag systems.

[0119] 2. The present invention introduces a power MPC method without PI control. While retaining the excellent dynamic response characteristics of traditional predictive control, it eliminates the complex parameter tuning process in the traditional control process with a PI loop. At the same time, the switch state is given by the computer and directly applied to the three-level converter. The number of electronic devices required by the system is less than that of traditional control methods, significantly reducing the system cost and control complexity.

[0120] 3. The present invention applies a three-level energy storage converter to a hybrid energy storage system, achieving breakthroughs in terms of power conversion efficiency, power density, and dynamic response speed compared with the traditional two-level structure. Among them, the multi-level modulation characteristics of the converter effectively reduce the voltage stress and switching losses of power devices through voltage stratification, and the combined compact topology design significantly reduces the volume of the equipment. It provides an energy conversion solution with high energy efficiency and high reliability for the urban rail transit energy storage system, having outstanding engineering practical value and economic advantages. Description of the Drawings

[0121] Figure 1 It is a schematic diagram of the circuit topology structure of a DC three-level converter;

[0122] Figure 2 It is a structure diagram of a fuzzy inference system;

[0123] Figure 3 It is a schematic diagram of a dynamic setting and power MPC control method;

[0124] Figure 4 It is a schematic diagram of the working mode switching principle of the supercapacitor part of the energy storage system;

[0125] Figure 5 It is a flowchart for selecting the optimal switching state. Detailed Embodiment

[0126] The present invention will be further described in detail below in conjunction with the drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0127] The dynamic setting and power MPC control method of the urban rail hybrid energy storage system proposed by the present invention are described in detail as follows in conjunction with the drawings and specific embodiments:

[0128] Figure 1 This is the circuit topology structure of the DC three-level converter in the present invention. The present invention includes two parts: dynamic setting and power MPC control. The specific setting steps are as follows:

[0129] Design a charge and discharge threshold fuzzy inference system, including membership function selection, membership function expression, determination of the fuzzy domain, determination of fuzzy subsets, and establishment of fuzzy inference rules;

[0130] Based on SOC sc The dynamic setting fuzzy inference system is as shown in Figure 2 . The fuzzy inference device selects the supercapacitor SOC sc as the input, and the charging threshold (U bat_char ) and discharging threshold (U bat_dis ) of the lithium battery as the output; the settings are as follows:

[0131] 1) Selection of membership functions for fuzzy inference, determination of fuzzy universes and fuzzy subsets:

[0132] SOC sc Seven fuzzy subsets are selected: NL (Negative Large), NM (Negative Medium), NS (Negative Small), ZO (Zero), PS (Positive Small), PM (Positive Medium), PL (Positive Large);

[0133] Lithium battery charging voltage threshold U bat_char and seven fuzzy subsets are selected: NL (Negative Large), NM (Negative Medium), NS (Negative Small), ZO (Zero), PS (Positive Small), PM (Positive Medium), PL (Positive Large);

[0134] Lithium battery discharging voltage threshold U bat_dis and seven fuzzy subsets are selected: NL (Negative Large), NM (Negative Medium), NS (Negative Small), ZO (Zero), PS (Positive Small), PM (Positive Medium), PL (Positive Large);

[0135] SOC sc Fuzzy universe [0.2 0.8]

[0136] U bat_char Fuzzy universe [1510 1530]

[0137] U bat_dis Fuzzy universe [1470 1490]

[0138] SOC sc 、U bat_char 、U bat_dis The membership function adopts the trigonometric function formula;

[0139] Triangle function formula:

[0140] It is required that a ≤ b ≤ c, where a, b, and c are the left, middle, and right coordinates of the membership degree respectively, and x is the input value;

[0141] For the present invention, the trigonometric function is adopted. For the triangular membership function, for SOC sc the following values are selected: 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8; for U bat_char the following values are respectively selected: 1510, 1514, 1516, 1520, 1524, 1526, 1530; for U bat_dis the following values are selected: 1470, 1474, 1476, 1480, 1484, 1486, 1490;

[0142] Gaussian type:

[0143] where c is the position of the center of the membership function, σ is the width of the membership function curve, and x is the input value;

[0144] 2) Fuzzy inference rules:

[0145] Discharge state:

[0146] If (SOC sc is PL) then (U bat_dis is NL)

[0147] If (SOC sc is PM) then (U bat_dis is NM)

[0148] If (SOC sc is PS) then (U bat_dis is NS)

[0149] If (SOC sc is ZO) then (U bat_dis is ZO)

[0150] If (SOC sc is NS) then (U bat_dis is PS)

[0151] If (SOC sc is NM) then (U bat_dis is PM)

[0152] If (SOC sc is NL) then (U bat_dis is PL)

[0153] Charging state:

[0154] If (SOC sc is NL) then (U bat_char is PL)

[0155] If (SOC sc is NM) then (U bat_char is PM)

[0156] If (SOC sc is NS) then (U bat_char is PS)

[0157] If (SOC sc is ZO) then (U bat_char is ZO)

[0158] If (SOC scis PS)then(U bat_char is NS)

[0159] If(SOC sc is PM)then(U bat_char is NM)

[0160] If(SOC sc is PL)then(U bat_char is NL)

[0161] Taking the supercapacitor SOC sc as the fuzzy input, through the fuzzy inference system, the charging and discharging threshold values of the lithium battery voltage are obtained, and the method of combining the non-PI power MPC control is adopted to solve the disadvantage of difficult setting of the PI controller control parameters in the PI double closed-loop control and the current prediction control, so that the energy storage system has a better dynamic response speed, gives full play to the fast charge and discharge characteristics of the supercapacitor, and further protects the lithium battery. The utilization degree of the energy storage system tends to be balanced and reasonable, avoiding overcharge and overdischarge of the supercapacitor. At the same time, a three-level DC converter is introduced to provide an additional voltage level selection for the converter, reduce the voltage stress of the switching tubes and diodes, reduce the switching loss, reduce the voltage fluctuation amplitude, and ensure the stability and reliability of the output of the energy storage system. The schematic diagram of the dynamic setting and power MPC control method based on the three-level DC converter is as Figure 3 shown, where the converters used in the lithium battery and supercapacitor modules are the same.

[0162] For the non-PI power MPC control method of the DC three-level converter, its characteristics are that the three-level converter adopts a non-isolated three-level DC / DC topology structure. The converter topology is composed of 4 IGBT tubes T1, T2, T3, T4, two equalizing capacitors C1, C2 and inductors L1, L2. The low-voltage side is connected to the hybrid energy storage system, and the high-voltage side is connected to the urban rail traction network;

[0163] This method obtains the optimal switching state through power prediction, has no PI link, simplifies the control structure, omits the adjustment of multiple parameters, and can significantly improve the dynamic response speed of the system; for different working modes of the energy storage system, corresponding prediction equations can be established;

[0164] For the switching state of the three-level DC / DC converter, the switching state can be summarized into 4 modes in the Boost / Buck mode;

[0165] The switching modes of the converter during operation are shown in Table (1):

[0166]

[0167] When the switching transistor T is working, T1 and T2 conduct complementarily, and T3 and T4 conduct complementarily. 0 indicates that the switching transistor T is turned off, and 1 indicates that the switching transistor T is turned on, U ab That is, it represents the three voltage levels of the three-level converter;

[0168] HESS adopts a dual-active parallel topology structure composed of supercapacitors and lithium batteries. The control structures of the two are completely symmetrical, only the parameters are different. Taking the supercapacitor as an example, the control method is introduced in detail, and then the control method of HESS is summarized.

[0169] The regenerative energy storage devices available in the urban rail HESS mainly include three working conditions: when the train is in the traction acceleration stage, the energy storage system is in the discharge state; when the train is in the coasting operation stage, the energy storage system is in the holding state; when the train is in the braking deceleration stage, the energy storage system is in the charging state; the energy management of the energy storage system is mainly designed by referring to the standard of the DC traction network voltage as a condition to specifically design the charge and discharge voltage thresholds of the energy storage system as whether HESS is put into use;

[0170] The HESS energy management strategy is studied based on the characteristics of the hybrid energy storage system. Since lithium batteries and supercapacitors have complementary characteristics, in terms of capacity ratio, the capacity of lithium batteries is greater than that of supercapacitors; in terms of energy management, the response priority of supercapacitors is higher than that of lithium batteries. The energy storage device must work within the discharge rate limit during operation, so the SOC constraint charge and discharge limit values are introduced.

[0171]

[0172] In the formula: SOC bat , SOC sc are the state of charge of the lithium battery and the supercapacitor respectively; E bat and E sc are the energies currently stored in the lithium battery and the supercapacitor respectively; E bat_max and E sc_max are the maximum energies stored in the lithium battery and the supercapacitor respectively; C bat , C sc are the capacities of the lithium battery and the supercapacitor respectively, U bat , U sc are the working voltages of the lithium battery and the supercapacitor respectively, U bat_max , U sc_max are the rated voltages of the lithium battery and the supercapacitor respectively.

[0173] During the entire process of charging and discharging, it is also necessary to pay attention that the state of charge of the energy storage device must be within the allowable operating range value.

[0174] SOC bat_min (t) ≤ SOCbat (t) ≤ SOC bat_max (t) (5)

[0175] SOC sc_min (t) ≤ SOC sc (t) ≤ SOC sc_max (t) (6)

[0176] Under normal circumstances, the safe values of the state of charge of lithium batteries and supercapacitors are between [0.2, 0.8].

[0177] The specific design process is as follows:

[0178] The schematic diagram of the working mode switching of the supercapacitor part of the energy storage system is as Figure 4 shown. The lithium battery part is similar to the supercapacitor, so it will not be elaborated here;

[0179] First, combined with Figure 4 analysis, to avoid overcharging and over-discharging of the energy storage system, the prohibited and normal working modes are set according to the state of charge SOC of the energy storage system. Assume that the state of charge SOC of the energy storage system is lower than the set minimum safety value SOC min or higher than the set maximum safety value SOC max , that is, when SOC ≤ SOC min or when SOC ≥ SOC max , it enters the prohibited state; on the contrary, when SOC max ≤ SOC ≤ SOC min , the energy storage works normally. Then, the charge and discharge state of the energy storage system is judged by voltage comparison. In the normal working state of the energy storage system, the specific steps for establishing the prediction model are as follows:

[0180] a) If the SC voltage U SC is higher than the set supercapacitor voltage threshold U sc_dis , at this time, the converter Boost charges / the supercapacitor discharges, with the SC discharge current as the reference direction. At this time, T2 and T3 are the main control tubes, and T1 and T4 are turned off. In this mode, the four switching states are: state 1 (T2 = 1, T3 = 1), state 2 (T2 = 1, T3 = 0), state 3 (T2 = 0, T3 = 0), state 4 (T2 = 0, T3 = 1). Among them, the binary variable T is used to represent the state of the switch tube, 1 represents the switch tube is on, and 0 represents the switch tube is off. The mathematical model of the three-level converter in different switching modes is established as follows:

[0181] Mode 1: T2 is on, T3 is on, and at this time the [T1, T2, T3, T4] switch state is [0, 1, 1, 0];

[0182] The corresponding mathematical model is:

[0183] Mode 2: T2 is turned on and T3 is turned off. At this time, the switching states of [T1, T2, T3, T4] are [0, 1, 0, 1].

[0184] The corresponding mathematical model is:

[0185] Mode 3: T2 is turned off and T3 is turned off. At this time, the switching states of [T1, T2, T3, T4] are [1, 0, 0, 1].

[0186] The corresponding mathematical model is:

[0187] Mode 4: T2 is turned off and T3 is turned on. At this time, the switching states of [T1, T2, T3, T4] are [1, 0, 1, 0].

[0188] The corresponding mathematical model is:

[0189] By performing Euler discretization on the inductor current mathematical models in equations (7) to (10), the values of the inductor current at times k and k + 1 under the corresponding operating modes can be obtained. The specific steps are as follows:

[0190] The derivative of the inductor current di / dt is approximately replaced by the backward Euler formula, as shown in equation (11):

[0191]

[0192] Furthermore, the inductor current i is obtained L That is, the capacitor current i sc The discrete form is as shown in equation (12):

[0193]

[0194] For the derivative of the capacitor voltage dU / dt, the backward Euler approximation is also used, as shown in equation (13):

[0195]

[0196] Furthermore, the discrete form of the capacitor voltage is obtained as shown in equation (14):

[0197]

[0198] Where: U cj (k + 1) and U cj (k) are the sampled values of the capacitor voltage at times k + 1 and k respectively; j = 1, 2; T s is the sampling period.

[0199] Substituting Eqs. (7)-(10) into Eq. (11), the inductor current \(i_{L}(k + 1)\) at the \((k + 1)\)-th moment under the Boost charging / SC discharging condition of the converter can be obtained, that is, the discrete form of the capacitor current \(i_{C}(k + 1)\) is as shown in Eqs. (15)-(18): L \(i_{L}(k + 1)\) sc For the four modes, they are as follows:

[0200] Mode 1:

[0201] Mode 2:

[0202] Mode 3:

[0203] Mode 4: They can be uniformly expressed as: For HESS:

[0204] b) If the SC voltage \(U_{SC}\) SC is lower than the voltage threshold \(U_{SC}^{set}\) sc_char set for the supercapacitor, the energy storage starts the charging process. At this time, the converter is in the Buck discharging / supercapacitor charging state, where T1 and T4 are the main control transistors, and T2 and T3 are turned off. In this mode, the four switching states are: State 1 (T1 = 1, T4 = 1), State 2 (T1 = 1, T4 = 0), State 3 (T1 = 0, T4 = 0), State 4 (T1 = 0, T4 = 1). The mathematical models are established as follows:

[0205] Mode 5: T1 is turned on, T4 is turned on, and the switching state of [T1, T2, T3, T4] is [1, 0, 0, 1];

[0206] The corresponding mathematical model is:

[0207] Mode 6: T1 is turned on, T4 is turned off, and the switching state of [T1, T2, T3, T4] is [1, 0, 1, 0];

[0208] The corresponding mathematical model is:

[0209] Mode 7: T1 is turned off, T4 is turned off, and the switching state of [T1, T2, T3, T4] is [0, 1, 1, 0];

[0210] The corresponding mathematical model is:

[0211] Mode 8: T1 is turned off, T4 is turned on, and the switching state of [T1, T2, T3, T4] is [0, 1, 0, 1];

[0212] The corresponding mathematical model is:

[0213] Substituting Eqs. (21) to (24) into Eq. (5), the discrete form of the capacitor current i sc (k + 1) in the Buck discharging / SC charging mode of the converter can be obtained as Eqs. (25) to (28):

[0214] Mode 5:

[0215] Mode 6:

[0216] Mode 7:

[0217] Mode 8:

[0218] They can be unified as:

[0219] For HESS:

[0220] c) If the voltage U of the supercapacitor SC is within the set operating range, the converter / supercapacitor enters the standby state. That is, when U sc_dis ≤ U sc ≤ U sc_char the converter / supercapacitor is in the standby state.

[0221] Where: L1 and L2 are the upper and lower inductance values in the converter, U sc (k), U b (k) are the voltages of the supercapacitor and the lithium battery on the energy storage side at time k, and U Hess (k) represents the output voltage U sc (k) or U b (k); T s is the sampling period; U c1 (k) and U c2 (k) are the voltage sampling values of C1 and C2 of the voltage-dividing capacitor at time k respectively; U dc is the DC bus voltage; the inductor current i L (k + 1) and the supercapacitor current i sc (k + 1).

[0222] Input the sampling values into the discrete prediction model corresponding to the charging / discharging state to predict the inductor current i L (k + 1), the voltages U c1 (k + 1), U c2 (k + 1) of the voltage-dividing capacitor at the next moment;

[0223] The setting of the objective function J for supercapacitor prediction is shown in Equation (31):

[0224]

[0225] Where P sc (k + 1) is the predicted power value of the supercapacitor at time k + 1. Substituting the discretized equations (19) and (29) into P sc (k + 1) = |i sc (k + 1) × U sc (k + 1)|, the power prediction model corresponding to all switching states of the supercapacitor can be obtained. i sc (k + 1) is the predicted value of the capacitor current at time k + 1, and U c1 (k + 1), U c2 (k + 1) is the predicted value of the voltage-dividing capacitor obtained according to Equation (14) at time k + 1, and U dc (k + 1) = U c1 (k + 1) + U c2 (k + 1); is the power reference value of the supercapacitor, i dc (k) is the current value flowing through the traction network at time k, is the bus reference voltage;

[0226] Then, for HESS, the objective function can be expressed as Equation (32):

[0227]

[0228] Where P Hess (k + 1) is the predicted power value of the energy storage system at time k + 1. Similarly, substituting the discretized equations (20) and (30) into P Hess (k + 1) = |i L (k + 1) × U dc (k + 1)|, the power prediction model corresponding to all switching states of the system can be obtained; U c1 (k + 1), U c2 (k + 1) is the predicted value of the voltage-dividing capacitor obtained according to Equation (14) at time k + 1; U dc (k + 1) = U c1 (k + 1) + U c2 (k + 1); is the power reference value of the energy storage system, is the bus reference voltage, set to 1500V.

[0229] In each cycle when the MPC algorithm works, when the objective function value J corresponding to the switching state is less than the current optimal value J op at this time, Jop is replaced by the new objective function value, and vice versa for the optimal value J op remains unchanged. The optimization of the control objective is achieved by minimizing the objective function. J1, J2, J3, and J4 are the objective functions corresponding to the four prediction models, respectively. The switching state J corresponding to min{J1, J2, J3, J4} is selected min as the switching state for the next moment. It can be seen that after comparing all possible switching states one by one, the switching state S finally acting on the three-level bidirectional DC converter i is unique and also optimal. The optimal switching state S obtained by this algorithm i is updated in real time in each working cycle of the algorithm and is selected according to the objective function value from the following four switching states: when the state j = 1, the switching tubes T2 and T3 are both turned on; when the state j = 2, T2 is turned on and T3 is turned off; when the state j = 3, T2 and T3 are both turned off; when the state j = 4, T2 is turned off and T3 is turned on. Since at each sampling moment, the optimization performance index only involves a finite future time starting from this moment, and by the next sampling moment, this optimization time period moves forward, the optimization process can be repeatedly carried out online, thus realizing cyclic optimization.

[0230] The flowchart for selecting the optimal switching state is as Figure 5 shown;

[0231] For the system operation part, the specific operation steps are as follows:

[0232] Step1: Real-time collect the traction network terminal voltage U dc and the supercapacitor SOC sc , and use the SOC sc as the input quantity of the fuzzy system;

[0233] Step2: Run the fuzzy inference system to obtain the corresponding dynamic setting values U bat_dis , U bat_char ;

[0234] Step3: Compare U dc with the charge and discharge thresholds of the energy storage system to judge the charge and discharge state of the system;

[0235] Step4: At the corresponding state, sample the input voltage U dc (k) and the input current i L (k) of the converter at the current moment k by the sensor, as well as the terminal voltage U b (k), U sc (k) and the terminal current i b (k), i sc (k) of the hybrid energy storage system;

[0236] Step 5: The computer calculates the predicted power value and the reference value of HESS at the k+1 moment respectively, and then calculates the objective function values J corresponding to all switching states according to the objective function, and selects the smaller value J min The corresponding switching state S i As the switching state at the k+1 moment, S i The switching signals of the 4 IGBT tubes T1, T2, T3, and T4 are composed of 0 and 1;

[0237] Step 6: Apply the new switching state to the DC three-level converter;

[0238] Step 7: Return to Step 1 to start the optimization of the next switching state; if a fault occurs in the energy storage system or the DC three-level converter, the program ends.

[0239] The present invention is not limited to the above examples. Of course, the present invention can also have many other implementation manners and can also be applied to other energy storage fields. The above specific implementation manners are only illustrative and not restrictive. Without departing from the spirit and essence of the present invention, those skilled in the art can make corresponding changes and deformations according to the present invention, but these corresponding changes and deformations should all fall within the protection scope of the claims of the present invention.

Claims

1. A model predictive control (MPC) and dynamic setting method without PI for a three-level DC / DC converter based on urban rail hybrid energy storage, characterized in that, Based on a non-isolated DC three-level converter, combined with a fuzzy inference system and a PI-free power MPC control method; by establishing fuzzy rules, according to the real-time monitoring of the state of charge (SOC sc ) of the supercapacitor, the charging and discharging thresholds of the lithium battery are dynamically generated; after passing through the switch selection module, the corresponding real-time signal input quantity is generated, combined with the PI-free power MPC control method, to realize the dynamic adjustment of the charging and discharging state of the energy storage system, as well as the optimization of energy management and lithium battery protection. The present invention is divided into a design part and an operation part, and is realized as follows: Design Part: This design relates to a coordinated control method for a hybrid energy storage system in rail transit. Aiming at the problem of insufficient energy management coordination between power-type supercapacitors and energy-type lithium batteries under the existing fixed voltage threshold strategy, a dynamic threshold setting method based on fuzzy control is proposed. This design constructs a collaborative architecture with supercapacitors responding first and lithium batteries compensating dynamically, constrained by the voltage fluctuation characteristics of the DC traction network. In the design, the discharge voltage threshold U of the supercapacitor sc_dis is initially set to 1480V, and the charge voltage threshold U sc_char is 1520V; by collecting the real-time feedback value of the supercapacitor SOC sc , the optimization parameter set of the lithium battery charge and discharge voltage thresholds (U bat_dis / U bat_char ) is dynamically adjusted using the fuzzy logic control algorithm to enable the hybrid energy storage system (HESS) to achieve coordinated control of power and energy. Compared with the traditional fixed threshold method, it realizes the dynamic coordination of the working modes of energy storage components. At the initial stage of train acceleration / braking, the supercapacitor quickly absorbs / releases high-power components. When the SOC value of the supercapacitor deviates from the set range, the voltage action boundary of the lithium battery is autonomously adjusted through fuzzy rules to trigger its energy compensation function. This design realizes the efficient distribution of the energy of the energy storage system and the coordinated optimization of energy management by optimizing the output of the supercapacitor and the lithium battery. Design a fuzzy inference system for charge and discharge thresholds, including membership function selection, membership function expression, determination of fuzzy domain, determination of fuzzy subsets, and establishment of fuzzy inference rules. The fuzzy inference unit selects the supercapacitor SOC sc as the input, and the lithium battery charging threshold (U bat_char ), and the discharging threshold (U bat_dis ) as the outputs; and they are set as follows: 1) Selection of membership function for fuzzy inference, determination of fuzzy domain and fuzzy subsets: SOC sc Seven fuzzy subsets are selected: NL (Negative Large), NM (Negative Medium), NS (Negative Small), ZO (Zero), PS (Positive Small), PM (Positive Medium), PL (Positive Large); Lithium-ion battery charging voltage threshold U bat_char , seven fuzzy subsets are selected: NL (Negative Large), NM (Negative Medium), NS (Negative Small), ZO (Zero), PS (Positive Small), PM (Positive Medium), PL (Positive Large); Lithium battery discharge voltage threshold U bat_dis , seven fuzzy subsets are selected: NL (negative large), NM (negative medium), NS (negative small), ZO (zero), PS (positive small), PM (positive medium), PL (positive large); SOC sc Fuzzy universe of discourse [0.2 0.8]; U bat_char Fuzzy domain of discourse [1510 1530]; U bat_dis Fuzzy universe of discourse [1470 1490]; SOC sc 、U bat_char 、U bat_dis The membership function adopts a trigonometric function formula or a Gaussian formula; Triangle function formula: It is required that a ≤ b ≤ c, where a, b, and c are the left, middle, and right coordinates of the membership degree respectively, and x is the input value. Gaussian type: Among them, The position of the center of the c membership function, σ is the width of the membership function curve, and x is the input value. 2) Fuzzy inference rules: Discharge state: If(SOC sc is PL)then(U bat_dis is NL) If(SOC sc is PM)then(U bat_dis is NM) If(SOC sc is PS)then(U bat_dis is NS) If(SOC sc is ZO)then(U bat_dis is ZO) If(SOC sc is NS)then(U bat_dis is PS) If(SOC sc is NM)then(U bat_dis is PM) If(SOC sc is NL)then(U bat_dis is PL) Charge state: If(SOC sc is NL)then(U bat_char is PL) If(SOC sc is NM)then(U bat_char is PM) If(SOC sc is NS)then(U bat_char is PS) If(SOC sc is ZO)then(U bat_char is ZO) If(SOC sc is PS)then(U bat_char is NS) If(SOC sc is PM)then(U bat_char is NM) If(SOC sc is PL)then(U bat_char is NL) Taking the supercapacitor SOC sc as the fuzzy input, through the fuzzy inference system, the charging and discharging thresholds of the lithium battery voltage are obtained, and the method of combining the non-PI power MPC control is adopted to solve the disadvantage of difficult setting of the PI controller control parameters in the PI double closed-loop control and the current prediction control, so that the energy storage system has a better dynamic response speed, gives full play to the fast charge and discharge characteristics of the supercapacitor, and further protects the lithium battery. The utilization degree of the energy storage system tends to be balanced and reasonable, avoiding overcharge and overdischarge of the supercapacitor. At the same time, a three-level DC converter is introduced to provide an additional voltage level selection for the converter, reduce the voltage stress of the switching tubes and diodes, reduce the switching loss, reduce the voltage fluctuation amplitude, and ensure the stability and reliability of the output of the energy storage system; For the non-PI power MPC control method of a three-level DC converter, it is characterized in that the three-level converter adopts a non-isolated three-level DC / DC topology structure. The converter topology consists of 4 IGBT tubes T1, T2, T3, T4, two equalizing capacitors C1, C2, and inductors L1, L2. The low-voltage side is connected to the hybrid energy storage system, and the high-voltage side is connected to the urban rail traction network. This method obtains the optimal switching state through power prediction, without a PI link, simplifies the control structure, omits the adjustment of multiple parameters, and can significantly improve the dynamic response speed of the system. For different working modes of the energy storage system, corresponding prediction equations can be established. For the switching state of the three-level DC / DC converter, the switching state can be summarized into 4 modes in the Boost / Buck mode. The switching modes during the operation of the converter are shown in Table (1): (Table 1) When the switching tube T works, T1 and T2 conduct complementarily, and T3 and T4 conduct complementarily. 0 indicates that the switch tube T is turned off, and 1 indicates that the switch tube T is turned on. U ab represents the three voltage levels of the three-level converter. U dc is the DC traction network voltage; The specific design process is as follows: a) If the Boost charging / storage system of the converter discharges, with the discharge as the reference direction, at this time, T2 and T3 are the main control transistors, and T1 and T4 are turned off. The switching transistors have the following four modes [T2 T3] = {[00], [01], [10], [11]}. Based on the equivalent circuits corresponding to all switching states in these four switching states, a current prediction model is established, and after discretization, the inductor current i of the HESS at the (k + 1)th moment can be obtained. L (k + 1) As shown in Equation (3), the predicted values of the voltages of the voltage-dividing capacitors C1 and C2 are shown in Equations (4) and (5); b) If the converter Buck discharges and the energy storage system charges, with the discharge as the reference direction, at this time, T1 and T4 are the main control transistors, and T2 and T3 are turned off. The switching transistors have the following four modes [T1 T4] = {[00], [01], [10], [11]}. Under these four switching states, the inductor current i of the HESS at time k + 1 L (k + 1) is as shown in Equation (6), and the predicted values of the voltages of the voltage-dividing capacitors C1 and C2 are as shown in Equations (7) and (8); Where: L1 and L2 are the upper and lower inductance values in the converter, U sc (k), U b (k) are the voltages of the supercapacitor and the lithium battery on the energy storage side at time k, U Hess (k) represents the output voltage U sc (k) or U b (k); T s is the sampling period; U c1 (k) and U c2 (k) are the voltage sampling values of C1 and C2 of the voltage-dividing capacitor at time k respectively; U c1 (k + 1) and U c2 (k + 1) are the voltage prediction values of C1 and C2 of the voltage-dividing capacitor at time k + 1 respectively; i c1 (k) and i c2 (k) are the currents of C1 and C2 of the voltage-dividing capacitor at time k respectively; Set the prediction objective function J of the energy storage system, as shown in Equation (9): where P Hess (k + 1) is the predicted power value of the energy storage system at the (k + 1)-th moment. Substitute the discretized equations (3) and (6) into P Hess (k + 1) = |i L (k + 1) × U dc (k + 1)| to obtain the power prediction model corresponding to all switch states; U c1 (k + 1), U c2 (k + 1) is the predicted value of the voltage-dividing capacitor obtained from equations (4), (5), (7), and (8) at the (k + 1)-th moment; U dc (k + 1) = U c1 (k + 1) + U c2 (k + 1); is the power reference value of the energy storage system, i dc (k) is the current value flowing through the traction network at the k-th moment, is the bus reference voltage, set to 1500V; In each cycle when the MPC algorithm works, when the objective function value J corresponding to the switch state is less than the current optimal value J op , J op is replaced by the new objective function value. Conversely, the optimal value J op remains unchanged. The optimization of the control objective is achieved by minimizing the objective function. J1, J2, J3, and J4 are the objective functions corresponding to 4 prediction models respectively. The switch state J min corresponding to min{J1, J2, J3, J4} is selected as the switch state at the next moment. It can be seen that after comparing all possible switch states one by one, the switch state S i acting on the three-level bidirectional DC converter finally is unique and also optimal. The optimal switch state S i obtained by this algorithm is updated in real time in each working cycle of the algorithm and is selected according to the objective function value from the following 4 switch states: when the state j = 1, both switch tubes T2 and T3 are turned on; when the state j = 2, T2 is turned on and T3 is turned off; when the state j = 3, both T2 and T3 are turned off; when the state j = 4, T2 is turned off and T3 is turned on. Since at each sampling moment, the optimization performance index only involves a limited time in the future starting from this moment, and by the next sampling moment, this optimization time period moves forward, the optimization process can be repeated online, thus realizing cyclic optimization; In the system operation part, the specific operation steps are as follows: Step1: Real-time collect the traction network terminal voltage U dc and the supercapacitor SOC sc , and use the SOC sc as the input quantity of the fuzzy system; Step 2: Run the fuzzy inference system to obtain the corresponding dynamic setpoint U bat_dis , U bat_char ; Step3: Compare U dc with the charge and discharge thresholds of the energy storage system to determine the charge and discharge state of the system; Step4: Sample the input voltage U(k) and input current i(k) of the converter at the current moment k under the corresponding state, as well as the terminal voltage U(k) and terminal current i(k) of the hybrid energy storage system; dc (k) and input current i L (k), and the terminal voltage U b (k), U sc (k) and terminal current i b (k), i sc (k); Step 5: The computer calculates the predicted power value and the reference value of HESS at the (k + 1)-th moment respectively, and then calculates the objective function values J corresponding to all switching states according to the objective function, and selects the smaller value J min The corresponding switching state S i As the switching state at the (k + 1)-th moment, S i is the switching signal composed of 0 and 1 for the 4 IGBT tubes T1, T2, T3, and T4; Step6: Apply the new switching state to the three-level DC converter. Step7: Return to Step1 to start the optimization of the next switching state. If a fault occurs in the energy storage system or the three-level DC converter, the program ends.

Images