DC-DC-based non-contact power supply series constant current control algorithm

Abstract

The application discloses a DC-DC-based non-contact power supply series constant current control algorithm and belongs to the field of non-contact power supply, and the control algorithm has the following specific steps: I, a dynamic mathematical model of the non-contact power supply system is established, and the current change trend at multiple future time points is predicted in each sampling period; II, according to the load prediction value and the working point of the resonant network, the compensation amount is calculated, the external disturbance is responded, the current signal is collected in real time, and the error is corrected through closed-loop adjustment; the application can significantly improve the track current tracking accuracy, avoid the lag and overshoot of the traditional PI / PID under dynamic disturbance, reduce the artificial threshold setting, improve the adaptability and fault tolerance capability under complex working conditions, ensure the energy efficient transmission, reduce the current impact and switching loss when transient switching or load mutation occurs, reduce the device stress and improve the system reliability.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of non-contact power supply, in particular to a non-contact power supply series constant current control algorithm based on DC-DC. BACKGROUND

[0002] Non-contact power transmission utilizes magnetic coupling or electromagnetic resonance mechanism to realize energy transmission between power supply end and power receiving end, avoiding wear, spark risk and maintenance cost caused by mechanical contact. In some application scenarios (such as track power supply, track vehicle ground energy transmission, AGV wireless power supply, mobile power supply of factory automation, etc.), the system requires continuous and stable current output, and often faces challenges such as load mutation, device switching and environmental disturbance. For series resonant type non-contact system, maintaining constant current on the primary side (track side) is not only related to power supply quality, but also directly affects coupling efficiency and device life. Traditional control strategies usually use PI / PID based closed loop regulation, frequency tracking (keeping resonance) and fixed point voltage adjustment methods. These methods can achieve good control in steady state, but are limited in strong disturbance or defect (device failure, switch delay) situations; therefore, it is particularly important to invent a non-contact power supply series constant current control algorithm based on DC-DC.

[0003] The track current tracking accuracy in the existing non-contact power supply series constant current control algorithm is low, and PI / PID has lag and overshoot under dynamic disturbance, artificial threshold setting is more, reducing the adaptability and fault tolerance ability under complex working conditions, which cannot guarantee energy efficient transmission, increases current impact and switching loss when transient switching or load mutation occurs; therefore, we propose a non-contact power supply series constant current control algorithm based on DC-DC. SUMMARY

[0004] The purpose of the present application is to solve the defects in the prior art, and to propose a non-contact power supply series constant current control algorithm based on DC-DC.

[0005] In order to achieve the above purpose, the technical scheme adopted by the present application is as follows:

[0006] A non-contact power supply series constant current control algorithm based on DC-DC, the specific steps of the control algorithm are as follows:

[0007] I. Establish a dynamic mathematical model of the non-contact power supply system, and predict the current change trend at multiple future time points in each sampling period;

[0008] II. According to the load prediction value and the working point of the resonant network, calculate the compensation amount, respond to external disturbance, real-time collect current signal, and correct error through closed loop regulation;

[0009] III. Identify the real-time running state of the non-contact power supply system, autonomously judge according to historical data and real-time input during the running process, and start the optimal control action;

[0010] IV. According to different load conditions, automatically execute multi-frequency collaborative impedance matching, and simultaneously estimate the current state in real time according to the system model;

[0011] V. Plan the optimal switching trajectory of energy transfer, simultaneously establish a system simulation model, and perform real-time mapping and multi-time scale prediction on the running state.

[0012] Compared with the prior art, the beneficial effects of the present application are:

[0013] The present application establishes the mapping of DC-DC duty cycle to output voltage, and models the coupling coil and load with resonant inductance, capacitance and equivalent resistance, forming a discretized state space model with inductance current and capacitance voltage as the state. Based on this model, MPC predicts the track current at each sampling period and solves the optimal DC-DC voltage adjustment amount by QP. In parallel, DQN is introduced to make autonomous decisions based on the state. Frequency and inverter output voltage are collaboratively adjusted to achieve multi-frequency impedance matching for maximum power transmission. Then a piecewise linear parameterized resonant soft switching sequence is constructed, and safe and smooth switching timing and voltage ramp are obtained through offline NLP and online fast optimization, with slope, dead zone and energy constraints. After that, the sampling signal is estimated by Kalman filtering, and the parameters are identified online by RLS. A system simulation model is constructed for short-term prediction to support control and long-term life prediction to drive preventive maintenance or non-inductive switching. Finally, the optimization results are sent to FPGA / drive, and safety is guaranteed through fault detection and rollback. This can significantly improve the tracking accuracy of the track current, avoid the lag and overshoot of traditional PI / PID under dynamic disturbance, reduce the artificial threshold setting, improve the adaptability and fault tolerance capability under complex working conditions, ensure efficient energy transmission, reduce current impact and switching loss during transient switching or load mutation, reduce device stress and improve system reliability. BRIEF DESCRIPTION OF DRAWINGS

[0014] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, to explain the present application, and do not constitute a limitation of the present application.

[0015] Figure 1 A flowchart of a DC-DC-based non-contact power supply series constant current control algorithm is proposed. DETAILED DESCRIPTION

[0016] REFERENCE Figure 1 A DC-DC-based non-contact power supply series constant current control algorithm, the specific steps of which are as follows:

[0017] A dynamic mathematical model of the non-contact power supply system is established, and the current trend at multiple future time points is predicted within each sampling period.

[0018] Specifically, the input DC bus voltage of the DC-DC converter and the duty cycle are calculated to obtain the output voltage of the DC-DC converter, and then a DC-DC output voltage model is established based on the relationship between the output voltage and the duty cycle. A resonance network is established by inductance, capacitance and equivalent resistance, and the DC-DC output voltage is added to the resonance network through the inverter. Then, the load is modeled based on the energy supply information from the coupling coil through electromagnetic coupling, so as to obtain the electrical relationship of the energy transmission from the resonance network to the load in real time. Based on the coupling inductance between the primary side and the secondary side, a coupling inductance model is established, and the inductance current and the capacitance voltage are taken as state variables to establish the state space equation of the non-contact power supply system, to generate the corresponding dynamic mathematical model, so as to obtain the dynamic behavior of the resonance network in real time. The final track current is obtained by superimposing the primary side resonance current and the mutual inductance effect.

[0019] Specifically, at each sampling time, the discrete form of the future trajectory of the track current, i.e. the predicted track current, is obtained based on the established dynamic mathematical model according to the preset prediction time domain length and the control time domain length. Then, the current tracking error between the predicted track current at the current time and the preset ideal current trajectory is calculated. Based on the DC-DC converter output voltage and the current tracking error, a corresponding optimization objective function is established, and the lower limit of the DC-DC output voltage and the upper limit of the DC-DC output voltage are taken as constraint conditions. Then, at each sampling time, a quadratic programming problem of the current non-contact power supply system is established according to the predicted track current, the current tracking error, the optimization objective function and the constraint conditions. Then, the quadratic programming problem is solved to obtain the optimal control sequence. The first group of control quantities in the optimal control sequence are extracted and applied, and then the current DC-DC output voltage is adjusted based on the extracted control quantities.

[0020] According to the load prediction value and the working point of the resonance network, the compensation amount is calculated to respond to external disturbances, and the current signal is collected in real time to correct the error through closed-loop regulation.

[0021] The real-time running state of the non-contact power supply system is identified, and the optimal control action is started according to the historical data and real-time input during the running process.

[0022] Specifically, the system state at each time is collected, and a corresponding system state space is established, including the track voltage, the junction temperature of the power device, the actual output power of the system, and the health state index of the power cabinet. Each control action of the non-contact power supply system is set, and a corresponding action space is established, including maintaining the current running mode, performing voltage compensation operation, and starting the switching of the redundant power cabinet. According to the track current deviation improvement amount, the system energy loss increase amount, and the switching cost, a reward function is established. Then, the Q value under any system state and control action at the current time is calculated, and the immediate reward obtained by the system at the current time is calculated through the reward function. Based on the immediate reward of the current system and all candidate actions at the next system state, the current Q value is updated using the Bellman equation. In the real-time running of the system, the control action with the maximum Q value is selected through the epsilon-greedy strategy or a random control action is selected. In each sampling period, the selected control action is directly fed back to the non-contact power supply system.

[0023] Reference Figure 1 A DC-DC-based non-contact power supply series constant current control algorithm, the specific steps of which are as follows:

[0024] According to different load conditions, automatic execution of multi-frequency collaborative impedance matching, and real-time estimation of current state according to system model.

[0025] Specifically, based on the equivalent resistance, equivalent inductance, equivalent capacitance and angular frequency of the resonant network, the equivalent input impedance of the primary side is determined by the coupling coil and the load to achieve maximum power transmission as the impedance matching condition. According to the equivalent inductance and the equivalent capacitance, the corresponding resonant frequency is calculated. According to the voltage regulation mechanism, the real-time output power is obtained through the effective value of the inverter output voltage at each time and the load impedance of the primary side. In actual operation, the system maintains constant current power supply and high efficiency operation under different impedance conditions by adjusting the effective value of the inverter output voltage. The coupling resonant frequency and the real-time output power are matched, and the collaborative matching control law is formed. Based on the established collaborative matching control law, the system restores the matching state when the load changes through the joint adjustment of frequency and the effective value of the inverter output voltage. Then, during system operation, the instantaneous efficiency of the system at the current time is calculated through the input power and the output power at the current time. If the instantaneous efficiency of the system at the current time is lower than the preset threshold, the frequency deviation and the voltage compensation are adjusted, and the real-time working frequency and the real-time output voltage are iteratively updated through the collaborative matching control law until the instantaneous efficiency of the system reaches the preset threshold.

[0026] Optimal switching trajectory of energy transfer is planned, and a system simulation model is established to map and predict the running state in multiple time scales in real time.

[0027] Specifically, the resonance energy deviation metric, the switching loss sum and the switching smoothness metric are taken as the three main metrics of the soft switching sequence, and a comprehensive cost function is established according to the three main metrics selected, the switching interval is divided into multiple sub-sections, and a switch event time vector is set, the DC-DC output voltage trajectory of each time section is calculated by using a piecewise linear function, then the instantaneous energy of the resonance network during switching is calculated according to the resonance equivalent inductance and the resonance equivalent capacitance, then the state variables of the resonance branch current are calculated according to the instantaneous energy of the resonance network during switching, the instantaneous switching loss power density at each time in the switching interval is calculated according to the instantaneous voltage across the switch and the switch current function, and the energy loss of each switch event is obtained according to the switching loss power density at each time, then the corresponding constraint conditions are set according to the voltage slope, the switch timing and the dead zone, the energy boundary and the total loss upper limit allowed in the switching process, the switching interval is discretized, and each discrete time index is set to the corresponding time, and at the same time, the voltage slope value at the discrete time, the switch state vector, the resonance current state and the capacitance voltage state at the discrete time are calculated, and the instantaneous energy of the resonance network during switching is calculated step by step, in the offline stage, the switch timing and the slope change parameters of the output voltage are optimized based on the nonlinear programming NLP to obtain a set of candidate parameters and a corresponding switching library, in the online stage, the parameters obtained in the offline stage are taken as the initial values, a gradient type method with limited iteration times is used for local optimization, the optimal parameter set is obtained by looking up or interpolating the working condition parameters measured in real time in the offline library, the switching interval in the parameter set obtained by optimization is converted into a timer trigger point, which is written into the event table of the IGBT driver or the FPGA, then the piecewise linear coefficient is converted into the DAC / PWM reference of the voltage slope, or into the reference trajectory of the DC-DC control loop, and detection is carried out before and after each switching action, if the detection is abnormal, a rollback or a safety shutdown sequence is triggered.

[0028] Specifically, the functions of real-time mapping of the electrical, thermal and stress states of the physical system, short-time scale prediction to support the controller, and long-time scale prediction of element aging and remaining life to support preventive maintenance are taken as the functional objectives of the system simulation model. A continuous-time model based on circuit laws is adopted. The instantaneous electromagnetic state is represented by the primary-side flux linkage and the secondary-side flux linkage. The relationship between the current and the output voltage is obtained based on the primary-side flux linkage and the secondary-side flux linkage to establish an electromagnetic model. The instantaneous loss calculated by the electromagnetic model is taken as the input. The equivalent thermal capacity of each device is calculated according to the device junction temperature, the environmental temperature, the thermal resistance of the actual environment and the instantaneous loss power. A corresponding single-junction temperature first-order thermal model is established. The temperature acceleration and cumulative damage accumulation rules are adopted. The cumulative damage model is constructed according to each stress factor. The device health index is calculated in real time through the cumulative damage model. When the device health index is lower than the preset health index, it is determined that the device needs maintenance and replacement. The electromagnetic model, the single-junction temperature first-order thermal model and the cumulative damage model are coupled as the system simulation model. The overall state vector in the form of a continuous nonlinear system is output. Then, the recursive model is obtained by first-order linearization and discretization at the current state point. The sensor observation data is synchronized to the system simulation model. The estimation is performed through Kalman filtering. The recursive least squares method is adopted. The parameters are updated based on the forgetting factor. Then, the prediction system state at the next time is output based on the matrix recursion of the system simulation model. At the same time, the short-term prediction output is taken as the initial value or constraint of each controller. Then, the long-time scale prediction and the remaining life assessment are performed based on the system simulation model through the Monte Carlo or deterministic integral method. Then, the prediction output is mapped to the operation and maintenance or control action. At the same time, the maintenance or non-inductive switching is planned according to the prediction results to trigger the early warning.

Claims

1. A non-contact power supply series constant current control algorithm based on DC-DC, characterized in that, The specific steps of the control algorithm are as follows: I. Establish a dynamic mathematical model of the non-contact power supply system and predict the current change trend at multiple future moments within each sampling period; specifically: Based on the relationship between the input bus voltage and duty cycle of the DC-DC converter, the output voltage of the DC-DC converter is calculated, and a DC-DC output voltage model is established based on the relationship between the output voltage and the duty cycle. Then, combined with the inverter, resonant network and load equivalent model, the electrical characteristics of the power transmission to the load through the coupling coil are obtained in real time. Based on the coupling inductance of the primary and secondary sides, the state space equation of the non-contact power supply system is established, and the corresponding dynamic mathematical model is generated. Finally, the track current is obtained by superimposing the primary resonant current and the mutual inductance effect. At each sampling moment, the current tracking error between the predicted track current and the preset ideal current trajectory is calculated based on the dynamic mathematical model. Based on the output voltage and current tracking error of the DC-DC converter, a corresponding optimization objective function is established, and constraints are set. Based on the predicted track current, current tracking error, optimization objective function, and constraints, the optimal control sequence is obtained. Then, the first set of control quantities in the optimal control sequence is extracted and applied. The current DC-DC output voltage is adjusted based on the extracted control quantities. II. Based on the load prediction value and the operating point of the resonant network, calculate the compensation amount, respond to external disturbances, collect current signals in real time, and correct errors through closed-loop adjustment; III. Identify the real-time operating status of the contactless power supply system, and autonomously judge based on historical data and real-time input during operation, and initiate optimal control actions: collect the system status at each moment, and establish a corresponding system state space and corresponding action space, including maintaining the current operating mode, performing voltage compensation operations, and initiating the switching of redundant power supply cabinets; establish a reward function, It calculates the Q-value of the system under any system state and control action at the current moment, and calculates the system Q-value at the current moment using the reward function. The instant reward received; During real-time system operation, the control action with the largest Q value is selected or a control action is randomly selected. In each sampling period, the selected control action is directly fed back to the non-contact power supply system. IV. Based on different load conditions, the system automatically performs multi-frequency coordinated impedance matching and simultaneously estimates the current state in real time based on the system model: it couples the resonant frequency with the real-time output power and forms a coordinated matching control law to obtain the instantaneous efficiency of the system at the current moment. The instantaneous efficiency of the system at the current moment is compared with a preset threshold, thereby adjusting the frequency deviation and voltage compensation. V. Plan the optimal switching trajectory for energy transfer, establish a system simulation model, and perform real-time mapping and multi-timescale prediction of the operating state: Construct a comprehensive cost function that includes resonant energy deviation metric, total switching loss, and switching smoothness metric, and set corresponding constraints based on voltage slope, switching timing and dead zone, energy boundary, and the upper limit of the allowable total loss during switching; In the offline phase, obtain a set of candidate parameters and corresponding switching libraries for various parameters of the switching timing and output voltage ramp change optimized based on nonlinear programming (NLP); In the online phase, use the parameters obtained in the offline phase as initial values, and obtain the optimal parameter set based on the real-time measured operating parameters; Convert the switching intervals in the optimal parameter set into timer trigger points and write them into the event table of the IGBT driver or FPGA; Perform detection before and after each switching action, and if an abnormality is detected, trigger rollback or enter a safe shutdown sequence.

2. The non-contact power supply series constant current control algorithm based on DC-DC as described in claim 1, characterized in that, The specific steps for establishing the dynamic mathematical model of the non-contact power supply system described in step I are as follows: S1.1: Calculate the output voltage of the DC-DC converter based on the input DC bus voltage and duty cycle, and then establish a DC-DC output voltage model based on the relationship between the output voltage and the duty cycle. S1.2: A resonant network is established through inductance, capacitance and equivalent resistance. The DC-DC output voltage is then applied to the resonant network through an inverter. Then, based on the information that the coupled coil supplies energy to the load through electromagnetic coupling, the load is modeled in an equivalent manner to obtain the electrical relationship of the power transmission from the resonant network to the load in real time. S1.3: Based on the existence of coupled inductance between the primary and secondary sides, a coupled inductance model is established. At the same time, the inductor current and capacitor voltage are used as state variables to establish the state space equation of the non-contact power supply system and generate the corresponding dynamic mathematical model to obtain the dynamic behavior of the resonant network in real time. Then, the final track current is obtained by superimposing the primary side resonant current and the mutual inductance effect.

3. The non-contact power supply series constant current control algorithm based on DC-DC as described in claim 2, characterized in that, The specific steps for predicting the current change trend at multiple future moments, as described in Step I, are as follows: S2.1: At each sampling time, based on the preset prediction time domain length and control time domain length, and based on the established dynamic mathematical model, the discretized form of the future trajectory of the track current is obtained, i.e., the predicted track current. Then, the current tracking error between the predicted track current at the current time and the preset ideal current trajectory is calculated. S2.2: Based on the DC-DC converter output voltage and current tracking error, establish a corresponding optimization objective function, and use the lower limit of DC-DC output voltage and the upper limit of DC-DC output voltage as constraints. S2.3: Then, at each sampling time, based on the predicted track current, current tracking error, optimization objective function and constraints, a quadratic programming problem for the current contactless power supply system is established. The quadratic programming problem is then solved to obtain the optimal control sequence. The first set of control quantities in the optimal control sequence is then extracted and applied. Finally, the current DC-DC output voltage is adjusted based on the extracted control quantities.

4. The non-contact power supply series constant current control algorithm based on DC-DC as described in claim 3, characterized in that, Step III involves identifying the real-time operating status of the contactless power system and making autonomous judgments based on historical data and real-time input during operation. The specific steps are as follows: S3.1: Collect the system status at each moment and establish the corresponding system state space, including the rail voltage, junction temperature of power devices, actual system output power and power cabinet health status indicators. Then set the control actions of the non-contact power system and establish the corresponding action space, including maintaining the current operating mode, performing voltage compensation operation and starting the switching of redundant power cabinets. S3.2: Based on the improvement of track current deviation, the increase in system energy loss and the switching cost, establish a reward function, then calculate the Q value under any system state and control action at the current moment, and calculate the instantaneous reward obtained by the system at the current moment through the reward function; S3.3: Based on the immediate reward of the current system and all candidate actions in the next system state, update the current Q value using the Bellman equation. During real-time operation of the system, select the control action with the largest Q value or randomly select a control action through an ε-greedy strategy. In each sampling period, the selected control action is directly fed back to the non-contact power supply system.

5. The non-contact power supply series constant current control algorithm based on DC-DC as described in claim 4, characterized in that, The specific steps for automatically performing multi-frequency coordinated impedance matching based on different load conditions, as described in step IV, are as follows: S4.1: Based on the equivalent resistance, equivalent inductance, equivalent capacitance, and angular frequency of the resonant network, the equivalent input impedance on the primary side is determined by the coupling coil and the load. The impedance matching condition is to achieve maximum power transmission. The corresponding resonant frequency is calculated based on the equivalent inductance and equivalent capacitance. S4.2: Based on the voltage regulation mechanism, the real-time output power is obtained by the effective value of the inverter output voltage at each moment and the load impedance on the primary side. In actual operation, the effective value of the inverter output voltage is adjusted to enable the system to maintain constant current power supply and high-efficiency operation under different impedance conditions. S4.3: Couple the resonant frequency and real-time output power to form a cooperative matching control law. Based on the established cooperative matching control law, the system restores the matching state when the load changes by jointly adjusting the frequency and the effective value of the inverter output voltage. After the system runs, the instantaneous efficiency of the system at the current moment is calculated by using the input power and output power at the current moment. S4.4: If the instantaneous efficiency of the system is lower than the preset threshold at the current moment, the frequency deviation and voltage compensation are adjusted, and the real-time operating frequency and real-time output voltage are iteratively updated again through the cooperative matching control law until the instantaneous efficiency of the system reaches the preset threshold.

6. The non-contact power supply series constant current control algorithm based on DC-DC as described in claim 1, characterized in that, The specific steps for planning the optimal switching trajectory for energy transfer as described in step V are as follows: S5.1: The resonant energy deviation metric, the total switching loss metric, and the switching smoothness metric are taken as the three main metrics of the soft switching sequence. A comprehensive cost function is established based on the three main metrics selected, the switching interval is divided into multiple sub-segments, and the switching event time vector is set. S5.2: Calculate the DC-DC output voltage trajectory for each time period using a piecewise linear function, then calculate the instantaneous energy of the resonant network during the switching period based on the resonant equivalent inductance and resonant equivalent capacitance, and finally calculate the state variable of the resonant branch current based on the instantaneous energy of the resonant network during the switching period. S5.3: Calculate the instantaneous switching power density at each moment in the switching interval based on the instantaneous voltage at both ends of the switch and the switching current function, and obtain the energy loss of each switching event based on the switching power density at each moment. Then set corresponding constraints based on the voltage slope, switching sequence and dead zone, energy boundary and the upper limit of the total allowable loss during the switching process. S5.4: Discretize the switching interval and set the corresponding time for each discrete time index. At the same time, calculate the instantaneous energy of the resonant network during the switching period based on the voltage ramp value, switch state vector, resonant current state and capacitor voltage state at the discrete time. S5.5: In the offline phase, the switching timing and output voltage ramp change parameters are optimized based on nonlinear programming (NLP) to obtain a set of candidate parameters and corresponding switching libraries. In the online phase, the parameters obtained in the offline phase are used as initial values, and a gradient method with a limited number of iterations is used for local optimization. The optimal parameter set is obtained by searching or interpolating in the offline library based on the real-time measured operating parameters. S5.6: Convert the switching interval in the optimized parameter set into timer trigger points and write it into the event table of the IGBT driver or FPGA. Then, convert the piecewise linear coefficients into the DAC / PWM reference of the voltage ramp, or into the reference trajectory of the DC-DC control loop, and perform detection before and after each switching action. If an abnormality is detected, trigger rollback or enter the safe shutdown sequence.

7. The non-contact power supply series constant current control algorithm based on DC-DC as described in claim 1, characterized in that, The specific steps for real-time mapping and multi-timescale prediction of the operating state described in step V are as follows: S6.1: The functional goal of the system simulation model is to map the electrical, thermal and stress states of the physical system in real time, make short-term predictions to support the controller, and make long-term predictions of component aging and remaining life to support preventive maintenance. S6.2: A continuous-time model based on circuit laws is adopted. The instantaneous electromagnetic state is represented by the primary side flux linkage and the secondary side flux linkage. The relationship between current and output voltage is obtained based on the primary side flux linkage and the secondary side flux linkage to establish an electromagnetic model. The instantaneous loss calculated by the electromagnetic model is then used as input. The equivalent heat capacity of each device is calculated based on the device junction temperature, ambient temperature, actual ambient thermal resistance, and instantaneous loss power. A first-order thermal model of the corresponding single-unit junction temperature is established. S6.3: Using temperature acceleration and cumulative damage accumulation rules, a cumulative damage model is constructed based on each stress factor, and the device health index is calculated in real time through the cumulative damage model. When the device health index is lower than the preset health index, it is determined that the device needs maintenance and replacement. S6.4: Couple the output results of the electromagnetic model, the first-order thermal model of the single-unit junction temperature and the cumulative damage model into a system simulation model, and output the global state vector in a continuous nonlinear form. Then, at the current state point, obtain the recursive model through first-order linearization and discretization, synchronize the sensor observation data to the system simulation model, and estimate it through Kalman filtering. S6.5: The recursive least squares method is adopted to update the parameters based on the forgetting factor. Then, based on the system simulation model, the system state at the next moment is predicted by matrix recursion. At the same time, the short-term prediction output is used as the initial value or constraint of each controller. Then, the long-term scale prediction and remaining lifetime assessment are performed based on the system simulation model by Monte Carlo or deterministic integration methods. Finally, the prediction output is mapped to operation and maintenance or control actions. At the same time, the prediction results trigger early warning and plan maintenance or seamless switching.

Images