Flexible direct current traction power supply method and system based on energy route cooperative control

Abstract

The application discloses a flexible DC traction power supply method and system based on energy route cooperative control and belongs to the technical field of power supply. The system is applied to a rail transit DC traction power grid and a 35kV medium-voltage looped network and is used as an extension application of an inverter feedback type regenerative energy absorption device and an over-capacity energy storage type regenerative energy absorption device. The system connects the DC traction power grid of the whole subway, cooperatively adjusts the dynamic demand of electric energy of each traction substation, adjusts relevant parameters of electric energy input equipment such as an inverter feedback device, an over-capacity energy storage device and photovoltaic equipment, guides the flow of electric energy between each traction substation, ensures that the electric energy input matches the demand of a train, rationally controls electric energy transmission and distribution, wastes as little electric energy as possible, reduces the demand for electric energy of a national power grid and saves the electric fee expenditure of a subway company.

Description

Technical Field

[0001] This invention belongs to the field of power supply technology, specifically a flexible DC traction power supply method and system based on energy routing coordinated control. Background Technology

[0002] In the operation of modern industrial and transportation equipment, hybrid renewable energy recovery technology is a key pillar for improving overall energy efficiency. Currently, most conventional energy recovery systems are limited to energy flow and storage within the equipment itself. When the equipment is operating at high intensity, this mechanism, which focuses solely on internal circulation, often leads to bottlenecks in energy processing, preventing the effective utilization of a large amount of valuable renewable energy.

[0003] This limitation directly leads to the problem of a mismatch between energy supply and demand. Under certain operating conditions, equipment can instantly generate regenerative energy far exceeding its own capacity limit, causing the internal energy storage space to quickly become saturated, leaving the excess energy with nowhere to go within the internal closed loop. Once the internal energy storage is full, this excess energy must find external channels for consumption to avoid being consumed as waste heat. This problem is particularly prominent in flexible DC traction power supply scenarios such as heavy-duty mining rail transport and heavy-haul railways. For example, when a heavy mining truck travels down a long downhill section of more than ten kilometers, the onboard battery is already fully charged at the beginning of the descent. Due to the lack of an efficient mobile energy routing and interaction mechanism, the continuously generated braking regenerative energy at the source cannot be flexibly injected into the flexible DC traction power supply network for grid connection and consumption, nor can it be precisely and flexibly routed and coordinated with other power consumption nodes along the line. Ultimately, it can only be converted into heat energy through braking resistors at the vehicle end or on the ground and wasted, or even cause serious safety hazards such as equipment overheating and DC bus overvoltage. Summary of the Invention

[0004] To address the above problems, this invention provides a flexible DC traction power supply method and system based on energy routing collaborative control. This system solves the problem of accurately controlling the routing power and timing of active power when the traction source and multiple nodes in the flexible DC traction power supply network are dynamically coordinated and flexibly connected to the grid for absorption, based on the energy routing mechanism, when the energy storage is saturated.

[0005] To achieve the above objectives, the present invention includes a flexible DC traction power supply method based on energy routing coordinated control, wherein the control method includes: Obtain the rotational speed data of the sliding window slice to predict the regeneration energy, obtain the maximum predicted energy value, and obtain the remaining energy value based on the maximum predicted energy value; The velocity range is obtained, and the velocity range and the remaining energy value are weighted and aggregated according to the random forest regression algorithm to obtain the initial outward transmission power of the remaining energy and the confidence interval of the remaining energy; The initial outward output power is constrained by the upper limit of the confidence interval according to the support vector regression algorithm, and after compensation, the target outward output power that meets the requirements of flexible DC grid connection is obtained. Based on the supply and demand matching results between the target outward output power and the real-time load data of each node in the flexible DC traction power supply network, when the matching is successful, the topology node detection and connectivity verification of the multi-energy routing node are triggered to determine the target DC access point. Based on the target DC access point, the current initial access point is retained under safe conditions or a flexible switching control of multiple energy routing nodes is executed to the secondary access point under overload conditions, thereby establishing the target access point and generating node configuration instructions; Based on the final target access point, the output voltage is evaluated for voltage attenuation and dynamically boosted to determine the boosting compensation voltage. Based on the harmonic suppression state of the boosting compensation voltage, waveform distortion filtering is performed to generate the final target transmission voltage that meets the grid connection standard of the flexible DC traction power supply network. Based on the control signal that the target transmission voltage meets the trigger threshold, the multi-energy router is driven to perform power routing action, inject the remaining energy value into the flexible DC traction power supply network, and perform closed-loop calculation of energy absorption by continuously sampling the operating current and phase parameters of the access point until the preset energy value is reached, thus completing the energy transfer and releasing the power upper limit constraint.

[0006] Preferably, the step of obtaining the sliding window slice rotation speed data to predict the regeneration energy, obtaining the maximum predicted energy value, and obtaining the remaining energy value based on the maximum predicted energy value includes: Acquire instantaneous rotational speed sampling data and internal energy storage upper limit; Based on the instantaneous rotational speed sampling data, a regenerative energy prediction sequence is obtained; The maximum predicted energy value is obtained based on the regenerative energy prediction sequence; The remaining energy value is obtained based on the difference between the maximum predicted energy value and the internal energy storage limit.

[0007] Preferably, the velocity range and the remaining energy value are weighted and aggregated using a random forest regression algorithm to obtain the initial outward transmission power of the remaining energy and the confidence interval of the remaining energy, including: Obtain the current energy value, current speed value, preset speed value, and preset power threshold; Based on the current speed value and the preset speed value, the speed range is obtained; Based on the stated speed range, a speed identifier is obtained; The sample feature quantity is obtained based on the velocity identifier; The sample features are input into the random forest model, and the path is traced by splitting the nodes of the decision trees. The energy decay ratio on the path is calculated to obtain the gradient value. The gradient values ​​are weighted and aggregated to calculate the initial outward output power. Determine whether the initial outward output power is greater than the preset power threshold; When the initial outward output power is greater than the preset power threshold, the corresponding confidence interval is determined based on the variance of the random forest model; When the initial outward output power is less than or equal to the preset power threshold, the prediction confidence level corresponding to the initial outward output power is the final confidence interval.

[0008] Preferably, the initial outward power output is constrained by the upper limit of the confidence interval using a support vector regression algorithm, and after compensation, the target outward power output that meets the requirements of flexible DC grid connection is obtained, including: The upper limit of the output power is obtained based on the confidence interval; Determine whether the initial outward output power is greater than the upper limit value of the output power; When it is determined that the initial outward output power is greater than the upper limit of the output power, the initial outward output power is truncated according to the upper limit of the output power, and a constraint amount is obtained; When it is determined that the initial outward output power is less than or equal to the upper limit of the output power, a constraint amount is output according to the initial outward output power; Obtain the load of the access node in the flexible DC traction power supply network, the system's preset maximum threshold for acceptance, and the grid transmission loss rate; The difference value is determined based on the load amount and the constraint amount; Determine whether the difference value is greater than the maximum threshold preset by the system; When the difference value is greater than the maximum threshold value preset by the system, an equivalent reduction amount is generated based on the excess amount, and the reduction amount is deducted from the constraint amount to determine the correction value of the external power grid absorption capacity. When the difference is less than or equal to the maximum threshold value preset by the system, a correction value for the external power grid's absorption capacity is determined based on the excess amount. The target outward output power is obtained by inversely amplifying and compensating the correction value based on the power grid transmission loss rate.

[0009] Preferably, based on the supply and demand matching results between the target outward output power and the real-time load data of each node in the flexible DC traction power supply network, when the matching is successful, topology node detection and connectivity verification of the multi-energy routing nodes are triggered to determine the target DC access point, including: Obtain the current real-time load value of the multi-energy routing node, and obtain the first load gap value based on the target outward output power and the real-time load. The first acceptance window corresponding to the first load gap value is determined based on the support vector machine model; Obtain the preset power output range; Determine whether the first receiving window is within the preset power output range; When the first receiving window is within the preset power output range, the supply and demand matching status is determined to be a successful matching status; Obtain the initial topology of the surrounding microgrid, and locate the initial topology to the first DC access point according to the breadth-first search algorithm; Obtain the communication address of the first DC access point, send a probe command to the communication address, and obtain the first probe response data; Extract the node connection identifiers from the first detection response data; Determine whether the node connection identifier is in a connected state; When the node connection identifier is in a connected state, the first DC access point is determined to be the target DC access point; When the node connection identifier is in a disconnected state, the initial topology is traversed to obtain the second DC access point, and a probe request is sent to the communication address of the second DC access point until a target DC access point with a connected node identifier is obtained. When the first receiving window exceeds the preset power output range, the supply and demand matching status is determined to be a matching failure status, the current power output is terminated, and the system waits for the next sampling period.

[0010] Preferably, based on the target DC access point, the current initial access point is retained under safe conditions, or a flexible switching control of multiple energy routing nodes is performed to the secondary access point under overload conditions, thereby establishing the target access point and generating node configuration instructions, including: Obtain the node number of the initial access point and the predicted current carrying threshold, and extract the resistance value, reactance value and location parameters based on the node number of the initial access point; Based on the target output power, the predicted current value is obtained using the support vector regression algorithm; The current difference is obtained based on the predicted current value and the preset current carrying threshold. Determine whether the current difference is within a preset safe range; When it is determined that the current difference is within the preset safe range, the current initial access point is retained as the final target access point; When it is determined that the current difference is not within the preset safe range, the backup secondary point is extracted as the final target access point; The location data is obtained based on the final target access point; Node configuration instructions are generated based on the location values.

[0011] Preferably, the output voltage is assessed for voltage attenuation and dynamically boosted based on the final target access point to determine the boost compensation voltage, and waveform distortion filtering is performed based on the harmonic suppression state of the boost compensation voltage, thereby generating the final target transmission voltage that meets the grid connection standard of the flexible DC traction power supply network, including: Obtain the preset attenuation threshold, and use the support vector regression algorithm to obtain the predicted voltage drop rate and transmission loss reduction rate based on the final target access point; The initial attenuation value of the output voltage is obtained based on the voltage drop rate and the transmission loss reduction rate. Determine whether the initial attenuation value is greater than the preset attenuation threshold; When the initial attenuation value is greater than the preset attenuation threshold, the phase offset angle of the output voltage is obtained based on the output voltage; The phase shift angle is smoothed using the Kalman filter algorithm to obtain a smoothed translation angle. The voltage rise is obtained based on the smooth translation angle. The voltage boost is superimposed on the output voltage to obtain the boost compensation voltage.

[0012] Preferably, the output voltage is assessed for voltage attenuation and dynamically boosted based on the final target access point to determine the boosting compensation voltage, and waveform distortion filtering is performed based on the harmonic suppression state of the boosting compensation voltage, thereby generating the final target transmission voltage that meets the grid connection standard of the flexible DC traction power supply network, and further includes: When the initial attenuation value is less than or equal to the preset attenuation threshold, the current output voltage is used as the boost compensation voltage; Obtain the harmonic suppression rate and preset suppression threshold of the raised compensation voltage; Determine whether the harmonic suppression rate of the raised compensation voltage is greater than a preset suppression threshold; When it is determined that the harmonic suppression rate of the boost compensation voltage is greater than or equal to the preset suppression threshold, the current boost compensation voltage is taken as the target transmission voltage. When the harmonic suppression rate of the boost compensation voltage is determined to be less than the preset suppression threshold, the waveform distortion value of the boost compensation voltage is calculated, and the distortion rate of the boost compensation voltage is divided according to the waveform distortion value to obtain the final target transmission voltage.

[0013] Preferably, based on a control signal indicating that the target transmission voltage meets a trigger threshold, a multi-energy router is driven to perform power routing, injecting the remaining energy value into the flexible DC traction power supply network. Energy absorption closed-loop calculations are then performed by continuously sampling the operating current and phase parameters of the access point until a preset energy value is reached, completing the energy transfer and releasing the power upper limit constraint. This includes: The target transmission voltage is classified according to the support vector machine algorithm to obtain the corresponding initial control signal; Determine whether the initial control signal is at a preset trigger threshold; When it is determined that the initial control signal is within a preset trigger threshold, the node configuration command is driven to output the system to perform flexible grid-connected energy routing interaction; The location of the current access point is determined based on the output of the system. The actual current and phase offset angle are obtained based on the location of the access point; The current absorption rate is obtained based on the actual current and phase offset angle. Determine whether the current absorption capacity is within the preset energy value; When the current absorption amount is at the preset energy value, it is determined that the flexible DC traction power supply network has completed the absorption of the remaining energy and the power upper limit constraint is released; When the current absorption amount is not at the preset energy value, the original constraint value is maintained, the current control command is stored in the sample pool to be verified, and the redundancy absorption query of adjacent nodes is calculated. When the initial control signal is not at the preset trigger threshold, the issuance of the current node configuration command is blocked, a warning is triggered, and the system waits for reclassification in the next cycle.

[0014] On the other hand, the present invention also includes a flexible DC traction power supply system based on energy routing cooperative control, wherein the control system includes a processor for executing any of the flexible DC traction power supply methods based on energy routing cooperative control as described above.

[0015] Through the above technical solutions, this invention achieves accurate prediction of regenerative energy, initial power planning, and safety constraints on the upper limit of the confidence interval by combining sliding window slice data with dual machine learning algorithms of random forest and support vector machine, effectively overcoming the randomness and volatility of distributed energy sources. After completing supply and demand matching with real-time loads, the system dynamically detects the topology of surrounding microgrids and verifies connectivity, supporting adaptive switching to secondary access points under overload conditions, significantly improving the flexible scheduling and overload resistance of microgrid groups. Simultaneously, addressing transmission path losses and grid pollution issues, this invention utilizes dynamic voltage rise compensation and harmonic distortion filtering technologies to significantly optimize grid-connected power transmission quality, ensuring efficient and safe power transmission. Finally, by continuously sampling the operating current and phase parameters to perform closed-loop calculations of energy absorption, the hardware is driven to accurately execute energy transfer. After the target is met, the power upper limit constraint is automatically released, and finally, an energy absorption control signal is generated to drive the power switch to perform outward output action, realizing the accurate transfer of remaining energy to the microgrid. This restores the internal energy storage of the equipment from a saturated state to its initial state, effectively avoiding energy waste and improving the overall energy utilization efficiency. This achieves efficient active power mutual assistance and flexible grid connection absorption between the traction source and the flexible DC traction network. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of one implementation of a flexible DC traction power supply method based on energy routing cooperative control; Figure 2 This is a schematic diagram of one implementation of a flexible DC traction power supply method based on energy routing cooperative control; Figure 3 This is a schematic diagram of one implementation of a flexible DC traction power supply method based on energy routing cooperative control; Figure 4 This is a schematic diagram of one implementation of a flexible DC traction power supply method based on energy routing cooperative control; Figure 5 This is a schematic diagram of one implementation of a flexible DC traction power supply method based on energy routing cooperative control; Figure 6 This is a schematic diagram of one implementation of a flexible DC traction power supply method based on energy routing cooperative control; Figure 7 This is a schematic diagram of one implementation of a flexible DC traction power supply method based on energy routing cooperative control; Figure 8 This is a schematic diagram of one implementation of a flexible DC traction power supply method based on energy routing cooperative control. Detailed Implementation

[0017] To enable those skilled in the art to better understand the technical solution, the present invention will be described in detail below with reference to embodiments. The description in this part is only exemplary and explanatory, and should not be used to limit the scope of protection of the present invention in any way.

[0018] like Figure 1 As shown, the Figure 1 This is a schematic diagram of a flexible DC traction power supply method based on energy routing cooperative control. Figure 1 It can be seen that the control method includes the following steps: S01. Obtain the rotation speed data of the sliding window slice to predict the regeneration energy, obtain the maximum predicted energy value, and obtain the remaining energy value based on the maximum predicted energy value.

[0019] S02. Obtain the velocity range. The velocity range and the remaining energy value are weighted and aggregated according to the random forest regression algorithm to obtain the initial outward transmission power of the remaining energy and the confidence interval of the remaining energy.

[0020] S03. The initial outward power is constrained by the upper limit of the confidence interval according to the support vector regression algorithm. After compensation, the target outward power that meets the requirements of flexible DC grid connection is obtained.

[0021] S04. Based on the supply and demand matching results between the target output power and the real-time load data of each node in the flexible DC traction power supply network, when the matching is successful, the topology node detection and connectivity verification of the multi-energy routing node are triggered to determine the target DC access point.

[0022] S05. Based on the target DC access point, retain the current initial access point under safe conditions or perform flexible switching control of multiple energy routing nodes to the secondary access point under overload conditions, thereby establishing the target access point and generating node configuration instructions.

[0023] S06. Based on the final target access point, the output voltage is evaluated for voltage attenuation and dynamically boosted to determine the boosted compensation voltage. Based on the harmonic suppression state of the boosted compensation voltage, waveform distortion filtering is performed to generate the final target transmission voltage that meets the grid connection standard of the flexible DC traction power supply network.

[0024] S07. Based on the control signal that the target transmission voltage meets the trigger threshold, drive the multi-energy router to perform power routing action, inject the remaining energy value into the flexible DC traction power supply network, and perform closed-loop calculation of energy absorption by continuously sampling the operating current and phase parameters of the access point until the preset energy value is reached to complete the energy transfer and release the power upper limit constraint.

[0025] This invention utilizes a sliding window slicing data approach combined with dual machine learning algorithms—random forest and support vector machine—to achieve accurate prediction of regenerative energy, initial power planning, and safety constraints on the upper limit of the confidence interval, effectively overcoming the randomness and volatility of distributed energy resources. After matching supply and demand with real-time loads, the system dynamically detects the topology of surrounding microgrids and verifies connectivity, supporting adaptive switching to secondary access points under overload conditions, significantly improving the flexible scheduling and overload resistance of microgrid clusters. Simultaneously, addressing transmission path losses and grid pollution issues, this invention employs dynamic voltage rise compensation and harmonic distortion filtering technologies to significantly optimize grid-connected power transmission quality, ensuring efficient and safe power transmission. Finally, by continuously sampling the operating current and phase parameters to perform closed-loop calculations of energy absorption, the hardware is driven to accurately execute energy transfer. After the target is met, the power upper limit constraint is automatically released, and finally, an energy absorption control signal is generated to drive the power switch to perform outward output action, realizing the accurate transfer of remaining energy to the microgrid. This restores the internal energy storage of the equipment from a saturated state to its initial state, effectively avoiding energy waste and improving the overall energy utilization efficiency. This achieves efficient active power mutual assistance and flexible grid connection absorption between the traction source and the flexible DC traction network.

[0026] In this embodiment, the methods for obtaining the sliding window slice rotation speed data to predict the regeneration energy, obtaining the maximum predicted energy value, and obtaining the remaining energy value based on the maximum predicted energy value can be various methods known to those skilled in the art. In this invention, such as... Figure 2 As shown, the process of obtaining the regeneration energy from the sliding window slice rotation speed data, obtaining the maximum predicted energy value, and then obtaining the remaining energy value based on the maximum predicted energy value includes the following steps: S011. Obtain instantaneous rotational speed sampling data and internal energy storage limit. There are various methods known in the art for obtaining instantaneous rotational speed sampling data. In one embodiment of this invention, the instantaneous rotational speed of the equipment under high-intensity operating conditions is obtained by acquiring the instantaneous rotational speed through sensors to obtain initial sampling data. The initial sampling data is then denoised using a long short-term memory network model to obtain the instantaneous rotational speed sampling data. Specifically, the system uses a high-frequency photoelectric encoder installed on the main shaft of the heavy-duty mining hoist to collect instantaneous rotational speed data in real time under full-load lowering conditions at a sampling frequency of 100 Hz, for example, obtaining a continuous fluctuation value of 1500 to 1800 revolutions per minute. Simultaneously, the system reads the current internal energy storage capacity limit of the supercapacitor bank as 25,000 kilojoules through the communication bus.

[0027] S012. Obtain the regenerative energy prediction sequence based on the instantaneous rotational speed sampling data. There are various methods known in the art for obtaining the regenerative energy prediction sequence. In one embodiment of this invention, a preset window length is first obtained. The instantaneous rotational speed sampling data is then time-sliced ​​using the window length to obtain multiple segments. The regenerative energy corresponding to each of these segments is then calculated, finally yielding the regenerative energy prediction sequence. Specifically, the system is set to a sliding window with a length of 2 seconds and a sliding step size of 0.5 seconds. The continuously flowing rotational speed sampling data is time-sliced, and the sliced ​​data is input into a pre-trained long short-term memory neural network model. Kinetic energy conversion integration is performed using rotational inertia parameters to generate a regenerative energy prediction sequence for the next ten seconds, for example, a set containing values ​​such as 24,000 kJ, 26,500 kJ, and 27,000 kJ.

[0028] S013. Obtain the maximum predicted energy value based on the regenerative energy prediction sequence. Specifically, the system uses an extreme value search algorithm to traverse the regenerative energy prediction sequence and extracts the maximum predicted value (27,000 kJ) by comparing each value. This maximum predicted energy value is then taken as the maximum predicted energy value.

[0029] S014. Obtain the remaining energy value based on the difference between the maximum predicted energy value and the internal energy storage limit. Various methods known to those skilled in the art can be used to obtain the remaining energy value. In one embodiment of the present invention, the internal energy storage limit is first obtained. If the maximum predicted energy value is greater than the internal energy storage limit, the difference between the maximum predicted energy value and the internal energy storage limit is calculated. This difference is then used to determine if it exceeds a preset overflow threshold. If the difference exceeds the overflow threshold, the difference is truncated to obtain the remaining energy value. Specifically, the control unit finally performs a threshold determination between the maximum predicted energy value and the internal energy storage limit of 25,000 kJ. After confirming that 27,000 kJ is greater than the internal energy storage limit, it automatically performs a subtraction operation to calculate the difference between the two, accurately calculating the remaining energy value of 27,000 kJ. A control command is immediately generated to guide this excess energy to an external braking resistor network for heat dissipation to prevent DC bus overvoltage.

[0030] In this embodiment, the method of obtaining the initial outward transmission power of the remaining energy and the confidence interval of the remaining energy by weighted aggregation of the velocity range and the remaining energy value using a random forest regression algorithm can be any of the methods known to those skilled in the art. In this invention, such as... Figure 3 As shown, the specific steps include the following: S021. Obtain the current energy value, current speed value, preset speed value, and preset power threshold. That is, the controller collects the system's current remaining energy value in real time. and current driving speed value Simultaneously, a pre-calibrated set of preset speed values ​​is read from memory. and preset power threshold This preset power threshold is used to determine whether the confidence interval of the prediction result needs to be adjusted. The method for obtaining the preset power threshold is as follows. This can be done in various ways known to those skilled in the art. In one embodiment of the present invention, during the development phase of a vehicle or energy storage system, historical operating data is collected under various typical operating conditions (such as urban road conditions, highway conditions, and hill climbing conditions). This data includes remaining energy value, vehicle speed, actual outward output power, and corresponding system stability indicators (such as whether power limiting, over-discharge warning, or thermal protection occurs). The collected historical actual outward output power is statistically analyzed to calculate its probability distribution or cumulative distribution function (CDF). Quantities that meet a preset coverage rate (such as 85%, 90%, or 95%) are selected as preset power thresholds. This ensures that the initial outward output power under most normal operating conditions is below or equal to this threshold, while operating conditions exceeding this threshold are considered high-power output conditions. Mathematically, this is expressed as:

[0031] in, It is the inverse function of the cumulative distribution function of historical actual outward power. This is a preset coverage rate. Specifically, it involves collecting the sequence of the maximum power output that the battery is actually allowed to deliver during normal driving / discharging. After sorting, take The value is .in The speed interval is the natural sequence number. S022. Based on the current speed value and the preset speed value, a speed interval is obtained. That is, the current speed value is compared and judged according to the preset speed value division rule to obtain the speed interval. There are various methods known to those skilled in the art for obtaining the speed interval. In one embodiment of the invention, the current speed value is compared and judged according to the preset speed value division rule to obtain the speed interval. Specifically, the current speed value... The speed range to which the value belongs is determined by comparing it with various thresholds in the preset speed value set. S023. Obtain the speed identifier based on the speed interval. Specifically, extract the corresponding speed identifier based on the speed interval. Assign a unique identifier to each speed interval. For example, integer encoding can be used: interval 1 corresponds to identifier 1, interval 2 corresponds to identifier 2, and so on. This identifier serves as one of the category features input to the subsequent model.

[0032] S024. Obtain sample feature quantities based on the velocity identifier. This involves concatenating the current energy value with the velocity identifier to obtain the sample feature quantities. Specifically, the velocity identifier obtained in step S023... Compared with the current remaining energy value obtained in step S021 The vectors are concatenated to form a feature vector. , eigenvector For details, please refer to formula (2). , (2).

[0033] S025. Input the sample features into the random forest model, determine the path to follow by splitting the nodes of the decision trees, and calculate the energy decay ratio on the path to obtain the gradient value.

[0034] There are various methods known to those skilled in the art for obtaining gradient values. In one embodiment of the present invention, the sample feature values ​​are first... The input is fed into a random forest model, which includes an N-tree regression decision tree. Secondly, the random forest algorithm uses a multi-tree decision tree to determine the node conditions of the sample features, thus obtaining the corresponding split points. Specifically, for each decision tree... Starting from the root node, based on the feature quantity The values ​​of each dimension are compared with the node splitting threshold to determine whether to enter the left or right child node, and this process is repeated recursively until a leaf node is reached. Then, based on the splitting point, the direction of the sample feature values ​​in the decision tree branches is tracked, and the energy decay ratio along the path is calculated. This energy decay ratio reflects the degree of energy loss caused by factors such as load changes and efficiency losses under the current operating state. Finally, the corresponding gradient value is obtained based on the energy decay ratio along the path. The gradient value is the first... The single-tree prediction of the power output of the remaining energy in the current state by each decision tree. Specifically, it is represented by formula (3). (3) S026. The gradient values ​​are weighted and aggregated to calculate the initial outward output power. That is, the gradient values ​​generated by the multi-tree decision tree are weighted and summed to determine the overall aggregation degree. The energy transfer rate of the device is calculated based on the aggregation degree to obtain the initial outward output power. There are many ways to weight and aggregate the gradient values, which are known to those skilled in the art. In one embodiment of the present invention, the equal-weighted average method is used for weighted aggregation, specifically represented by formula (4). (4) in, For the initial outward output power, This represents the total number of decision trees in the random forest model. For the first The gradient values ​​output by each decision tree.

[0035] S027. Determine whether the initial outward output power is greater than the preset power threshold. That is, compare the initial outward output power obtained in step S026. Compared with the preset power threshold obtained in step S021 .

[0036] S028. When the initial outward output power is greater than a preset power threshold, the corresponding confidence interval is determined based on the variance of the random forest model. That is, when... At this time, the system is considered to be operating under high load, and the prediction results of each decision tree may have large dispersion, resulting in high uncertainty in the prediction. Therefore, the statistical variance of the predicted values ​​of the random forest model is used to determine the confidence interval. The variance of the predicted values ​​of all decision trees is calculated according to formula (5). (5) Then, at a preset confidence level (e.g., 95%), the confidence interval is determined according to formula (6). (6) in, The quantiles corresponding to the pre-set confidence level for the standard normal distribution. The standard deviation is denoted as .

[0037] S029. When the initial outward output power is less than or equal to a preset power threshold, the prediction confidence level corresponding to the initial outward output power is the final confidence interval. That is, when... At this time, the system is considered to be in a normal or low-load operating state, and the prediction results of each decision tree are relatively consistent, with low uncertainty in model prediction. Therefore, the prediction confidence level corresponding to the initial outward output power is directly used as the final confidence interval. The prediction confidence level can be obtained through the cross-validation accuracy or calibration curve during model training. At this time, the confidence interval can be expressed according to formula (7). (7) in, A fixed offset, determined based on historical prediction deviation statistics, reflects the typical prediction error range of the model in low-power output scenarios.

[0038] In this embodiment, the method of constraining the upper limit of the confidence interval of the initial outward output power using a support vector regression algorithm, and obtaining the target outward output power that meets the requirements of flexible DC grid connection after compensation, can be various methods known to those skilled in the art. In this invention, such as... Figure 4 As shown, the specific steps include: S0301. Obtain the upper limit of output power based on the confidence interval. That is, based on the confidence interval determined in step S028 or S029. Extract the upper limit of the confidence interval as the upper limit of the output power. According to formula (8), (8) in, This is the upper boundary value of the confidence interval. This upper limit of output power represents the maximum power boundary that the system can reliably output at the current confidence level.

[0039] S0302. Determine whether the initial outward output power is greater than the upper limit of the output power. That is, compare the initial outward output power. With upper limit of output power .

[0040] S0303. When it is determined that the initial outward output power is greater than the upper limit of the output power, the initial outward output power is truncated according to the upper limit of the output power, and a constraint quantity is obtained. If the initial outward output power exceeds the reliable output range, posing an overload risk, a hard cutoff method is used to constrain the initial outward output power, i.e., the upper limit of the output power is used as the constraint. According to formula (9), , (9).

[0041] S0304. When it is determined that the initial outward output power is less than or equal to the upper limit of the output power, output a constraint amount based on the initial outward output power. At this point, it is assumed that the initial outward output power is within a reliable output range, and no truncation is required. Therefore, the initial outward output power is directly used as the constraint quantity. According to formula (10), (10) In one specific embodiment of the present invention, the system calculates the 95% prediction confidence interval of the power generation in the next hour based on the Gaussian process regression algorithm, extracts the corresponding preset power threshold of 2000 kilowatts for power fluctuation, and when the current initial outward output power is detected to reach 2300 kilowatts, the system automatically triggers the limiting mechanism to hard constrain the initial outward output power to 2000 kilowatts.

[0042] S0305. Obtain the load capacity of the access nodes within the surrounding flexible DC traction power supply network, the system's preset maximum load capacity threshold, and the grid transmission loss rate. That is, obtain the load capacity of the access nodes within the flexible DC traction power supply network. The system's preset maximum threshold for accepting [something]. and power grid transmission loss rate Load capacity of nodes connected to the flexible DC traction power supply network This indicates the current actual power load demand of the access node within the flexible DC traction power supply network; the system's preset maximum load capacity threshold. This indicates the maximum power that the surrounding microgrid and its associated energy storage equipment can safely accept; grid transmission loss rate. This indicates the proportion of power loss during power transmission, usually expressed as a percentage. In a specific embodiment of the present invention, after obtaining the system's constraint quantity, the system pulls real-time load data of three adjacent microgrids from the regional energy management platform through a dedicated fiber optic communication line. After analysis, the load of the surrounding microgrids is found to be 1500 kilowatts. Combined with the remaining 300 kilowatts of rechargeable capacity of its energy storage equipment, the system's preset maximum load capacity threshold is calculated to be 1800 kilowatts.

[0043] S0306. Determine the difference value based on the load and constraint values. The system uses a differential comparison algorithm to determine the constraint value. The load of the surrounding microgrid Perform the difference operation to obtain the difference value. According to formula (11), (10) This difference reflects the extent to which the constraint exceeds the absorption capacity of the surrounding microgrid. When When, it indicates that there is an excess portion that needs further processing; when When the constraint is within the absorption capacity of the surrounding microgrid, it indicates that the constraint is within the absorption capacity of the surrounding microgrid.

[0044] Specifically, the system uses a differential comparison algorithm to calculate the difference between the constrained 2000 kW output power and the system's preset maximum load threshold of 1800 kW, accurately identifying a difference of 200 kW.

[0045] S0307. Determine whether the difference value is greater than the system's preset maximum threshold. That is, determine the difference value. Compared with the system's preset maximum threshold Size.

[0046] S0308. When the difference value exceeds the system's preset maximum acceptance threshold, an equivalent reduction amount is generated based on the excess, and the reduction amount is deducted from the constraint amount to determine the correction value for the external power grid's absorption capacity. That is, when... At this point, the excess is considered very significant and requires substantial reduction. Therefore, an equivalent reduction amount is generated based on the excess. The reduction amount is then subtracted from the constraint amount to obtain the correction value. That is, the correction value is obtained according to formula (11). , (11) That is, the correction value is equal to the system's preset maximum threshold for acceptance.

[0047] S0309. When the difference is less than or equal to the system's preset maximum threshold, a correction value for the external power grid's absorption capacity is determined based on the excess amount. That is, when... At that time, it was considered that the excess was within a controllable range. At this point, the correction value... According to formula (12), (12) Since the difference is small, the system can absorb the excess through flexible scheduling methods (such as local consumption, energy storage charging and discharging regulation, etc.).

[0048] S0310. The target outward output power is obtained by inversely amplifying and compensating the correction value based on the grid transmission loss rate. Considering the line losses during power transmission, the correction value needs to be inversely amplified and compensated to ensure that the actual power delivered to the surrounding microgrids meets the demand. Target outward output power According to formula (13), (13) in, This represents the power grid transmission loss rate.

[0049] In one specific embodiment of the present invention, for the 200 kW excess power component, the system calls a dynamic programming algorithm to reduce and correct it, automatically sends a 100 kW local consumption command to the local high-energy-consuming temperature control equipment, and simultaneously controls the inverter to perform a power derating operation on the remaining 100 kW. After the double correction, the system accurately locks the final power transmitted to the outside at 1800 kW and uses it as the target power output to the outside for grid dispatch.

[0050] In one embodiment of the present invention, the method for determining the target DC access point by triggering topology node detection and connectivity verification of the surrounding microgrid upon successful matching of the target outward output power and real-time load data can be varied and can be known to those skilled in the art, such as... Figure 5 As shown, the specific steps in this invention are as follows: S0401. Obtain the current real-time load value at the multi-energy routing node, and determine the first load gap value based on the target outward output power and real-time load. The system obtains the current real-time load value. And, the total maximum available power of local renewable energy power generation and energy storage systems. The first load gap value is calculated according to formula (14). , (14) This first load gap value reflects the shortfall where local power supply capacity cannot meet local load demand, which needs to be made up by power input from an external microgrid.

[0051] In one specific embodiment of the present invention, the microgrid energy management system collects local grid operation data through a sensor network every 5 seconds. The current measured local real-time load data is 3500 kilowatts, while the maximum available power of the local new energy power generation and energy storage system is 2500 kilowatts. The system calculates the difference to determine that there is a first load gap of 1000 kilowatts.

[0052] S0402. Determine the first acceptance window corresponding to the first load gap value based on the support vector machine model. (The first load gap value is then...) The data is input into a pre-trained support vector machine model, which, based on historical load data and power allocation records, outputs a first acceptance window that matches the current load gap. This receiving window is a power range. This indicates the range of candidate external power supplies that can effectively fill the current load gap.

[0053] In one specific embodiment of the present invention, the system receives a target outward output power of 850 kilowatts from the adjacent microgrid No. 1 via the cross-network data bus. Subsequently, the system starts the supply and demand matching calculation program and uses a dynamic threshold determination algorithm to generate the first acceptance window.

[0054] S0403. Obtain the preset power output range. In this invention, the method for obtaining the preset power output range can be any method known to those skilled in the art. In a specific embodiment of this invention, the lower limit of the preset power output range is 80% of the first load gap value, and the upper limit is 120% of the first load gap value. The preset power output range is obtained according to formula (15). (15) The preset power output range is used to determine whether the power provided by the external microgrid is within a reasonable matching range.

[0055] Specifically, the system then initiates a supply and demand matching calculation program, uses a dynamic threshold determination algorithm to generate a receiving window, sets the lower limit of the receiving window to 80% of the first load gap value, i.e., 800 kilowatts, and sets the upper limit to 120% of the first load gap value, i.e., 1200 kilowatts.

[0056] S0404. Determine whether the first receiving window is within the preset power output range. That is, compare the first receiving window... With the preset power output range The inclusion relationship. Specifically, determine the first receiving window. lower limit Is it greater than or equal to the lower limit of the set power output range? And the first receiving window upper limit Is it less than or equal to the upper limit of the preset power output range? S0405. When the first receiving window is within the preset power output range, the supply and demand matching status is determined to be a successful matching status. When the first receiving window... Completely falls within the preset power output range When the power provided by the external microgrid is deemed to be able to effectively match the local load gap, the supply and demand matching status is determined to be a successful matching status.

[0057] In a specific embodiment of the present invention, the system compares the 850 kW output power of the adjacent No. 1 microgrid with the receiving window value and finds that the 850 kW accurately falls within the range of 800 kW to 1200 kW. At this time, the system logic controller rewrites the supply and demand matching status flag from zero to one and officially determines that the supply and demand matching status is a successful matching status.

[0058] S0406. Obtain the initial topology of the surrounding microgrid and locate the first DC connection point using a breadth-first search algorithm. The system obtains the initial topology of the surrounding microgrid. This topology includes the physical connections and electrical parameters between the microgrid nodes. A breadth-first search algorithm is used to traverse the initial topology, starting from the local node and searching layer by layer for connected microgrid nodes until the first reachable DC connection point is located, denoted as the first DC connection point. .

[0059] S0407. Obtain the communication address of the first DC access point, send a probe command to the communication address, and obtain the first probe response data. The access point's communication address (such as IP address, MAC address, or dedicated communication protocol address) is used to send a probe command to that address. This probe command includes power scheduling requirements and connectivity test requests. Feedback data from this access point is received and recorded as the first probe response data. .

[0060] S0408. Extract the node connection identifiers from the first probe response data. That is, extract the node connection identifiers from the first probe response data. Extracting node connection identifiers from parsing This identifier is used to indicate whether the DC access point is currently in a connected state, and is usually represented by binary bits: 1 indicates connected, and 0 indicates disconnected.

[0061] S0409. Determine whether the node connection identifier is in a connected state.

[0062] S0410. When the node connection identifier is in a connected state, the first DC access point is determined as the target DC access point. That is, when... At that time, confirm the first DC access point. The current state is connected; therefore, it is identified as the target DC access point. .

[0063] S0411. When the node connection identifier is in a disconnected state, traverse the initial topology, obtain the second DC access point, and send a probe request to the communication address of the second DC access point until a target DC access point with a connected node identifier is obtained. That is, when... At this point, the first DC access point is considered unavailable. Therefore, the initial topology is then traversed again. Following the order of the breadth-first search algorithm, the next candidate DC access point is obtained and denoted as the second DC access point. Repeat steps S0407 to S0410, namely, obtain its communication address, send a probe command, extract the node connection identifier and make a judgment, until an access point with a node connection identifier in a connected state is found, and it is identified as the target DC access point. .

[0064] In one specific embodiment of the invention, the system's underlying communication module is automatically woken up and uses the shortest path first algorithm to send high-frequency probe request messages to four backup microgrid access nodes within a 10-kilometer physical distance. The messages encapsulate an 850-kilowatt power scheduling command and the current millisecond-level timestamp to test the communication delay and line impedance of each node, thereby establishing the optimal transmission route channel for subsequent physical closing and power transmission, and obtaining the final target DC access point.

[0065] S0412. When the first receiving window exceeds the preset power output range, the supply and demand matching status is determined to be a matching failure state, the current power output is terminated, and the system waits for the next sampling cycle. That is, when the first receiving window... Failed to fall within the preset power output range If, at this time, it is determined that there is no suitable external microgrid to effectively match the local load gap, the supply-demand matching status is identified as a matching failure, the system terminates the current power output operation, and waits for the next sampling cycle to re-execute the above steps.

[0066] In one embodiment of the present invention, the method for establishing the target access point and generating node configuration instructions, based on the target DC access point, can be one of various methods known to those skilled in the art, such as: (e.g., retaining the current initial access point under safe conditions or performing flexible switching control of multiple energy routing nodes to a secondary access point under overload conditions). Figure 6 As shown, the specific steps in the invention are as follows: S051. Obtain the node number of the initial access point and the predicted current carrying capacity threshold. Extract the resistance value, reactance value, and location parameters based on the node number of the initial access point. The system obtains the node number of the initial access point. Its corresponding predicted current carrying capacity threshold The predicted current carrying capacity threshold represents the maximum current value that the cable at this node can withstand under safe operating conditions.

[0067] S052. Based on the target output power, the predicted current value is obtained using the support vector regression algorithm. The system obtains the node number of the initial access point. Its corresponding predicted current carrying capacity threshold This predicted current carrying capacity threshold represents the maximum current value that the cable at this node can withstand under safe operating conditions.

[0068] In one specific embodiment of the invention, the system automatically reads the line resistance component of the target access node as 15 milliohms and the line reactance component as 8 milliohms from the power grid topology database, and simultaneously obtains the access location parameter of the node as 3500 meters from the main transformer. Subsequently, the system receives that the target outward output power of the node is 500,000 watts and the operating voltage is 10,000 volts. Using a complex power flow algorithm, it first calculates the total line impedance modulus as 17 milliohms. Combining this with the set power factor of 90%, and using the formula that the base current equals the power divided by the product of the voltage and the power factor, the base current is calculated to be 55 amperes. Then, the system introduces an impedance loss coefficient of 102% derived from the location parameters for correction, resulting in the expected current flowing through the node being 56 amperes.

[0069] S053. Obtain the current difference based on the predicted current value and the preset current carrying threshold. That is, the predicted current value. With predicted current carrying threshold The difference between them yields the current difference. The current difference is obtained according to formula (16). , (16) This current difference reflects the degree to which the current predicted current deviates from the safe carrying capacity. When the predicted current does not exceed the carrying capacity threshold, it indicates that the current is within the specified range. When the predicted current exceeds the carrying capacity threshold, it indicates a potential safety hazard.

[0070] S054. Determine if the current difference is within the preset safe range. Preset safe range , is a symmetric interval centered at zero, in which This is the allowable current deviation margin. Determine the current difference. Does it satisfy formula (17)? (17) If the above conditions are met, the current difference is considered to be within the preset safe range; otherwise, the current difference is considered to be outside the preset safe range.

[0071] S055. When the current difference is determined to be within a preset safe range, the current initial connection point is retained as the final target connection point. That is, when... When the current initial connection point is within the preset safety range, it is assumed that the electrical parameters meet the safety operation requirements. In this case, the current initial connection point is retained. As the final target access point .

[0072] S056. When the current difference is determined to be outside the preset safe range, the backup secondary point is selected as the final target connection point. That is, when... If the initial access point is not within the preset safe zone, it is considered to have an overload risk and cannot be used as the final target access point. In this case, a backup secondary access point is retrieved from the backup routing table. As the final target access point .

[0073] S057. Obtain the location data based on the final target access point. That is, based on the final target access point... Extract the location parameters of the access point from the power grid topology database. This includes its geographical coordinates, distance relative to the main transformer, and feeder number. This location information is used for subsequent node configuration command generation.

[0074] S058. Generate node configuration instructions based on the location quantity.

[0075] In one specific embodiment of the invention, after obtaining the expected current, the system automatically retrieves the preset current carrying threshold of the node cable as 60 amps and executes a comparison analysis program. Since the calculated expected current value of 56 amps is lower than the preset current carrying threshold of 60 amps, the system determines that the current line is in a safe operating range. It then locks and retains the current access node in the database. If the expected current value calculated under another high-load condition reaches 65 amps and exceeds the threshold, the system will immediately block the current connection command and automatically switch to a secondary access location 4200 meters away from the main transformer with a carrying threshold as high as 80 amps according to the backup routing table.

[0076] In one embodiment of the invention, the method for determining the boost compensation voltage by evaluating voltage attenuation and dynamically boosting the output voltage based on the final target access point, and performing waveform distortion filtering based on the harmonic suppression state of the boost compensation voltage, thereby generating the final target transmission voltage that meets the grid connection standard of the flexible DC traction power supply network, can be various methods known to those skilled in the art, such as... Figure 7 As shown, the present invention includes the following steps: S0601. Obtain the preset attenuation threshold, and use a support vector regression algorithm to obtain the predicted voltage drop rate and transmission loss reduction rate based on the final target access point. That is, the system takes the preset attenuation threshold. This threshold is used to determine whether the line voltage attenuation exceeds the allowable range. (Based on the final target access point) Extract the line parameters and real-time operating parameters of the node, such as equivalent impedance, from the power grid topology database. and real-time power transmission These parameters are used as input features and fed into a pre-trained support vector regression model to output the predicted voltage drop rate. and transmission loss reduction rate The voltage drop rate is obtained according to formulas (18) and (19). and transmission loss reduction rate , (18) (19) in, This is the reference voltage.

[0077] In one specific embodiment of the present invention, the system first extracts the retained access node parameters, wherein the node's equivalent impedance is 0.5 ohms and the real-time transmission current is 100 amperes. Using 10,000 volts as the reference voltage, the voltage drop factor is calculated to be 0.005 using Ohm's law and a ratio algorithm. At the same time, the transmission loss power is calculated to be 5,000 watts using the formula of current squared multiplied by impedance. This is then divided by the reference current to convert it to a transmission loss of 5 volts.

[0078] S0602. Obtain the initial attenuation value of the output voltage based on the voltage sag rate and the transmission loss reduction rate. That is, the voltage sag rate... Convert to the absolute value of voltage drop and compare with the transmission loss rate. By adding them together, the initial attenuation value of the output voltage can be obtained according to formula (20). , , (20).

[0079] S0603. Determine whether the initial attenuation value is greater than the preset attenuation threshold. That is, compare the initial attenuation value. With the preset attenuation threshold .

[0080] S0604. When the initial attenuation value is greater than the preset attenuation threshold, the phase offset angle of the output voltage is obtained based on the output voltage. That is, when... At this point, it is considered that the line voltage drop is too large, and phase compensation is required. At this time, based on the current output voltage... Calculate the phase offset angle caused by line parameters (especially line inductance and capacitance) and distributed line parameters. This angle reflects the voltage phase lag caused by reactive power flowing in the line. S0605, the phase shift angle is smoothed using the Kalman filter algorithm to obtain a smoothed translation angle. Due to measurement noise and system disturbances, the directly measured phase shift angle... There may be jitter. A Kalman filter algorithm is used to recursively smooth the angle, filtering out noise and obtaining the smoothed translation angle. .

[0081] S0606. The voltage rise is obtained based on the smooth translation angle. In order to offset the reactive voltage drop caused by the phase shift, the voltage rise needs to be calculated based on the smooth translation angle. The voltage rise is calculated according to formula (21). , ,(twenty one) S0607. The voltage rise is superimposed on the output voltage to obtain the voltage rise compensation voltage. The calculated voltage rise is... Compared with the original output voltage By superimposing these values, a preliminary lifting compensation pressure is obtained. The lifting compensation pressure is obtained according to formula (22). , ,(twenty two).

[0082] In one specific embodiment of the invention, the system detects that the distributed capacitance and inductance of the line cause a 15-degree phase shift angle. In order to eliminate the voltage attenuation caused by reactive power, the system adopts a vector compensation algorithm to divide the reference voltage by the cosine value of 15 degrees, 0.965, to obtain a theoretical boost value of 10362.69 volts. Then, it directly adds a loss conversion amount of 5 volts to complete the boost compensation of the output voltage. At this time, the boost compensation voltage reaches 10367.69 volts.

[0083] S0608. When the initial attenuation value is less than or equal to the preset attenuation threshold, the current output voltage is used as the boost compensation voltage. That is, when... At this point, it is assumed that the line voltage drop is within the allowable range and no phase compensation is needed. In this case, the current output voltage is directly... As a lifting compensation pressure .

[0084] S0609. Obtain the harmonic suppression rate and preset suppression threshold of the boost compensation voltage. That is, the system obtains the boost compensation voltage. Harmonic suppression rate and preset suppression threshold Harmonic suppression rate It reflects the severity of harmonic content in the voltage waveform, and is usually measured by the total harmonic distortion (THD). The preset suppression threshold is a value that is pre-calibrated and stored in the controller according to the voltage THD limit specified in the national power quality standard corresponding to the voltage level of the system access point.

[0085] S0610. Determine whether the harmonic suppression rate of the boosted compensation voltage is greater than the preset suppression threshold. That is, compare the harmonic suppression rates. With the preset suppression threshold The size relationship.

[0086] S0611. When the harmonic suppression rate of the boost compensation voltage is determined to be greater than or equal to the preset suppression threshold, the current boost compensation voltage is used as the target transmission voltage. That is, when... At this point, the harmonic content is considered to be within an acceptable range, and the voltage quality is deemed acceptable. At this time, the current boost compensation voltage is... Directly used as the final target transmission voltage .

[0087] S0612. When the harmonic suppression rate of the boost compensation voltage is determined to be less than the preset suppression threshold, the waveform distortion value of the boost compensation voltage is calculated. The waveform distortion value is then divided by the distortion rate of the boost compensation voltage to obtain the final target transmission voltage. That is, when... At that time, it was determined that the harmonic content exceeded the standard, requiring filtering. The waveform distortion value of the boost compensation voltage was calculated. The boost compensation voltage is corrected based on the waveform distortion value. One feasible correction method is to inject reverse harmonic current using an active filter, which is equivalent to performing a distortion rate division operation on the voltage mathematically to eliminate harmonic components. The pure target transmission voltage is then obtained according to formula (23). , ,(twenty three) In one specific embodiment of the invention, the waveform will produce nonlinear distortion during the voltage rise process. The system then starts the Fast Fourier Transform algorithm to perform frequency domain analysis on the waveform and identifies that the total amplitude of the third and fifth harmonic components is 150 volts. At this time, the system introduces a preset harmonic suppression ratio of 0.98 and reduces the harmonic amplitude to 3 volts through a digital filter. The pure fundamental voltage after filtering out the distorted components is then inversely transformed back into a time domain signal. Finally, the system accurately calculates and outputs the target transmission voltage of 10364.69 volts.

[0088] In one embodiment of the present invention, the method for driving a multi-energy router to perform power routing based on a control signal that satisfies a trigger threshold, injecting the remaining energy value into a flexible DC traction power supply network, and performing closed-loop calculations of energy absorption by continuously sampling the operating current and phase parameters of the access point until a preset energy value is reached to complete energy transfer and remove the power upper limit constraint can be one of various methods known to those skilled in the art, such as... Figure 8 As shown, the present invention specifically includes the following steps: S071. The target transmission voltage is classified according to the support vector machine algorithm to obtain the corresponding initial control signal. That is, the system classifies the target transmission voltage determined in step S06. The input features are fed into a pre-trained support vector machine classification model. This model maps the voltage amplitude range to the corresponding control category and outputs the initial control signal. This control signal is typically the duty cycle or switching frequency parameter of a pulse width modulation signal.

[0089] S072. Determine whether the initial control signal is within the preset trigger threshold. The system reads the preset trigger threshold range. Determine the initial control signal Does it satisfy formula (24)? ,(twenty four) If the inequality is true, the initial control signal is determined to be within the preset trigger threshold; otherwise, it is determined to be outside the preset trigger threshold.

[0090] S073. When the initial control signal is determined to be within the preset trigger threshold, the drive node configuration command outputs the system to perform flexible grid-connected energy routing interaction. When within the effective range, the system is allowed to perform power output. The initial control signal... The signal is transmitted via an opto-isolation circuit to the control terminal of the power switching device (such as the gate of an insulated gate bipolar transistor (IGBT) or a metal-oxide-semiconductor field-effect transistor (MOSFET)) to drive the hardware circuit.

[0091] In one specific embodiment of the present invention, after receiving the target transmission voltage parameter of 380.5 volts, the system uses a pulse width modulation algorithm to convert the voltage value into a high-frequency square wave pulse with a duty cycle of 65.5%, which serves as the energy absorption control signal. This control signal is directly transmitted to the control terminal of the insulated gate bipolar transistor array via an opto-isolation circuit, triggering the underlying hardware power switch to alternately perform on and off operations at a fixed frequency of 20,000 times per second, thereby distributively controlling the internal DC bus to release electrical energy to the external AC power grid, completing the external output action.

[0092] S074. Determine the location of the current access point based on the system output. During power output, the system determines the physical location where electrical energy is actually injected into the power grid, i.e., the current access point, based on the currently activated circuit topology. This location information corresponds to the target access point determined in step S05 and is used for subsequent electrical quantity acquisition and positioning.

[0093] S075. Obtain the actual current and phase offset angle based on the location of the access point. The system is installed at the access point. The sensor at the location collects electrical parameters in real time, obtaining the actual current. and phase offset angle .

[0094] In one specific embodiment of the present invention, during the entire cycle of continuous power output, the built-in Hall sensor and phase-locked loop module continuously sample the electrical status of the grid-connected node at a sampling frequency of once per millisecond, and obtain the current instantaneous actual current as 45.2 amperes. At the same time, the central processing unit uses the fast discrete Fourier transform algorithm to analyze the collected voltage and current waveforms and calculates the phase offset angle between them as 0.08 radians.

[0095] S076. The current absorption capacity is obtained based on the actual current and phase offset angle. That is, the system calculates the absorption capacity based on the actual current collected. and phase offset angle Calculate the active power that the system actually transmits to the outside and is absorbed by the grid at the current moment, and obtain the current absorption amount according to formula (25). , (25) S077. Determine if the current energy consumption is within the preset energy value. The system obtains the preset energy value. The current absorption capacity is determined according to formula (26). Does it meet the preset energy value? , (26) in, This represents the allowable error band. If satisfied, the current energy consumption is determined to be within the preset energy value; otherwise, it is determined not to be within the preset energy value.

[0096] S078. When the current energy absorption rate is at the preset energy value, it is determined that the flexible DC traction power supply network has completed the absorption of the remaining energy, and the power upper limit constraint is released. That is, when... If the accuracy requirements are met, the energy transfer task is considered successfully completed. The system outputs a "energy transfer complete" status flag and removes the hard constraint on the initial outward output power setting in step S03 (i.e., no longer limiting the power upper limit), restoring the system to standby or normal monitoring state.

[0097] S079. When the current absorption capacity is not at the preset energy value, maintain the original constraint value, store the current control command in the sample pool to be verified, and calculate the redundancy absorption query of adjacent nodes. If the accuracy requirements are not met, the system determines that the current output is abnormal. In this case, the original power upper limit constraint remains unchanged to prevent overload; the current control command is then... The actual output data and target voltage are packaged and stored in the sample pool to be verified for subsequent model iteration and optimization; a redundancy query mechanism is started to calculate and query whether there is idle capacity in adjacent nodes, and attempt to transfer the remaining energy to other available nodes to improve the overall absorption efficiency.

[0098] S0710. When the initial control signal is not at the preset trigger threshold, block the issuance of the current node configuration command and trigger a warning, waiting for reclassification in the next cycle. That is, when it is determined... If the signal is outside the valid range, the system immediately blocks the output of the control signal, prohibiting the operation of power devices such as IGBTs to prevent hardware damage due to abnormal control parameters. At the same time, the system triggers a software alarm (such as an indicator light flashing or sending an error code to the host computer) and suspends the current scheduling cycle, waiting for the next scheduling cycle to re-execute the classification process of step S071.

[0099] In one specific embodiment of the present invention, the edge computing unit substitutes the real-time current value and phase offset angle into the active power integral equation for continuous calculation. When the cumulative transferred energy reaches the preset energy value of 50.0 kilowatts, the control program immediately issues a lockout command to cut off the power switch and packages the dataset containing the action start time, average output current, dynamic phase deviation and final total transferred energy into the relational database, automatically generating an energy transfer completion record with a unique hash value.

[0100] In one specific embodiment of the present invention, the system automatically reads the historical transmission archive data of the distributed nodes through the log parsing script, extracts the actual transmitted power of the current operation as 500 watt-hours using a string matching algorithm, and then the control center calls the numerical comparison algorithm to calculate the difference between the actual transmitted power and the energy index to be processed updated in real time in memory. When the system finds that the energy index to be processed read is also 500 watt-hours and the difference between the two is 0, it triggers the state machine transition logic, determines that the underlying hardware has successfully released excess charge, and the flag bit in its status register is automatically overwritten from code 2 representing full load to code 0 representing the baseline state, indicating that the battery capacity has accurately dropped from the 100% full charge threshold to the 20% safe baseline. Then, the energy management system sends a parameter reset command to the control module through the internal bus, directly modifying the maximum threshold parameter of the charge and discharge rate from 50 watts in the limited state to 200 watts in the rated state, completely releasing the current limiting protection mechanism and allowing the hardware to resume full-speed bidirectional inverter operation.

[0101] Furthermore, the present invention also includes a flexible DC traction power supply system based on energy routing cooperative control, the system including a processor for executing any of the above-described flexible DC traction power supply methods based on energy routing cooperative control.

[0102] Through the above technical solutions, this invention achieves accurate prediction of regenerative energy, initial power planning, and safety constraints on the upper limit of the confidence interval by combining sliding window slice data with dual machine learning algorithms of random forest and support vector machine, effectively overcoming the randomness and volatility of distributed energy sources. After completing supply and demand matching with real-time loads, the system dynamically detects the topology of surrounding microgrids and verifies connectivity, supporting adaptive switching to secondary access points under overload conditions, significantly improving the flexible scheduling and overload resistance of microgrid groups. Simultaneously, addressing transmission path losses and grid pollution issues, this invention utilizes dynamic voltage rise compensation and harmonic distortion filtering technologies to significantly optimize grid-connected power transmission quality, ensuring efficient and safe power transmission. Finally, by continuously sampling the operating current and phase parameters to perform closed-loop calculations of energy absorption, the hardware is driven to accurately execute energy transfer. After the target is met, the power upper limit constraint is automatically released, and finally, an energy absorption control signal is generated to drive the power switch to perform an outward output action, realizing the accurate transfer of surplus energy to the microgrid. This restores the internal energy storage of the equipment from a saturated state to its initial state, effectively avoiding energy waste and improving the overall energy utilization efficiency. It fundamentally solves the core pain points of difficult renewable energy absorption, large grid connection impact, and high transmission loss, and realizes efficient active power mutual assistance and flexible grid connection absorption between the traction source end and the flexible DC traction network.

[0103] It should be noted that, in this document, the terms "comprising," "including," and any other variations are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Specific examples have been used in this document to illustrate the principles and implementation methods of the present invention. These examples are merely for the purpose of helping to understand the method and core ideas of the present invention. The above are only preferred embodiments of the present invention. It should be pointed out that, due to the limitations of written expression and the objective existence of infinite specific structures, those skilled in the art can make several improvements, modifications, or variations without departing from the principles of the present invention, and can also combine the above technical features in an appropriate manner. These improvements, modifications, variations, or combinations, or the direct application of the concept and technical solution of the present invention to other situations without modification, should all be considered within the scope of protection of the present invention.

Claims

1. A flexible DC traction power supply method based on energy routing coordinated control, characterized in that, The control method includes: Obtain the rotational speed data of the sliding window slice to predict the regeneration energy, obtain the maximum predicted energy value, and obtain the remaining energy value based on the maximum predicted energy value; The velocity range is obtained, and the velocity range and the remaining energy value are weighted and aggregated according to the random forest regression algorithm to obtain the initial outward transmission power of the remaining energy and the confidence interval of the remaining energy; The initial outward output power is constrained by the upper limit of the confidence interval according to the support vector regression algorithm, and after compensation, the target outward output power that meets the requirements of flexible DC grid connection is obtained. Based on the supply and demand matching results between the target outward output power and the real-time load data of each node in the flexible DC traction power supply network, when the matching is successful, the topology node detection and connectivity verification of the multi-energy routing node are triggered to determine the target DC access point. Based on the target DC access point, the current initial access point is retained under safe conditions or a flexible switching control of multiple energy routing nodes is executed to the secondary access point under overload conditions, thereby establishing the target access point and generating node configuration instructions; Based on the final target access point, the output voltage is evaluated for voltage attenuation and dynamically boosted to determine the boosting compensation voltage. Based on the harmonic suppression state of the boosting compensation voltage, waveform distortion filtering is performed to generate the final target transmission voltage that meets the grid connection standard of the flexible DC traction power supply network. Based on the control signal that the target transmission voltage meets the trigger threshold, the multi-energy router is driven to perform power routing action, inject the remaining energy value into the flexible DC traction power supply network, and perform closed-loop calculation of energy absorption by continuously sampling the operating current and phase parameters of the access point until the preset energy value is reached, thus completing the energy transfer and releasing the power upper limit constraint.

2. The control method according to claim 1, characterized in that, The process of obtaining the sliding window slice rotation speed data to predict the regeneration energy, obtaining the maximum predicted energy value, and obtaining the remaining energy value based on the maximum predicted energy value includes: Acquire instantaneous rotational speed sampling data and internal energy storage upper limit; Based on the instantaneous rotational speed sampling data, a regenerative energy prediction sequence is obtained; The maximum predicted energy value is obtained based on the regenerative energy prediction sequence; The remaining energy value is obtained based on the difference between the maximum predicted energy value and the internal energy storage limit.

3. The control method according to claim 1, characterized in that, The velocity range and the remaining energy value are weighted and aggregated using a random forest regression algorithm to obtain the initial outward transmission power of the remaining energy and the confidence interval of the remaining energy, including: Obtain the current energy value, current speed value, preset speed value, and preset power threshold; Based on the current speed value and the preset speed value, the speed range is obtained; Based on the stated speed range, a speed identifier is obtained; The sample feature quantity is obtained based on the velocity identifier; The sample features are input into the random forest model, and the path is traced by splitting the nodes of the decision trees. The energy decay ratio on the path is calculated to obtain the gradient value. The gradient values ​​are weighted and aggregated to calculate the initial outward output power. Determine whether the initial outward output power is greater than the preset power threshold; When the initial outward output power is greater than the preset power threshold, the corresponding confidence interval is determined based on the variance of the random forest model; When the initial outward output power is less than or equal to the preset power threshold, the prediction confidence level corresponding to the initial outward output power is the final confidence interval.

4. The control method according to claim 1, characterized in that, The initial outward power output is constrained by the upper limit of the confidence interval using a support vector regression algorithm. After compensation, the target outward power output that meets the requirements of flexible DC grid connection is obtained, including: The upper limit of the output power is obtained based on the confidence interval; Determine whether the initial outward output power is greater than the upper limit value of the output power; When it is determined that the initial outward output power is greater than the upper limit of the output power, the initial outward output power is truncated according to the upper limit of the output power, and a constraint amount is obtained; When it is determined that the initial outward output power is less than or equal to the upper limit of the output power, a constraint amount is output according to the initial outward output power; Obtain the load of the access node in the flexible DC traction power supply network, the system's preset maximum threshold for acceptance, and the grid transmission loss rate; The difference value is determined based on the load amount and the constraint amount; Determine whether the difference value is greater than the maximum threshold preset by the system; When the difference value is greater than the maximum threshold value preset by the system, an equivalent reduction amount is generated based on the excess amount, and the reduction amount is deducted from the constraint amount to determine the correction value of the external power grid absorption capacity. When the difference is less than or equal to the maximum threshold value preset by the system, a correction value for the external power grid's absorption capacity is determined based on the excess amount. The target outward output power is obtained by inversely amplifying and compensating the correction value based on the power grid transmission loss rate.

5. The control method according to claim 1, characterized in that, Based on the supply and demand matching results between the target outward output power and the real-time load data of each node in the flexible DC traction power supply network, when the matching is successful, topology node detection and connectivity verification of multi-energy routing nodes are triggered to determine the target DC access point, including: Obtain the current real-time load value of the multi-energy routing node, and obtain the first load gap value based on the target outward output power and the real-time load. The first acceptance window corresponding to the first load gap value is determined based on the support vector machine model; Obtain the preset power output range; Determine whether the first receiving window is within the preset power output range; When the first receiving window is within the preset power output range, the supply and demand matching status is determined to be a successful matching status; Obtain the initial topology of the surrounding microgrid, and locate the initial topology to the first DC access point according to the breadth-first search algorithm; Obtain the communication address of the first DC access point, send a probe command to the communication address, and obtain the first probe response data; Extract the node connection identifiers from the first detection response data; Determine whether the node connection identifier is in a connected state; When the node connection identifier is in a connected state, the first DC access point is determined to be the target DC access point; When the node connection identifier is in a disconnected state, the initial topology is traversed to obtain the second DC access point, and a probe request is sent to the communication address of the second DC access point until a target DC access point with a connected node identifier is obtained. When the first receiving window exceeds the preset power output range, the supply and demand matching status is determined to be a matching failure status, the current power output is terminated, and the system waits for the next sampling period.

6. The control method according to claim 1, characterized in that, Based on the target DC access point, the current initial access point is retained under safe conditions, or a flexible switching control of multiple energy routing nodes is executed to the secondary access point under overload conditions, thereby establishing the target access point and generating node configuration instructions, including: Obtain the node number of the initial access point and the predicted current carrying threshold, and extract the resistance value, reactance value and location parameters based on the node number of the initial access point; Based on the target output power, the predicted current value is obtained using the support vector regression algorithm; The current difference is obtained based on the predicted current value and the preset current carrying threshold. Determine whether the current difference is within a preset safe range; When it is determined that the current difference is within the preset safe range, the current initial access point is retained as the final target access point; When it is determined that the current difference is not within the preset safe range, the backup secondary point is extracted as the final target access point; The location data is obtained based on the final target access point; Node configuration instructions are generated based on the location values.

7. The control method according to claim 1, characterized in that, Based on the final target access point, the output voltage is evaluated for voltage attenuation and dynamically boosted to determine the boost compensation voltage. Then, based on the harmonic suppression state of the boost compensation voltage, waveform distortion filtering is performed to generate the final target transmission voltage that meets the grid connection standards of the flexible DC traction power supply network, including: Obtain the preset attenuation threshold, and use the support vector regression algorithm to obtain the predicted voltage drop rate and transmission loss reduction rate based on the final target access point; The initial attenuation value of the output voltage is obtained based on the voltage drop rate and the transmission loss reduction rate. Determine whether the initial attenuation value is greater than the preset attenuation threshold; When the initial attenuation value is greater than the preset attenuation threshold, the phase offset angle of the output voltage is obtained based on the output voltage; The phase shift angle is smoothed using the Kalman filter algorithm to obtain a smoothed translation angle. The voltage rise is obtained based on the smooth translation angle. The voltage boost is superimposed on the output voltage to obtain the boost compensation voltage.

8. The control method according to claim 7, characterized in that, Based on the final target access point, voltage attenuation assessment and dynamic boost compensation are performed on the output voltage to determine the boost compensation voltage. Based on the harmonic suppression state of the boost compensation voltage, waveform distortion filtering is performed to generate the final target transmission voltage that meets the grid connection standards of the flexible DC traction power supply network. The process also includes: When the initial attenuation value is less than or equal to the preset attenuation threshold, the current output voltage is used as the boost compensation voltage; Obtain the harmonic suppression rate and preset suppression threshold of the raised compensation voltage; Determine whether the harmonic suppression rate of the raised compensation voltage is greater than a preset suppression threshold; When it is determined that the harmonic suppression rate of the boost compensation voltage is greater than or equal to the preset suppression threshold, the current boost compensation voltage is taken as the target transmission voltage. When the harmonic suppression rate of the boost compensation voltage is determined to be less than the preset suppression threshold, the waveform distortion value of the boost compensation voltage is calculated, and the distortion rate of the boost compensation voltage is divided according to the waveform distortion value to obtain the final target transmission voltage.

9. The control method according to claim 1, characterized in that, Based on the control signal that the target transmission voltage meets the trigger threshold, the multi-energy router is driven to perform power routing, injecting the remaining energy value into the flexible DC traction power supply network. Energy absorption closed-loop calculations are performed by continuously sampling the operating current and phase parameters of the access point until a preset energy value is reached, completing the energy transfer and releasing the power upper limit constraint. This includes: The target transmission voltage is classified according to the support vector machine algorithm to obtain the corresponding initial control signal; Determine whether the initial control signal is at a preset trigger threshold; When it is determined that the initial control signal is within a preset trigger threshold, the node configuration command is driven to output the system to perform flexible grid-connected energy routing interaction; The location of the current access point is determined based on the output of the system. The actual current and phase offset angle are obtained based on the location of the access point; The current absorption rate is obtained based on the actual current and phase offset angle. Determine whether the current absorption capacity is within the preset energy value; When the current absorption amount is at the preset energy value, it is determined that the flexible DC traction power supply network has completed the absorption of the remaining energy and the power upper limit constraint is released; When the current absorption amount is not at the preset energy value, the original constraint value is maintained, the current control command is stored in the sample pool to be verified, and the redundancy absorption query of adjacent nodes is calculated. When the initial control signal is not at the preset trigger threshold, the issuance of the current node configuration command is blocked, a warning is triggered, and the system waits for reclassification in the next cycle.

10. A flexible DC traction power supply system based on energy routing coordinated control, characterized in that, The control system includes a processor for executing the control method as described in any one of claims 1 to 9.

Images