Container type transformer anti-phase load method and device

Abstract

The invention relates to the technical field of transformer control, in particular to a container type transformer anti-phase load method and device, and the method comprises the steps: constructing a digital twinborn model, a reinforcement learning algorithm and a time series prediction model, and collecting operation data in real time through a multi-parameter sensor array, and dynamically adjusting the tap switch gear and cooperating with the reactive compensation module for switching, thereby realizing active suppression and accurate control of a reverse-phase load. The control response time can be remarkably shortened, the control precision and the dynamic adaptability are improved, meanwhile, the power loss of the transformer is reduced, it is ensured that the voltage stability and the temperature rise constraint meet the requirements, and a brand new technical path is provided for intelligent upgrading of the container type transformer.

Description

[0001] A method and apparatus for reverse-phase load on a containerized transformer Technical Field

[0002] This invention relates to the field of transformer technology, and in particular to a method and apparatus for reverse-phase load on a containerized transformer. Background Technology

[0003] As a modular core device in power distribution systems, containerized transformers (also known as box-type transformers) are widely used in new energy power generation, industrial parks, residential communities and municipal engineering projects due to their advantages such as "ready to use, small footprint and convenient maintenance". They play a key role in the "conversion-distribution-transmission" of electrical energy. With the upgrading of power systems towards "intelligent and low-carbon", the operational reliability and energy efficiency of containerized transformers are particularly important.

[0004] In actual operation, containerized transformers frequently face the challenge of reverse-phase loads. When distributed power sources backfeed, reactive power compensation devices malfunction, industrial loads experience sudden pulse changes, or three-phase loads become unbalanced, the load phase and power supply phase become opposite, leading to the following problems: First, reverse-phase circulating current exacerbates copper and iron losses in the transformer windings, significantly increasing energy efficiency losses; second, reverse-phase current causes localized overheating in the windings, accelerating the aging of insulation materials and shortening the transformer's lifespan; furthermore, reverse-phase voltage fluctuations may damage downstream sensitive loads, and traditional relay protection devices can only avoid equipment damage through "passive tripping," which is insufficient to meet the demand for uninterrupted power supply; finally, reverse-phase loads are "time-varying and random," and existing solutions rely on "fixed parameter thresholds" or "single-phase detection," resulting in long control response times and an inability to achieve precise control.

[0005] In response to the problem of reverse load, the existing technology mainly adopts three types of solutions, but all of them have certain limitations: (1) The passive protection solution triggers the trip after detecting the reverse signal through the relay protection device. Although it can avoid equipment damage, it will cause power outage and cannot meet the continuous power supply requirements of scenarios such as hospitals and data centers. (2) Fixed parameter type scheme presets voltage / current thresholds and adjusts the transformer tap changer or reactive power compensation module when the reverse load exceeds the threshold. However, it has limited adaptability to time-varying load characteristics and large control accuracy error. (3) The single detection type scheme relies on the single phase signal of the voltage transformer (PT) or current transformer (CT) to detect the reverse load. It is easily affected by harmonics and electromagnetic interference, has a higher false judgment rate, and a slow response speed.

[0006] Therefore, a method for reverse-phase load on a containerized transformer is needed to improve dynamic adaptability and control efficiency. Summary of the Invention

[0007] This invention provides a method and apparatus for reverse-phase load on a containerized transformer, used to optimize the transformer's operating performance under reverse-phase load conditions and improve its energy efficiency and reliability. A first aspect of this invention provides a method for reverse-phase load on a containerized transformer, comprising the following steps: S1. Construct a digital twin model of a containerized transformer, collect operating status data in real time through a multi-parameter sensor array deployed on the transformer body, and transmit the data to the processing unit through an edge computing gateway to set the dynamic parameters of the digital twin model to be updated periodically, and establish a high-fidelity virtual mirror system synchronized with the physical transformer; S2. Based on the virtual mirror system, the collected operating status data is fused and processed, including noise suppression and status estimation, in order to correct measurement deviations and accurately extract key electrical parameters that reflect the reverse-phase load characteristics of the transformer; S3. In the virtual mirror system, using reinforcement learning algorithms and taking the key electrical parameters as input, the system simulates the reverse load scenario and previews its development process. Through strategy optimization iteration, the system generates the optimal control strategy that is ahead of the actual state changes of the physical transformer. S4. The optimal control strategy is converted into specific control commands, and the tap position of the containerized transformer is dynamically adjusted to stabilize the winding voltage. The switching of the reactive power compensation module is controlled in coordination to compensate for reactive power and suppress circulating current, thereby achieving active suppression of reverse-phase loads. S5. By combining the time series prediction model, the development trend of the opposite load is predicted in advance, and coupled with the reinforcement learning decision-making mechanism, active predictive control with multi-objective optimization is realized in the virtual space.

[0008] As a further aspect of the present invention, the generation of the optimal control strategy in step S3 is based on a multi-objective optimization function, which includes at least the following optimization objectives: Minimize the overall power loss of the transformer caused by the reverse circulating current; The load-side voltage deviation is controlled within ±5% of the rated value; Suppress the three-phase current imbalance to below 10%.

[0009] As a further aspect of the present invention, the multi-objective optimization function also introduces constraint objectives related to transformer life, including: using the winding hot spot temperature predicted based on the thermal model as a hard constraint to ensure that it does not exceed the tolerance limit of the insulation material.

[0010] As a further aspect of the present invention, the reinforcement learning algorithm described in step S3 is a proximal policy optimization algorithm, whose reward function is designed based on the multi-objective optimization function. During the training process, the algorithm solves the Pareto optimal solution set by weighing the weights of each optimization objective.

[0011] As a further aspect of the present invention, when generating the optimal control strategy, the digital twin model is used to perform millisecond-level forward simulation verification of the strategy. The strategy is determined to be "optimal" and output only when the verification results simultaneously satisfy the constraints of voltage stability, loss reduction and temperature rise.

[0012] As a further aspect of the present invention, the operation of dynamically adjusting the tap changer position in step S4 is specifically as follows: when it is predicted that the phase difference between the load current phase and the power supply voltage phase is greater than 150 degrees, the tap changer is adjusted to reduce the transformation ratio by 1 to 2 positions to actively weaken the reverse load effect.

[0013] As a further aspect of the present invention, the switching of the collaborative control reactive power compensation module in step S4 is specifically as follows: after adjusting the tap changer, if the phase difference still has not recovered to the safe range, the optimal compensation capacity is calculated based on the real-time reactive power deficit, and the capacitor bank that can provide the maximum capacitive reactive power is put into operation first.

[0014] As a further aspect of the present invention, the issuance of control commands in step S4 follows a priority logic: tap changer adjustment is performed first to correct the phase at the source; after the tap changer adjustment is completed, reactive power compensation requirements are evaluated to avoid oscillation operations between devices.

[0015] As a further aspect of the present invention, the execution of step S4 has an adaptive cycle. When the rate of change of the key electrical parameter exceeds a set threshold, the control cycle is automatically shortened to less than 10ms to achieve rapid suppression of sudden reverse loads.

[0016] A second aspect of the present invention provides an apparatus for implementing the aforementioned reverse-phase load method for a containerized transformer, characterized in that it comprises: A multi-parameter sensor array is installed inside the containerized transformer enclosure to collect voltage, current, temperature and harmonic data in real time. An edge computing gateway, connected to the sensor array, is used for data collection, preprocessing, and transmission; The processing unit receives data from the edge computing gateway and internally builds and runs the digital twin model, reinforcement learning algorithm, and time series prediction model to generate control strategies. The actuator, including the drive motor of the tap changer and the switching switch of the reactive power compensation module, receives and executes control commands from the processing unit.

[0017] The beneficial effects of this invention are as follows: This invention constructs a digital twin model to collect and process transformer operating status data in real time, generating a high-fidelity virtual mirror system, providing a precise data foundation for subsequent control strategy formulation. Within the virtual mirror system, reinforcement learning algorithms simulate reverse-phase load scenarios and generate optimal control strategies. By dynamically adjusting tap changer positions and coordinating reactive power compensation module control, the impact of reverse-phase circulating current on the transformer is effectively suppressed. Combined with a time-series prediction model, advanced prediction of reverse-phase load development trends is achieved, and proactive predictive control through multi-objective optimization improves the transformer's energy efficiency and operational reliability. The above technical solution, through the coordinated operation of each step, solves the problems of low control accuracy and slow response speed in existing technologies, significantly reducing the damage of reverse-phase loads to transformers and extending equipment lifespan. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic flowchart of the reverse-phase load control method for a containerized transformer provided in an embodiment of the present invention; Figure 2 The structural connection block diagram of the containerized transformer reverse-phase load control device provided in the embodiment of the present invention is shown. Detailed Implementation

[0020] This invention provides a method and apparatus for reverse-phase load of a containerized transformer. The apparatus of this invention includes a multi-parameter sensor array 1, an edge computing gateway 2, a processing unit 3, and an execution mechanism, wherein the execution mechanism further includes a tap changer drive motor 4 and a reactive power compensation module switching switch 5.

[0021] The multi-parameter sensor array 1 is bolted to the inner wall of the containerized transformer tank. Its signal output terminal is connected to the edge computing gateway 2 via a shielded cable. The edge computing gateway 2 has a built-in low-pass filter and data compression module. It is connected to the multi-parameter sensor array 1 via an RS485 communication interface and to the processing unit 3 via an optical fiber communication interface. The processing unit 3 uses a high-performance embedded processor to receive data from the edge computing gateway 2 and run digital twin models, reinforcement learning algorithms, and time series prediction models. The tap changer drive motor 4 receives control commands from the processing unit 3 via a CAN bus. The reactive power compensation module switching switch 5 is controlled by a relay circuit.

[0022] In the specific implementation process, firstly, the voltage, current, temperature and harmonic data inside the containerized transformer tank are collected in real time through the multi-parameter sensor array 1. These data are transmitted to the edge computing gateway 2 through shielded cables. In the edge computing gateway 2, the data is initially filtered by a low-pass filter to remove high-frequency noise. Then, the data is compressed by the data compression module. Finally, the processed data is transmitted to the processing unit 3 through the fiber optic communication interface. The processing unit 3 updates the dynamic parameters in the digital twin model according to the received data, including winding resistance, core permeability and thermal aging coefficient of insulation material, so as to ensure that the virtual mirror system is consistent with the state of the physical transformer. This process constitutes the core content of step S1, namely, constructing a digital twin model of the containerized transformer.

[0023] In step S2, processing unit 3 performs fusion processing on the collected operating status data based on the virtual mirror system. Specifically, a Kalman filter algorithm is used to suppress noise in the collected operating status data, while an extended state observer is used to estimate the internal state variables of the transformer and correct measurement deviations caused by sensor errors or external interference. The gain matrix of the extended state observer is determined through offline simulation optimization to ensure the accuracy and real-time performance of the state estimation. By analyzing key electrical parameters such as voltage phase difference, current imbalance, and winding hot spot temperature, indicators reflecting the characteristics of reverse-phase load are extracted. This process achieves accurate processing of the operating status data, laying the foundation for the generation of subsequent control strategies.

[0024] In step S3, processing unit 3 uses reinforcement learning algorithms to simulate a reverse-phase load scenario and preview its development process in a virtual mirror system. Specifically, a near-end policy optimization algorithm is used as the core of the reinforcement learning algorithm. Key electrical parameters are used as inputs, and a reward function is designed to comprehensively consider the transformer's overall power loss caused by the reverse-phase circulating current, the load-side voltage deviation, and the three-phase current imbalance. The reward function also introduces the winding hot spot temperature predicted based on the thermal model as a hard constraint to ensure that it does not exceed the tolerance limit of the insulation material. During the policy optimization process, the Pareto optimal solution set is solved by weighing the weights of each optimization objective, and the effectiveness of the policy is verified by millisecond-level forward simulation. This process generates an optimal control policy that is ahead of the actual changes in the physical transformer state.

[0025] In step S4, the processing unit 3 converts the optimal control strategy into specific control commands, dynamically adjusts the tap position of the containerized transformer to stabilize the winding voltage, and coordinates the switching of the reactive power compensation module to compensate for reactive power and suppress circulating current.

[0026] When the phase difference between the load current and the power supply voltage is predicted to be greater than 150 degrees, the processing unit 3 generates a control command to adjust the tap changer by 1 to 2 positions in the direction of reducing the transformation ratio to weaken the reverse load effect. After adjusting the tap changer, if the phase difference still does not return to the safe range, the optimal compensation capacity is calculated based on the real-time reactive power deficit, and the capacitor bank that can provide the maximum capacitive reactive power is put into operation first. The control command is issued according to priority logic, and the tap changer adjustment is executed first to correct the phase. Then the reactive power compensation demand is evaluated to avoid oscillation operation between equipment. The tap changer drive motor 4 receives the control command through the CAN bus, and the reactive power compensation module switching switch 5 is controlled through the relay circuit.

[0027] In step S5, processing unit 3 combines a time series prediction model to predict the development trend of the reverse load. The time series prediction model is built based on a long short-term memory network (LSTM), and the training data includes historical operating status data and corresponding reverse load characteristic parameters. The prediction results are coupled with a reinforcement learning decision mechanism to generate a multi-objective optimized active predictive control strategy. When the rate of change of key electrical parameters exceeds the set threshold, the control cycle is automatically shortened to less than 10ms to quickly suppress sudden reverse loads. This process realizes active predictive control of the development trend of the reverse load and significantly improves the dynamic response capability of the system.

[0028] In practical applications, such as hospitals or data centers where continuous power supply is crucial, the device of this invention can effectively address reverse-phase load issues under complex operating conditions. Because the multi-parameter sensor array 1 collects real-time operating status data from the transformer tank and transmits it to the processing unit 3 via the edge computing gateway 2, the processing unit 3 generates an optimal control strategy based on a digital twin model, reinforcement learning algorithms, and a time series prediction model. It then executes control commands through a tap changer driving the motor 4 and the reactive power compensation module switching switch 5. Thus, throughout the entire process, the various components cooperate with each other through clear connections and data flow directions, ensuring the system's efficient operation.

[0029] The present invention achieves active suppression and multi-objective optimized control of reverse-phase loads in containerized transformers through the above-described specific embodiments, significantly improving the reliability and energy efficiency of transformer operation.

[0030] The device of the present invention collects the operating status data inside the transformer tank in real time through a multi-parameter sensor array 1, and transmits these data to an edge computing gateway 2 for preliminary processing. The edge computing gateway 2 has a built-in low-pass filter, which can effectively remove high-frequency noise interference. At the same time, it reduces the amount of data transmission through a data compression module, ensuring that the data is transmitted to the processing unit 3 efficiently. This process realizes high-precision monitoring of the transformer's operating status and lays the foundation for the generation of subsequent control strategies.

[0031] After receiving the preprocessed data, processing unit 3 first updates the dynamic parameters in the digital twin model, including winding resistance, core permeability, and thermal aging coefficient of insulation material. Real-time updates of these parameters ensure that the virtual mirror system maintains consistency with the physical transformer's state. Subsequently, processing unit 3 performs noise suppression on the collected operating status data based on the Kalman filter algorithm, and simultaneously uses an extended state observer to estimate the transformer's internal state variables, correcting measurement deviations caused by sensor errors or external interference. By analyzing key electrical parameters such as voltage phase difference, current imbalance, and winding hot spot temperature, indicators reflecting the characteristics of reverse-phase loads are extracted. This process achieves accurate processing of operating status data, providing reliable input for subsequent reinforcement learning algorithm applications.

[0032] In the virtual mirror system, processing unit 3 uses a near-end policy optimization algorithm as the core of the reinforcement learning algorithm, taking key electrical parameters as input to simulate a reverse-phase load scenario and preview its development process. The design of the reward function in this invention comprehensively considers the transformer's overall power loss caused by the reverse-phase circulating current, the load-side voltage deviation, and the three-phase current imbalance, and introduces the winding hot spot temperature predicted based on the thermal model as a hard constraint to ensure that it does not exceed the tolerance limit of the insulation material. During the policy optimization process, the Pareto optimal solution set is solved by weighing the weights of each optimization objective, and the effectiveness of the policy is verified by millisecond-level forward simulation. This process generates an optimal control policy that anticipates the actual changes in the physical transformer's state.

[0033] When the predicted phase difference between the load current and the power supply voltage is greater than 150 degrees, the processing unit 3 generates a control command to adjust the tap changer by 1 to 2 positions to reduce the ratio, thereby weakening the reverse load effect. The tap changer drive motor 4 receives the control command via the CAN bus and executes the corresponding position adjustment operation. If the phase difference still does not return to the safe range after adjusting the tap changer, the optimal compensation capacity is calculated based on the real-time reactive power deficit, and the capacitor bank that can provide the maximum capacitive reactive power is prioritized for activation. The reactive power compensation module switching switch 5 is controlled by a relay circuit, with the relay contact rated current not less than 50A to ensure the safety and reliability of the control process. The issuance of control commands follows priority logic, prioritizing tap changer adjustment to correct the phase, followed by assessment of reactive power compensation requirements to avoid oscillating operation between equipment.

[0034] Meanwhile, processing unit 3 combines a time series prediction model to predict the development trend of reverse-phase loads in advance. The time series prediction model is built based on a long short-term memory network (LSTM), and the training data includes historical operating status data and corresponding reverse-phase load characteristic parameters. The prediction results are coupled with a reinforcement learning decision mechanism to generate a multi-objective optimized active predictive control strategy. When the rate of change of key electrical parameters exceeds the set threshold, the control cycle is automatically shortened to less than 10ms to quickly suppress sudden reverse-phase loads. This process significantly improves the dynamic response capability of the system and ensures that the transformer can still maintain stable operation under complex working conditions.

[0035] Throughout the process, the various components cooperate with each other through clear connections and data flow directions to ensure the efficient operation of the system. For example, the multi-parameter sensor array 1 is connected to the edge computing gateway 2 through a shielded cable, the edge computing gateway 2 is connected to the processing unit 3 through a fiber optic communication interface, the tap changer drive motor 4 receives control commands through a CAN bus, and the reactive power compensation module switching switch 5 is controlled through a relay circuit. This modular design not only improves the reliability of the system, but also facilitates later maintenance and upgrades.

[0036] Through the above steps, the device of the present invention successfully addressed the reverse-phase load problem under complex operating conditions in a hospital setting. The multi-parameter sensor array 1 collects real-time operating status data from within the transformer tank and transmits it to the processing unit 3 via the edge computing gateway 2. The processing unit 3 generates the optimal control strategy based on a digital twin model, reinforcement learning algorithm, and time series prediction model, and executes the control commands through the tap changer driving the motor 4 and the reactive power compensation module switching switch 5. Throughout the process, the various components cooperate with each other through clear connections and data flow directions, ensuring the efficient operation of the system.

[0037] In summary, the present invention, through the above specific embodiments, achieves active suppression and multi-objective optimized control of reverse-phase loads in containerized transformers, significantly improving the reliability and energy efficiency of transformer operation.

[0038] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for reverse-phase load on a containerized transformer, characterized in that, Includes the following steps: S1. Construct a digital twin model of a containerized transformer, collect operating status data in real time through a multi-parameter sensor array deployed on the transformer body, and transmit the data to the processing unit through an edge computing gateway to set the dynamic parameters of the digital twin model to be updated periodically, and establish a high-fidelity virtual mirror system synchronized with the physical transformer; S2. Based on the virtual mirror system, the collected operating status data is fused and processed, including noise suppression and status estimation, in order to correct measurement deviations and accurately extract key electrical parameters that reflect the reverse-phase load characteristics of the transformer; S3. In the virtual mirror system, using reinforcement learning algorithms and taking the key electrical parameters as input, the system simulates the reverse load scenario and previews its development process. Through strategy optimization iteration, the system generates the optimal control strategy that is ahead of the actual state changes of the physical transformer. S4. The optimal control strategy is converted into specific control commands, and the tap position of the containerized transformer is dynamically adjusted to stabilize the winding voltage. The switching of the reactive power compensation module is controlled in coordination to compensate for reactive power and suppress circulating current, thereby achieving active suppression of reverse-phase loads. S5. By combining the time series prediction model, the development trend of the opposite load is predicted in advance, and coupled with the reinforcement learning decision-making mechanism, active predictive control with multi-objective optimization is realized in the virtual space.

2. The method for reverse-phase load on a containerized transformer according to claim 1, characterized in that, The generation of the optimal control strategy described in step S3 is based on a multi-objective optimization function, which includes at least the following optimization objectives: Minimize the overall power loss of the transformer caused by the reverse circulating current; The load-side voltage deviation is controlled within ±5% of the rated value; Suppress the three-phase current imbalance to below 10%.

3. The method for reverse-phase load of a containerized transformer according to claim 2, characterized in that, The multi-objective optimization function also introduces constraint objectives related to transformer life, including: using the winding hot spot temperature predicted based on the thermal model as a hard constraint to ensure that it does not exceed the tolerance limit of the insulation material.

4. The method for reverse-phase load of a containerized transformer according to claim 3, characterized in that, The reinforcement learning algorithm described in step S3 is a proximal policy optimization algorithm. Its reward function is designed based on the multi-objective optimization function. During the training process, the algorithm solves the Pareto optimal solution set by weighing the weights of each optimization objective.

5. The method for reverse-phase load of a containerized transformer according to claim 4, characterized in that, When generating the optimal control strategy, the digital twin model is used to perform millisecond-level forward simulation verification of the strategy. The strategy is only determined to be "optimal" and output when the verification results simultaneously meet the constraints of voltage stability, loss reduction and temperature rise.

6. The method for reverse-phase load of a containerized transformer according to claim 1, characterized in that, The specific operation of dynamically adjusting the tap changer position in step S4 is as follows: when it is predicted that the phase difference between the load current and the power supply voltage is greater than 150 degrees, the tap changer is adjusted to reduce the transformation ratio by 1 to 2 positions to actively weaken the reverse load effect.

7. The method for reverse-phase load of a containerized transformer according to claim 6, characterized in that, The switching of the collaborative control reactive power compensation module in step S4 is as follows: after adjusting the tap changer, if the phase difference still has not returned to the safe range, the optimal compensation capacity is calculated based on the real-time reactive power deficit, and the capacitor bank that can provide the maximum capacitive reactive power is put into operation.

8. The method for reverse-phase load of a containerized transformer according to claim 7, characterized in that, The control commands issued in step S4 follow a priority logic: tap changer adjustment is performed first to correct the phase at the source; after the tap changer adjustment is completed, reactive power compensation requirements are evaluated to avoid oscillation between devices.

9. The method for reverse-phase load on a containerized transformer according to claim 1, characterized in that, The execution of step S4 has an adaptive cycle. When the rate of change of the key electrical parameter exceeds the set threshold, the control cycle is automatically shortened to less than 10ms to achieve rapid suppression of sudden reverse loads.

10. An apparatus for implementing the method for reverse-phase load of a containerized transformer according to any one of claims 1 to 9, characterized in that, include: A multi-parameter sensor array is installed inside the containerized transformer enclosure to collect voltage, current, temperature and harmonic data in real time. An edge computing gateway, connected to the sensor array, is used for data collection, preprocessing, and transmission; The processing unit receives data from the edge computing gateway and internally builds and runs the digital twin model, reinforcement learning algorithm, and time series prediction model to generate control strategies. The actuator, including the drive motor of the tap changer and the switching switch of the reactive power compensation module, receives and executes control commands from the processing unit.

Images