Photovoltaic box transformer ventilation system with capacity expansion temperature control protection

Abstract

This invention relates to the field of photovoltaic equipment control technology and discloses a photovoltaic transformer substation ventilation system with extended capacity and temperature control protection, including a sensing and acquisition module, a control module, and an execution module. The control module acquires current and temperature data, extracts the load ramp-up slope, and when the current and slope exceed limits, it starts the winding fan before the winding temperature reaches the target, achieving load feedforward pre-cooling and reducing transient temperature overshoot. Simultaneously, it integrates the effective temperature rise of the transformer substation volume over time and controls the start and stop of the casing fan based on the volume heat accumulation degree. Combined with contact forced operation sequence isolation logic, it effectively eliminates frequent fan operation and contact erosion problems caused by environmental disturbances. Furthermore, by extracting the steady-state observation interval, constructing equivalent heat capacity features, and combining them with a pre-set neural network model to calculate the heat dissipation degradation index, the system achieves early warning of progressive faults such as filter blockage, improving equipment overload capacity and hardware lifespan, and enabling on-demand condition-based maintenance.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic equipment control technology, specifically to a photovoltaic transformer ventilation system with expanded capacity temperature control protection. Background Technology

[0002] With the increasing global demand for clean energy, photovoltaic (PV) power generation technology has been widely adopted. As one of the core components of a PV power generation system, the PV dry-type transformer substation undertakes critical functions such as voltage transformation, power transmission, and system protection. Its operating efficiency and safety directly affect the performance of the entire PV power station. During long-term operation, especially under high-load conditions, a large amount of heat is generated inside the dry-type transformer substation. If faced with a high-temperature environment and insufficient heat dissipation, it can lead to increased iron and copper losses, decreased operating efficiency, and even serious faults such as insulation aging, partial discharge, and overload. Therefore, equipping the transformer substation with an effective ventilation system and automatic temperature control device is essential to ensure the safe operation of the equipment.

[0003] However, existing photovoltaic transformer substation ventilation and temperature control systems still have many structural and control deficiencies in practical applications. In terms of hardware and daily maintenance, traditional transformer substation casing fans often adopt a fixed design structure, and their inlet filters easily accumulate dust during daily operation. When it is necessary to clean clogged filters or repair aging and damaged fans, due to the existing structure, it is often necessary to shut down the entire transformer substation before any work can be done. This directly results in power loss from the photovoltaic system and increases operation and maintenance costs. At the same time, existing winding fans are limited by traditional air intake methods and filter structures, restricting their actual ventilation volume and making it difficult to provide sufficient large-volume heat dissipation support when the equipment faces high-temperature loads.

[0004] In terms of automated temperature control, existing temperature controllers generally have small control contact capacity and relatively simple control logic. When the internal temperature of the transformer substation fluctuates slightly due to environmental or load conditions, simple threshold comparison logic can easily lead to frequent start-stop of the fans. This frequent operation not only makes it difficult to effectively maintain the stability of the internal temperature of the substation, but also generates electric arcs due to frequent contactor engagement and disengagement, causing contact erosion and damage to the temperature controller, severely reducing the overall reliability of the control system.

[0005] Furthermore, traditional equipment condition monitoring methods primarily rely on real-time data transmitted by physical temperature probes, which suffers from significant delays in physical heat conduction. Faced with gradual declines in heat dissipation capacity, such as dust filters becoming clogged or fan bearings aging, traditional systems lack the ability to predict and assess trends in advance. They typically only trigger over-temperature alarms passively after the internal temperature has exceeded limits and heat accumulation has occurred. This delayed monitoring mechanism places immense pressure on on-site manual inspections, failing to meet the demands of modern photovoltaic systems for refined equipment management and high-reliability operation. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a photovoltaic transformer substation ventilation system with expanded capacity temperature control protection. The technical problems solved by this invention are: photovoltaic dry-type transformer substations face high-temperature environments during long-term operation, leading to insufficient heat dissipation, which in turn causes insulation aging, partial discharge, and overload faults; traditional transformer substation casing fans often require power outages for routine cleaning and maintenance, resulting in power loss from the photovoltaic system; existing temperature controllers have small control contact capacity and simple control logic, making them prone to contact damage under conditions of frequent fan start-stop cycles due to temperature fluctuations, making it difficult to maintain stable internal temperature of the transformer substation; furthermore, traditional temperature monitoring methods suffer from physical conduction delays, lacking the ability to predict gradual heat dissipation faults such as filter blockage and fan aging, resulting in high workload for manual inspections.

[0007] To address the above problems, the present invention provides the following technical solution:

[0008] This invention provides a photovoltaic transformer substation ventilation system with expanded capacity temperature control protection, including a sensing and acquisition module, a control module, and an execution module connected to the control module; the sensing and acquisition module is used to acquire current data and temperature data of the photovoltaic transformer substation, the temperature data including at least the winding temperature and the substation volume temperature; the execution module includes at least a winding fan and a casing fan; The control module is used to: calculate the time derivative of the current data to obtain the load ramp-up slope; when the absolute value of the current data is greater than a preset base load threshold and the load ramp-up slope is greater than a preset extreme ramp-up threshold, generate and output a winding fan start-up signal before the winding temperature reaches a preset winding fan start-up temperature threshold; calculate the effective temperature rise difference between the transformer substation volume temperature and the preset base reference temperature, and perform cumulative integration on the effective temperature rise difference to obtain a temperature rise time integral value; perform a start-up determination for the casing fan based on the temperature rise time integral value and output a casing fan start-up signal; and control the execution module to drive the corresponding fan according to the winding fan start-up signal or the casing fan start-up signal.

[0009] This invention optimizes the triggering logic of the ventilation system by combining load feedforward precooling and volumetric heat accumulation integral control. For the transformer windings, the load rise slope is used as an input parameter to activate the fan before Joule heat conduction reaches the surface temperature sensor, establishing forced convection, reducing transient temperature overshoot, and improving the transformer's dynamic overload capacity. For the internal space of the transformer substation, the temperature rise time integral value is used to characterize the accumulated heat exceeding the baseline reference temperature, measuring the degree of heat accumulation in the space and avoiding false triggering caused by short-term fluctuations in ambient temperature.

[0010] Furthermore, the control module is also used to perform digital filtering on the current data and the temperature data to filter out high-frequency interference signals and generate time series parameters. These time series parameters include at least a steady-state current time series and a steady-state temperature time series. Specifically, the control module is used to extract the steady-state parameters from the steady-state current time series and the steady-state temperature time series as the current data and the temperature data, respectively, for use in various threshold determinations and slope calculations. This processing filters out transient distortions caused by photovoltaic inverter switching actions and electromagnetic interference in the sampling circuit, providing a stable data source for subsequent calculations.

[0011] Furthermore, the control module is also used to perform backup protection determination: when the conditions that the absolute value of the current data is greater than the basic load threshold and the load ramp rate is greater than the rapid ramp threshold are not simultaneously met, if the absolute value of the current data is greater than the preset high load protection threshold, or if the winding temperature is greater than the winding fan start-up temperature threshold, a winding fan start-up signal is generated and output. This setting ensures that the system can still perform backup heat dissipation protection under conditions of slow load ramp-up or rising ambient temperature.

[0012] Furthermore, the specific logic for the control module to determine the start-up of the casing fan based on the integral value of the temperature rise time is as follows: when the integral value of the temperature rise time reaches a preset volumetric heat accumulation trigger threshold, a start signal for the casing fan is output; when the volumetric temperature of the transformer is detected to be less than or equal to the baseline reference temperature, the effective temperature rise difference is set to zero, and the integral value of the temperature rise time is simultaneously cleared to zero. This clearing mechanism aligns with the natural heat dissipation process of the transformer in a low-temperature environment, preventing small positive temperature differences from causing invalid long-term integral accumulation.

[0013] In a preferred embodiment of the present invention, the execution module further includes a temperature controller and a contact expansion unit connected to the control module. The output contact of the temperature controller is connected to the control terminal of the contact expansion unit, and the main contact of the contact expansion unit is connected in series with the power supply circuit of the corresponding fan. After generating the corresponding fan start signal, the control module controls the temperature controller to close the corresponding output contact, thereby driving the contact expansion unit to close the corresponding main contact and start a forced operation timer. During the activation interval of the forced operation timer, the shutdown logic is suspended to keep the output contact and its corresponding main contact closed. When the accumulated value of the forced operation timer reaches a preset forced operation time threshold, and the shutdown condition of the fan corresponding to the main contact is met, the control module controls the temperature controller to disconnect the output contact, thereby driving the contact expansion unit to disconnect the corresponding main contact and start a forced cooling timer. During the activation interval of the forced cooling timer, the start request is suspended to keep the output contact and its corresponding main contact open. Specifically, when the corresponding fan is the winding fan, the shutdown conditions include at least the winding temperature being lower than a preset winding shutdown temperature threshold, the steady-state current generated after digital filtering of the current data being lower than a preset shutdown load threshold, and the load ramp-up slope being lower than a preset reset slope threshold. When the corresponding fan is the casing fan, the shutdown conditions include at least the transformer substation volume temperature being lower than a preset casing fan shutdown temperature threshold or lower than the base reference temperature. By setting timer activation intervals for forced operation and forced cooling, wind pressure build-up time and contact cooling time are reserved for the actuator, avoiding frequent contactor operation caused by data jitter under critical conditions.

[0014] Furthermore, the control module establishes a state machine for each wind turbine control channel. The state machine includes at least four states: shutdown lockout, shutdown standby, forced operation, operation pending judgment, and emergency protection. The emergency protection state corresponds to the detection of short circuits, severe overloads, severe overheating, or wind turbine faults. When the winding temperature or the transformer substation volume temperature exceeds a preset emergency protection threshold, the control module releases the forced cooling isolation restriction established by the forced cooling timer. The state machine manages the contact action states of the wind turbine control channels and sets the highest response priority for the emergency protection state, ensuring timely release of the delay lockout and execution of protection actions when a fault occurs.

[0015] Furthermore, the control module is also used to intercept the current data and the transformer volume temperature within the activation interval of the forced operation timer corresponding to the casing fan; calculate the weighted sum of the square integral value of the current and the linear integral value of the current within the interval to obtain the comprehensive thermodynamic input equivalent, and obtain the effective cooling range between the temperature at the start and end of the interval; when the effective cooling range is less than the preset minimum temperature difference, the preset minimum temperature difference is used in subsequent calculations; calculate the first heat dissipation decay index based on the comprehensive thermodynamic input equivalent and the effective cooling range; when the first heat dissipation decay index of multiple cycles exceeds the set maintenance limit after sliding average, output a filter cleaning command to the external terminal.

[0016] This setup uses the heat input and cooling rate during forced ventilation to quantitatively assess the heat dissipation degradation of the equipment.

[0017] In another preferred embodiment of the present invention, the sensing and acquisition module further includes a solar irradiance sensor for acquiring external energy input and a load power acquisition unit for acquiring internal energy dissipation; the control module acquires a solar irradiance sequence, a load parameter sequence, and a winding temperature sequence generated based on the winding temperature within a fixed sampling period, and calculates the winding temperature change rate of the winding temperature sequence; a valid deterministic steady-state observation interval is confirmed only when the standard deviation of the fluctuation of the solar irradiance sequence is lower than a preset solar irradiance stability threshold, the standard deviation of the fluctuation of the load parameter sequence is lower than a preset load parameter stability threshold, and the absolute value of the winding temperature change rate is lower than a preset thermal balance micro-change threshold, and the above conditions are continuously satisfied for a period exceeding a preset shortest steady-state interval threshold.

[0018] By using a multi-condition joint judgment to extract the steady-state observation interval in thermal equilibrium, the data bias caused by unsteady-state processes is reduced.

[0019] Furthermore, after extracting the deterministic steady-state observation interval, the control module calculates the average load parameters within the deterministic steady-state observation interval, and calculates the internal electrical loss heat equivalent by combining the transformer no-load loss and rated load loss; calculates the average solar irradiance within the deterministic steady-state observation interval, and calculates the external radiation heat equivalent by combining the effective light-receiving area of ​​the photovoltaic transformer and the comprehensive absorption coefficient of the transformer surface; after algebraically adding the internal electrical loss heat equivalent and the external radiation heat equivalent, it divides by the preset system limit heating benchmark value for normalization processing to generate a dimensionless total equivalent heat equivalent feature.

[0020] This process unifies internal electrical losses and external environmental radiation into a dimensionless total equivalent thermal characteristic, simplifying the input parameters and conforming to the actual thermodynamic characteristics of the system.

[0021] Furthermore, the control module integrates a pre-trained and fixed feedforward neural network model; the control module inputs the dimensionless total equivalent thermal characteristics into the feedforward neural network model and outputs the predicted reference winding temperature; it calculates the winding temperature at the end of the deterministic steady-state observation interval as the actual winding temperature, and calculates the difference between the actual winding temperature and the predicted reference winding temperature to generate a temperature residual; it calculates the arithmetic mean of the temperature residuals within the statistical period to obtain a second heat dissipation decay index, and outputs an alarm message when the second heat dissipation decay index exceeds a preset decay alarm threshold.

[0022] A temperature prediction model under healthy conditions is established using neural networks. The heat dissipation degradation is assessed by calculating the residual between the actual winding temperature and the predicted reference winding temperature, thus eliminating the interference of ambient temperature and solar radiation changes on the judgment of filter blockage.

[0023] Furthermore, a pull-out filter assembly is installed at the air inlet of the transformer substation casing. The pull-out filter assembly includes a slide rail, a filter frame, dustproof filter material, and a locking component. The filter frame can be pulled out from the outside of the transformer substation along the slide rail to achieve non-damaging disassembly, cleaning, and reuse of the filter. The casing fan is located in an independent maintenance fan box outside the exhaust vent of the transformer substation casing. The independent maintenance fan box is isolated from the high-voltage space inside the transformer substation and is equipped with an openable maintenance door to allow maintenance of the casing fan and adjacent filter assemblies without entering the high-voltage space inside the transformer substation. The air inlet duct of the winding fan is equipped with a bottom air inlet cavity, a guide shroud, and a guide plate to change the air inlet direction of the winding fan and increase the effective air intake. The temperature controller drives the fan power circuit through a contact expansion unit. The contact expansion unit is at least one of an intermediate relay, an AC contactor, or a solid-state relay to expand the output contact capacity of the temperature controller and reduce the risk of contact erosion.

[0024] This invention provides a photovoltaic transformer substation ventilation system with extended capacity temperature control protection. It has the following beneficial effects: 1. This invention constructs a load feedforward pre-cooling control logic based on current data and load ramp rate, which can trigger the winding fan to start before the Joule heat inside the transformer is substantially conducted outward. This method overcomes the physical conduction delay defect of traditional temperature measurement methods, improves the forced convection cooling efficiency of the transformer body, and reduces the maximum operating temperature of the equipment, thereby reducing the risk of insulation aging and thermal runaway of photovoltaic transformers under extreme high temperature and high load operating conditions.

[0025] 2. This invention eliminates the interference of external environmental disturbances and transient temperature jumps on the fan operation by introducing volumetric heat accumulation integral control and contact forced operation sequence isolation logic. The system uses the temperature rise time integral value for action anti-jitter, and in conjunction with forced operation and cooling timers to shield reset commands in critical states, it changes the traditional single threshold contact triggering method from the underlying control logic. This not only maintains the stability of the internal volume temperature of the transformer, but also solves the technical problem of easy burn-out and damage to the temperature controller contacts caused by frequent fan start-stop, thus improving the long-term reliability of the control hardware.

[0026] 3. This invention extracts deterministic steady-state observation intervals and constructs photoelectric equivalent heat equivalence features, combined with an internally embedded feedforward neural network, to achieve accurate calculation and dynamic tracking of the system's heat dissipation degradation index. This mechanism endows the system with early warning capabilities for progressive filter clogging and fan performance aging, changing the traditional maintenance mode that relies on manual periodic power outage inspections. It realizes data-driven on-demand condition-based maintenance, reducing the pressure of manual inspections and minimizing photovoltaic power generation losses caused by unnecessary power outages for cleaning and maintenance. Attached Figure Description

[0027] Figure 1 This is an architectural topology diagram of a photovoltaic transformer substation ventilation system according to an embodiment of the present invention; Figure 2 This is a flowchart of a ventilation system control method according to an embodiment of the present invention; Figure 3 This is a control logic diagram for the feedforward precooling of a winding fan according to an embodiment of the present invention; Figure 4 This is a control logic diagram for the volumetric heat accumulation of the casing fan according to an embodiment of the present invention; Figure 5 This is a flowchart of the deterministic steady-state observation interval extraction according to an embodiment of the present invention; Figure 6 This is a flowchart of the photoelectric homogeneous equivalent heat equivalent calculation according to an embodiment of the present invention; Figure 7 This is a flowchart of the heat dissipation decay index sequence calculation and judgment according to an embodiment of the present invention; Figure 8 This is a comprehensive data diagram illustrating the control logic deduction and experimental verification of an embodiment of the present invention. Detailed Implementation

[0028] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] See attached document Figure 1 This invention provides a photovoltaic transformer ventilation system with expanded capacity temperature control protection, which is applied to photovoltaic dry-type transformers and mainly includes a sensing and acquisition module, a control module and an execution module.

[0030] The sensor acquisition module includes a current transformer, temperature probes, a solar irradiance sensor, and a load power acquisition unit. The current transformer is installed on the low-voltage output busbar of the photovoltaic transformer substation to acquire real-time AC load current. The temperature probes include winding temperature probes, core temperature probes, and substation volume temperature probes, used to acquire the transformer winding temperature, core temperature, and internal air volume temperature of the substation, respectively. The solar irradiance sensor is located at the external sunlight-receiving position of the substation to acquire solar irradiance data of the surrounding environment. The load power acquisition unit is used to acquire low-voltage side voltage, current, and power factor, or to receive load power data uploaded by the inverter or substation monitoring and control device via a communication interface.

[0031] The control module is a microcontroller or programmable logic controller (PLC). It connects to the sensor acquisition module via a data bus. Internally, the control module incorporates filtering algorithms and a state machine program to process acquired signals and generate control pulses. The control module also connects to a communication alarm module, which outputs filter cleaning commands, fan maintenance commands, and heat dissipation alarm messages to external terminals via RS485, Ethernet, wireless communication, or the transformer substation monitoring protocol.

[0032] The execution module includes a temperature controller, a contact expansion unit, a winding fan, and a casing fan. The input terminal of the temperature controller is connected to the output terminal of the control module. The control output contacts of the temperature controller are connected to the control terminal of the contact expansion unit. The contact expansion unit is at least one of an intermediate relay, an AC contactor, or a solid-state relay. The main contacts of the contact expansion unit are connected in series in the power supply circuit of the corresponding fan. The control output contacts of the temperature controller are only used to drive the control terminal of the contact expansion unit. This hierarchically isolates the temperature controller contact circuit from the fan power circuit, expands the carrying capacity of the temperature controller output contacts, and meets the contact capacity requirements under frequent start-stop conditions of the fan.

[0033] The winding fan is installed at the bottom of the transformer winding. The air inlet of the winding fan includes a bottom air inlet cavity, a guide shroud, and at least one guide plate. The bottom air inlet cavity is connected to the air inlet of the transformer. The guide shroud guides the incoming cooling airflow to the lower gap of the transformer winding and the heat dissipation channel on the side of the winding. The guide plate is set as an arc-shaped guide plate or an inclined guide plate to reduce the sudden expansion resistance and sharp bend resistance in the air inlet channel. This changes the air inlet direction of the winding fan from a single end face air inlet to a combination of bottom air inlet and side guide flow, thereby increasing the effective air intake of the winding fan.

[0034] The transformer substation enclosure is equipped with an air inlet and an air outlet. A pull-out filter assembly is installed at the air inlet. The pull-out filter assembly includes a slide rail fixed to the outside of the air inlet, a filter frame that can be pulled out along the slide rail, dustproof filter media installed inside the filter frame, and a spring-loaded clip or knob plate for locking the filter frame. A circumferential sealing strip is installed between the filter frame and the air inlet. Maintenance personnel can release the spring-loaded clip or knob plate on the outside of the transformer substation and pull the filter frame out along the slide rail to clean, replace, or reinstall the dustproof filter media, achieving non-damaging disassembly and reuse of the filter.

[0035] The casing fan is installed in an independent maintenance fan box outside the exhaust vent of the transformer substation. This independent maintenance fan box is separated from the high-voltage space inside the substation by an exhaust isolation plate, which has ventilation holes corresponding to the air inlet side of the casing fan. An openable maintenance door is provided on the outside of the independent maintenance fan box. The casing fan is secured inside the box using a detachable bracket or quick-release fasteners. During maintenance, personnel can open the maintenance door from the outside of the substation to disassemble or clean the casing fan and its adjacent filter components without entering the high-voltage space inside the substation, thus enabling external maintenance without power interruption.

[0036] See attached document Figure 2 This invention provides a ventilation control method for photovoltaic transformer substations with extended capacity temperature control protection, comprising the following steps: S10: Obtain the physical parameters of the photovoltaic transformer box. The control module reads the discrete current data and temperature data transmitted by the sensor acquisition module, performs digital filtering on the current data and temperature data to filter out high-frequency interference signals, and generates time series parameters. S20, execute the graded trigger judgment, the control module calculates the time derivative of the real-time current to obtain the load rise slope, and combines the absolute value of the current, the load rise slope and the winding temperature to execute the winding fan start judgment; the control module performs cumulative integral calculation with reset condition on the effective temperature rise difference of the transformer volume temperature exceeding the basic reference temperature, calculates the temperature rise time integral value, and outputs the shell fan start signal when the temperature rise time integral value reaches the set volume heat accumulation threshold. S30, execute contact timing control. After the control module generates the fan start signal, it controls the temperature controller to close the corresponding output contact. The output contact drives the contact expansion unit to close the corresponding main contact and starts the forced run timer. The forced run timer starts accumulating when the output contact or the corresponding main contact closes. During the active interval of the forced run timer, the shutdown logic is suspended to keep the output contact and its corresponding main contact closed. When the accumulated value of the forced run timer reaches the forced run time threshold and the shutdown condition corresponding to the winding fan or the shell fan is met, the control module controls the temperature controller to disconnect the output contact. The output contact drives the contact expansion unit to disconnect the corresponding main contact and starts the forced cooling timer. The forced cooling timer starts accumulating when the output contact or the corresponding main contact opens. During the active interval of the forced cooling timer, the start request is suspended to keep the output contact and its corresponding main contact open. The shutdown conditions for winding fans include winding temperature below the winding shutdown temperature threshold, steady-state current below the shutdown load threshold or maintained within the allowable long-term operating range, and load ramp-up slope below the reset slope threshold; the allowable long-term operating range is determined based on the transformer's rated capacity, long-term allowable load rate, and winding temperature margin; the shutdown conditions for shell fans include transformer volume temperature below the shell fan shutdown temperature threshold or below the base reference temperature; the starting temperature thresholds for winding fans and shell fans are respectively higher than their corresponding shutdown temperature thresholds to form a temperature hysteresis; S40, perform filter status assessment, the control module preferably intercepts the current data and transformer volume temperature data within the activation interval of the forced operation timer of the casing fan, calculates the weighted sum of the square integral value of the current and the first term integral value of the current within the interval to obtain the comprehensive thermodynamic input equivalent, and calculates the effective cooling range between the temperature at the beginning and end of the interval. The effective cooling range is calculated by substation volume temperature at the start of the forced operation period of the casing fan minus substation volume temperature at the end. When the effective cooling range is less than the preset minimum temperature difference, the preset minimum temperature difference is used in the calculation of the first heat dissipation degradation index. The control module calculates the first heat dissipation degradation index based on the comprehensive thermodynamic input equivalent and the effective cooling range. The first heat dissipation degradation index increases with the increase of the comprehensive thermodynamic input equivalent and decreases with the increase of the effective cooling range. When the first heat dissipation degradation index of multiple cycles shows an upward trend after sliding average and exceeds the set maintenance limit, a filter cleaning command is output to the external terminal.

[0037] To avoid terminological confusion, in this embodiment, load parameters include at least one of load power, apparent load, load current, or equivalent load factor; when performing internal electrical loss heat equivalent calculation, steady-state average apparent load or equivalent load factor is preferred. The transformer substation volumetric steady-state temperature parameter is used for volumetric heat accumulation integral control of the casing fan, and the winding temperature parameter is used for winding fan control and reference winding temperature prediction. The first heat dissipation decay index characterizes the change in heat dissipation capacity of the casing ventilation path during forced ventilation, and the second heat dissipation decay index characterizes the degree of deviation of the actual winding temperature from the healthy reference winding temperature.

[0038] The underlying control principle and calculation logic of the technical solution of this invention will be explained in detail below with reference to specific functional modules.

[0039] The control module reads the discrete current and temperature data transmitted by the sensor acquisition module and converts the continuous analog physical quantities inside the transformer into discrete digital sequences that the control module can calculate. The specific process is as follows.

[0040] S101 sets the data acquisition time base and interrupt priority. The hardware timer inside the control module is set to have a fixed period. The interrupt source. To balance the computational overhead of the control module with the accuracy of sampling and reconstruction, The value range is typically set to 10 milliseconds to 50 milliseconds. When the current is less than one power frequency cycle, the control module calculates the root mean square value of the current based on a sliding sampling window covering at least one complete power frequency cycle; when... When the current is greater than or equal to one power frequency cycle, the control module calculates the root mean square value of the current based on at least one complete power frequency cycle covered within the current sampling period. When the timer generates an overflow interrupt, the control module suspends background routine tasks and prioritizes the execution of the analog-to-digital converter sampling service routine to maintain the time series parameters with equal time intervals.

[0041] S102, Discretize the load current signal. The control module triggers the analog-to-digital converter (ADC) channel electrically connected to the current transformer. To accurately assess the heating effect of the AC load inside the transformer, in one embodiment, the small AC signal output from the current transformer is converted into an AC sampling signal adapted to the ADC channel by isolation, limiting, and bias conditioning circuits outside the control module. The control module performs multiple discrete samplings of the conditioned AC sampling signal within at least one power frequency cycle and calculates the root mean square (RMS) value to generate the effective current data corresponding to the current moment. .

[0042] In another implementation, the current transformer output signal is converted into a DC voltage signal corresponding to the effective value of the current via a true RMS conversion circuit or a rectifier-filter circuit. The control module samples this DC voltage signal and converts it into effective current data. Simultaneously, the control module performs data out-of-bounds validity checks, eliminating abnormal data that exceeds the physical range of the current transformer. It should be noted that if the current data sampled in the current cycle is determined to be abnormal and eliminated, the control module will automatically retrieve valid historical current data from the previous cycle to fill the gap, ensuring the physical continuity of the discrete data sequence on the time axis and avoiding logical dead zones in subsequent mathematical operations.

[0043] When current data from multiple consecutive sampling cycles is determined to be abnormal, the control module generates a current acquisition fault alarm and enters a conservative protection mode. In conservative protection mode, the control module maintains the operation of the corresponding fan or outputs fan maintenance and sensor acquisition fault alarms to avoid heat dissipation control failure due to current acquisition failure. Regarding the root mean square calculation of AC signals and the configuration of hardware conditioning circuits, those skilled in the art can implement them according to conventional instrument design manuals. The specific circuit topology is well-known in the field and will not be described in detail here.

[0044] S103 performs discrete sampling of the volume and heat source temperature signals. The control module triggers the analog-to-digital converter channel electrically connected to the temperature probe. The resistance of the temperature probe changes with the internal temperature of the transformer substation and is converted into a voltage signal by the drive circuit. The control module reads this voltage signal and converts it into actual physical temperature data by consulting a calibration table or performing linear interpolation. Among them, physical temperature data At least include transformer substation volume temperature data; and physical temperature data when used for winding fan control or reference winding temperature prediction. It also includes winding temperature data or temperature measurement data near the winding.

[0045] To prevent hardware disconnection from causing algorithm dead zones, the control module is equipped with fault identification logic. If the acquired resistance value exceeds the upper or lower limit of the analog-to-digital conversion channel, the current period's data is discarded and replaced with a valid historical value from the previous period. When temperature data from multiple consecutive sampling periods are determined to be abnormal, the control module generates a temperature acquisition fault alarm and enters a conservative protection mode. In conservative protection mode, the control module maintains the operation of the corresponding fan or outputs a temperature probe fault, fan maintenance, or heat dissipation protection alarm to the external terminal to avoid over-temperature protection failure due to temperature acquisition failure.

[0046] S104 constructs a time-series storage structure for discrete parameters. The control module stores the effective current data acquired in the current sampling period. With physical temperature data The data is stored in a first-in, first-out (FIFO) circular buffer. The automatic overwrite feature of the circular buffer is used to update stale data. As hardware timer interrupts continue to trigger, the control module generates discrete current and temperature data sequences with continuous timestamps in memory, providing a continuous stream of raw data for subsequent calculations.

[0047] In the operation of photovoltaic transformer substations, the switching of power electronic devices typically generates high-frequency electromagnetic interference. To extract low-frequency effective signals that accurately reflect the heating status of the equipment, the control module employs a first-order low-pass digital filtering algorithm to smoothly reconstruct the acquired raw discrete data. The control module uses this algorithm to smooth the raw discrete data by assigning different weights to the newly sampled data and historical smoothed data to filter out transient interference spikes and obtain steady-state time series parameters.

[0048] S105, set the filter smoothing coefficients and perform start-up boundary initialization. The control module internally presets dimensionless filter smoothing coefficients for both current and temperature. The value range of each smoothing coefficient is set to 0 to 1. In practice, the smoothing coefficients are calibrated based on the intensity of electromagnetic interference at the site: when high-frequency interference is strong, a smaller value is used to increase the weighting of historical data retention and improve filter smoothness; conversely, when the system is required to respond quickly to load changes, a larger value is used. To prevent numerical step oscillations caused by a lack of historical reference data during the initial power-on phase or sensor reset and reconnection, the control module skips weighted calculations and directly assigns the first valid raw sampled data to the internal historical filter variable register when reading it, thereby establishing the initial steady-state reference for the filtering algorithm.

[0049] S106 executes iterative filtering operations to output steady-state time series parameters. Within each sampling period, the control module performs independent filtering operations on the raw current data and raw temperature data, respectively.

[0050] For current data, the discrete iterative formula for its first-order low-pass digital filter is as follows: In the formula, This refers to the discrete time step sequence number; Calculate the steady-state current parameters for the current cycle. The raw, valid current data retrieved from the ring buffer for the current cycle (i.e., the data obtained previously). In the Discrete values ​​at each time step); The steady-state current parameters of the previous cycle are stored in the historical filter variable register; This is the current filtering smoothing coefficient.

[0051] Similarly, for temperature data, the discrete iterative formula is as follows: In the formula, The steady-state temperature parameters are calculated and output for the current cycle. The raw physical temperature data for the current period, i.e., the physical temperature data obtained above. In the Discrete values ​​at each time step; The steady-state temperature parameter for the previous cycle; This is the temperature filtering smoothing coefficient.

[0052] The control module will calculate the steady-state parameters and The data is stored in the steady-state data storage area, and the historical filter variable register is overwritten simultaneously to generate a smooth steady-state current time series and a steady-state temperature time series. Regarding the code implementation of the digital filter within the microcontroller, those skilled in the art can refer to the microcontroller instruction set for conventional configuration; this is well-known technology in the field and will not be elaborated upon here.

[0053] See attached document Figure 3 Considering the physical time delay in the conduction of internal heat to the surface temperature probe, relying solely on temperature exceeding the limit trigger could easily lead to internal hot spots. Therefore, the control module constructs a load-feedforward pre-cooling control logic for the winding fan based on current parameters with no measurement delay. By capturing transient changes in current, the control module intervenes in the impending heat accumulation before Joule heat has substantially diffused outwards.

[0054] S201, calculate the time derivative of the steady-state current sequence. After obtaining the previously generated steady-state current parameter sequence, the control module extracts the ramp slope characterizing the load change trend by performing discrete first-order difference operations.

[0055] This calculation process is equivalent to differentiating the current over continuous time. The load ramp-up slope of the steady-state current sequence is calculated according to the following formula: In the formula, Calculate the load ramp-up slope for the current moment; The steady-state current parameters for the current cycle; These are the steady-state current parameters from the previous cycle; This refers to the time interval between two adjacent steady-state parameters, i.e., the aforementioned fixed interruption period. To avoid calculation errors due to missing historical data during the initial system startup phase, the control module is equipped with first-cycle bypass logic: within the time step of acquiring the first steady-state current parameter, the load ramp-up slope of the current cycle is forcibly set. If the value is zero, the above difference formula operation will resume from the second time step.

[0056] S202, executes multi-condition start-up judgment for the winding fan. The control module has a pre-set basic load threshold. Rapidly rising threshold High load protection threshold and the starting temperature threshold of the winding fan Basic load threshold The value is determined based on the rated nameplate capacity and the factory copper loss characteristic curve of the photovoltaic dry-type transformer. It is used to characterize the current reference point at which the transformer windings begin to generate significant Joule heating, and is preferably set between 60% and 80% of the rated current. Rapid Rise Threshold The value is calibrated based on the maximum historical solar irradiance variation rate and the photovoltaic array installed capacity in the area where the power station is located. In engineering applications, in order to balance pre-cooling sensitivity and disturbance resistance, The specific value is usually set as an equivalent of a ramp rate of 5% to 15% per minute increase in rated current, and converted into amperes per second. High load protection threshold. The starting temperature threshold of the winding fan is used to characterize that the load current has entered the high load range requiring forced cooling. This is a criterion used to characterize the temperature range in which the winding temperature enters the range requiring forced convection cooling.

[0057] Within each time step, the control module generates the winding fan start-up determination result according to the following logic: In the formula, The result of the winding fan start-up determination; The steady-state current parameters for the current cycle; To participate in the basic load threshold and the high load protection threshold The absolute values ​​of the steady-state currents are compared; Calculate the load ramp-up slope for the current moment; The winding temperature at the current time step; when When true, the control module outputs a winding fan start signal.

[0058] S203 triggers a feedforward pre-cooling intervention command. When the absolute value of the steady-state current parameter is greater than the base load threshold and the load ramp-up slope is greater than the extreme ramp-up threshold, the control module determines that the photovoltaic transformer load is in a state of rapid increase. At this time, the control module bypasses the conventional temperature comparison process and directly generates a start signal for the winding fan at the bottom layer. When the absolute value of the steady-state current parameter does not meet the aforementioned slope triggering condition, but has exceeded the high load protection threshold, or the winding temperature has exceeded the winding fan start-up temperature threshold, the control module also generates a start signal for the winding fan to avoid the winding fan failing to start under the condition of slow high load ramp-up; the above logic belongs to the backup protection determination. This start signal instructs the electromechanical execution module to close the power supply circuit of the winding fan in advance, establishing a forced convection cooling air channel before the transformer body substantially heats up, completing the feedforward pre-cooling.

[0059] Furthermore, to prevent frequent start-stop dead zones caused by minor fluctuations in the load slope at the threshold critical point, the control module synchronously triggers a forced-run timer after outputting the start signal. This timer forcibly maintains the winding fan's operation for a preset hold time (e.g., three to five minutes) until the hold time expires, at which point control is then reassigned to the current load, slope, and temperature status. The forced-run timer and the forced-run timer in the subsequent contact forced-run interlocking logic are the same timer or the same state machine timing unit. For the logic branch decision program of the threshold comparator within the microcontroller, those skilled in the art can perform conventional development based on standard control logic statements. Its specific code structure is well-known in the field and will not be elaborated upon here.

[0060] See attached document Figure 4 Since the temperature rise inside the transformer substation is a slow, cumulative process, to avoid frequent start-stop of the casing fan due to instantaneous temperature exceeding limits and external environmental disturbances (such as cloud cover or gusts of wind), the control module constructs volumetric heat accumulation integral control logic for the casing fan. The control module uses the temperature rise time integral value to represent the cumulative area of ​​the effective temperature rise difference exceeding the baseline reference temperature on the time axis, using this as an equivalent indicator to measure the degree of heat accumulation in the enclosed space, thereby filtering out high-frequency temperature spikes and completing anti-shake operation.

[0061] S204 calculates the effective temperature rise difference. The control module has a preset baseline reference temperature. This base reference temperature The temperature is typically set to the optimal long-term operating temperature allowed for the transformer, preferably 30 to 35 degrees Celsius. The control module reads the aforementioned steady-state temperature parameters of the transformer volume at each discrete time step. .in, In This represents the output parameter after filtering, not the external ambient temperature of the transformer substation. The control module determines the steady-state temperature parameter of the transformer substation volume. Compared with the base reference temperature Size relationship: when Greater than At that time, calculate the difference between the two and assign it to the effective temperature rise difference. ;when Less than or equal to At that time, the effective temperature rise difference is forced. The value is zero to eliminate the interference of normal low-temperature environment on heat accumulation calculation.

[0062] S205 executes the temperature rise time-integral iterative calculation. The control module is equipped with an internal integration register, which multiplies the current effective temperature rise difference by the time step size in each time step and adds it to the historical integration result of the previous cycle.

[0063] The effective temperature rise difference is calculated according to the following formula: The integral value of the temperature rise time is calculated according to the following formula: In the formula, Calculate the temperature rise time integral value for the current time step; The historical integral value of the previous cycle is stored in the internal integral register; This represents the effective temperature rise difference at the current time step. This is the time interval between two adjacent discrete parameters, i.e., the fixed interrupt period of the aforementioned timer; These are the steady-state temperature parameters of the transformer substation volume. Based on the reference temperature.

[0064] It should be noted that, to prevent the equipment from experiencing meaningless cumulative effects due to the small positive temperature difference caused by the day-night cycle during long-term operation, which could lead to malfunctions and dead zones, the control module is equipped with a natural heat dissipation reset mechanism: when the steady-state temperature parameter of the transformer volume is detected... Less than or equal to the base reference temperature At that time, the control module not only sets the increment to zero, but also simultaneously updates the internal integration register. Force a reset to zero to simulate a real physical heat dissipation process.

[0065] S206 executes temperature rise time integral threshold determination and control. The control module has a pre-set volumetric heat accumulation trigger threshold. In engineering applications, to balance system response speed and anti-interference capability, the volumetric heat accumulation trigger threshold is... The specific value is usually set to the integral equivalent of a temperature rise exceeding the limit by 10 degrees Celsius for 300 to 600 seconds. The control module compares the currently calculated integral value of the temperature rise over time. With volumetric heat accumulation trigger threshold When the integral value of temperature rise time Greater than the volumetric heat accumulation trigger threshold At this point, the control module determines that static heat has accumulated inside the transformer and needs to be forcibly dissipated. The control module then outputs a start signal for the casing fan, instructing the electromechanical actuator module to connect the power supply circuit for the casing fan and execute ventilation and cooling.

[0066] Furthermore, at the instant the control module outputs the fan start signal, it simultaneously updates the internal integral register. Force a zeroing process and suspend the above iterative accumulation calculation during wind turbine operation until the steady-state temperature parameter of the transformer volume is reached. When the temperature substantially drops below the baseline reference temperature, the integral monitoring logic is reactivated. This mechanism helps prevent the trailing effect during shutdown determination. For the programming of numerical accumulation and register clearing operations within the microcontroller, those skilled in the art can perform conventional development based on standard C language mechanisms. The specific code structure is well-known in the field and will not be elaborated upon here.

[0067] To prevent frequent start-stop cycles (i.e., relay bouncing) of the electromechanical actuator contacts due to fluctuations in sensor data at critical states, which could lead to contact damage, the control module incorporates a forced-operation interlocking logic at the action output stage in its software. By pre-setting a minimum action sustaining window, the control module forcibly blocks reset commands caused by environmental disturbances during the initial contact closure phase, ensuring the physical validity of each actuator action.

[0068] S301, Trigger Action and Timing Initialization. In the specific implementation process, after receiving the winding fan start signal or the casing fan start signal output by the aforementioned algorithm steps, the control module immediately drives the electromechanical execution module to close the corresponding physical power supply circuit contact, and simultaneously starts the software timer bound to the channel to accumulate the current continuous running time.

[0069] S302, Execute runtime interlocking determination. The control module has a preset mandatory runtime threshold. The threshold is determined based on the physical time required for the wind turbine to establish a stable air pressure flow field and the recommended minimum continuous engagement interval of the electromechanical contactor, and is typically set between 180 and 300 seconds in engineering practice. If the accumulated continuous running time of the software timer is less than this threshold... During this phase, the control module actively blocks any shutdown commands issued by the upper-layer algorithm due to data fallback, forcibly maintaining the closed conductive state of the contacts. When the accumulated value of the software timer reaches... Afterwards, the control module enters the operation pending state and decides whether to disconnect the contact based on the shutdown conditions of the corresponding fan.

[0070] S303, release the latch and resume normal decision-making. This occurs when the continuous running time reaches or exceeds the forced running time threshold. At this time, the control module automatically releases the software interlock. The on / off state of the electromechanical actuator module will then be taken over by the upper-level trigger logic: if the transformer substation's operating condition has returned to normal, shutdown is permitted, and the timer is simultaneously reset to zero; if the transformer substation still requires cooling, the contacts remain closed. Furthermore, to prevent the timing variable from overflowing and reaching zero due to prolonged continuous operation of the fan, thus triggering a secondary false interlock logic dead zone, the control module is equipped with a timing suspension mechanism: once the continuous running time accumulates to equal a threshold... The software timer will stop accumulating and maintain the maximum value until an actual shutdown action is performed, at which point a reset will be triggered to clear the value.

[0071] It should be added that the execution priority of the aforementioned software interlocking logic is lower than that of the system's hardware-level fault protection interrupt. This ensures that in the event of electrical faults such as short circuits or severe overloads, the protection circuit can prioritize cutting off the power supply to the actuator, reducing the risk of secondary damage. Regarding the programming of the software state machine and timers at the microcontroller's underlying layer, those skilled in the art can perform conventional configurations based on standard C language logic branch statements. The specific implementation mechanism is well-known in the field and will not be elaborated upon here.

[0072] When the electromechanical actuator disconnects the load of an inductive exhaust fan motor, an electric arc and localized high temperature are generated between its physical contacts. If a closing command is received again within the brief and incomplete cooling period after the disconnection, the contacts are still not cooled and the ionized gas from the arc has not dissipated. Re-closing increases the risk of contact ablation and welding. To mitigate this risk, the control module synchronously constructs contact forced cooling isolation logic at the software level, which works in conjunction with the runtime interlock. By presetting a minimum disconnection maintenance window, the control module forcibly blocks the start request in the initial stage of contact disconnection, allowing the contacts sufficient physical cooling and insulation recovery time.

[0073] S304, execute shutdown and cooling timer. In specific implementation, after confirming that the upper-level trigger logic meets the shutdown conditions and the channel has left the aforementioned forced operation lockout phase, the control module instructs the electromechanical execution module to disconnect the corresponding physical power supply circuit contact. At the synchronous moment of contact disconnection, the control module starts the software cooling timer bound to that channel to accumulate the duration of this disconnection. In particular, to avoid dead zones caused by memory variable loss due to abnormal power failure or watchdog reset of the microcontroller, the control module will force all channel cooling timers to zero in the first running cycle of each cold start initialization and will trigger a full cycle of cooling isolation shielding by default, thereby preventing relays from momentarily malfunctioning due to power fluctuations.

[0074] S305, perform isolation and shielding determination. The control module has a preset forced cooling time threshold. The specific value of this threshold is calibrated based on the heat dissipation characteristics of the electromechanical contactor contacts and the arc extinction relaxation time, and is usually set between 60 and 120 seconds in practical engineering applications. The accumulated disconnection duration of the cooling timer must be strictly less than this threshold. During this phase, the control module forcibly intercepts and discards any restart commands issued by the upper-level algorithm, maintaining the contact in an open state. However, when the winding temperature, core temperature, or transformer volume temperature exceeds the emergency protection threshold, the control module releases the forced cooling isolation restriction, or outputs transformer load reduction, tripping, or emergency alarm signals; among these, the priority of the emergency protection logic and the hardware-level fault protection interrupt is higher than that of the forced cooling isolation logic.

[0075] In a preferred embodiment, the control module establishes a state machine for each fan control channel. The state machine includes at least a shutdown lockout state, a shutdown standby state, a forced operation state, an operation pending judgment state, and an emergency protection state. The shutdown lockout state corresponds to the stage where the output contact of the temperature controller and its corresponding contact expansion unit main contact are open and the cooling timer has not reached the forced cooling time threshold. The shutdown standby state corresponds to the stage where the output contact of the temperature controller and its corresponding contact expansion unit main contact are open and the cooling timer has reached the forced cooling time threshold. The forced operation state corresponds to the stage where the output contact of the temperature controller and its corresponding contact expansion unit main contact are closed and the operation timer has not reached the forced operation time threshold. The operation pending judgment state corresponds to the stage where the output contact of the temperature controller and its corresponding contact expansion unit main contact are closed and the operation timer has reached the forced operation time threshold. The emergency protection state corresponds to the stage where a short circuit, severe overload, severe over-temperature, or fan failure is detected.

[0076] S306, Release isolation and timer suspension. When the disconnect duration reaches or exceeds the forced cooldown time threshold. At this time, the control module automatically releases the software isolation dead zone of the channel. The underlying driver channel then regains access to the upper-layer algorithm's start signal. Furthermore, to prevent erroneous secondary isolation locking caused by the timing variable overflowing and returning to zero during prolonged shutdown, the control module has a timing suspension mechanism: once the disconnection duration reaches a threshold... The cooling timer stops accumulating and maintains its maximum value until the next contact closure action, at which point the cooling timer is synchronously reset to zero. The programming of the aforementioned shielding and interception logic, power-on initialization debouncing, and timer suspension mechanism at the microcontroller's underlying layer can be conventionally configured by those skilled in the art using standard C language logic branch statements. The specific code structure is well-known in the field and will not be elaborated upon here.

[0077] See attached document Figure 5To construct an accurate temperature prediction model, high-quality training data is required. In the actual operating environment of photovoltaic transformer substations, fluctuations in solar irradiance and electrical load are non-steady-state random disturbances. During this period, the temperature and operating condition data collected contain excessive noise and hysteresis effects. Considering the large thermal inertia of the transformer system, transient data collected under non-steady-state conditions cannot accurately reflect the true mapping relationship between the heat source and the steady-state winding temperature. Therefore, the control module automatically identifies and extracts the observation range in which the photovoltaic transformer substation operates in a deterministic steady state through a steady-state joint constraint judgment algorithm.

[0078] S401 collects and caches multi-dimensional parameter data. In practical implementation, the control module synchronously collects key parameters characterizing external energy input, internal energy dissipation, and system thermal response at a fixed sampling period. These parameters include solar irradiance, transformer load parameters, and current winding temperature parameters. Solar irradiance is collected by a solar irradiance sensor; transformer load parameters are collected by a load power acquisition unit or input from an external measurement and control device via a communication interface; and winding temperature parameters are obtained from a winding temperature probe or a nearby temperature measurement point. Load parameters include at least one of load power, apparent load, load current, or equivalent load factor. The control module internally maintains a fixed-length, first-in-first-out (FIFO) data queue of length N to cache the historical sequences of the aforementioned multi-dimensional parameters within the most recent observation window. In engineering implementation, to cover brief environmental disturbances, the sampling period is typically set to 1 to 5 seconds, and the queue length N is typically set to cover the number of data points spanning 180 to 300 seconds.

[0079] S402 evaluates the fluctuation rate of operating parameters and the thermal response rate. Specifically, to prevent division-by-zero errors or logic dead zones due to referencing invalid memory data when the microcontroller is first powered on or when the queue is not full, the control module performs a queue fullness check before executing calculations: subsequent calculations are only activated when the queue contains N valid sampling points. Under this premise, the control module uses a rolling standard deviation algorithm to quantify and calculate the fluctuation standard deviation of the solar irradiance sequence within the current observation window. Standard deviation of fluctuation of the load parameter series Simultaneously, the control module calculates the initial and final differences in the winding temperature sequence within the window, divided by the window time span, to obtain the current winding temperature change rate. The absolute value of the rate of change of the winding temperature is taken and used to determine the steady state. This rate of change is used to characterize whether the system has entered a state of thermal equilibrium.

[0080] S403, execute steady-state joint constraint determination. The control module has a preset solar irradiance stability threshold. Load parameter stability threshold and the threshold of slight change in thermal equilibrium The values ​​of these three thresholds were experimentally calibrated based on the sensor's measurement noise floor and the engineering-acceptable thermal steady-state drift, among which the thermal equilibrium micro-variation threshold... In practical applications, it is often set to the level of 0.1 degrees Celsius per minute.

[0081] The control module performs steady-state determination based on the following logical intersection formula: In the formula, This is a Boolean value for determining the steady state at the current moment; a true value indicates that the steady state condition is met. For logical AND operation; This is the absolute value of the rate of change of winding temperature, used to avoid misjudging a steady state when the winding temperature drops rapidly. The control module determines that the current moment is in a physically stable thermal equilibrium state only when the two standard deviations of the fluctuations at the input are below their respective thresholds and the absolute value of the rate of change of temperature at the output is below the small change threshold.

[0082] S404, confirm the observation interval and construct model feature samples. To avoid misjudgment due to coincidental instantaneous data, the control module is equipped with a steady-state duration counter and a minimum steady-state interval threshold. Considering the hysteresis characteristics of heat conduction in large box-type transformers, this threshold... It is usually set to 10 to 15 minutes. Only when the above joint determination results are met... The time for consecutive true values ​​exceeds Only then is this period confirmed as a valid deterministic steady-state observation interval. After this interval is confirmed, the control module performs feature extraction to construct the sample data required for subsequent model processing: extracting the average solar irradiance and average load parameters within the interval to construct the original operating condition feature data for subsequent thermal equivalent conversion, and simultaneously extracting the final winding temperature parameter at the end of the interval when thermal equilibrium is reached as the output physical label. The average load parameter is the average value of load power, apparent load, load current, or equivalent load rate within the current steady-state observation interval, and is selected according to the load type used in the subsequent thermal equivalent calculation.

[0083] The aforementioned original operating condition characteristic data, after subsequent photoelectric equivalent heat equivalent calculation, is converted into dimensionless total equivalent heat equivalent characteristics, which are then used as input features for the feedforward neural network model. The combined effective training data pairs will be stored in non-volatile memory for subsequent model training. For the calculation of standard deviation and floating-point comparison, those skilled in the art can perform conventional programming based on standard mathematical library functions; the specific implementation mechanism is well-known in the field and will not be elaborated here.

[0084] See attached document Figure 6The steady-state observation samples extracted in the aforementioned steps include average solar irradiance and average load parameters. To reduce the difficulty of nonlinear fitting introduced by multidimensional heterogeneous features, the control module converts the two energy sources, light and electricity, into thermal power of the same dimension through physical calculations, thereby achieving dimensionality reduction of input features and implanting prior thermodynamic physical constraints into the model.

[0085] S405 calculates the internal electrical loss thermal equivalent. The control module pre-sets the transformer's factory parameters in non-volatile memory and performs a non-zero check on the rated capacity parameters to prevent division by zero anomalies in the underlying calculations. Furthermore, to avoid negative overflow interference caused by sensor disconnections or ADC temperature drift, the control module performs non-negative clamping protection on the input average load parameters; that is, if its value is determined to be less than zero, it is forcibly set to zero.

[0086] When the average load parameter is the steady-state average apparent load When the equivalent load capacity is obtained from the load power acquisition unit, the specific formula for calculating the internal electrical loss heat equivalent based on the steady-state average apparent load after clamping is as follows: In the formula, This is the internal electrical loss heat equivalent, measured in watts. This refers to the preset no-load loss of the transformer. The steady-state average apparent load is extracted, or the equivalent load capacity is calculated by the load power acquisition unit. The preset rated capacity of the transformer; This refers to the pre-set rated load loss of the transformer.

[0087] Without directly obtaining the steady-state average apparent load In this implementation, the control module can calculate the equivalent load factor based on the effective value of the low-voltage side output current and the rated current, and calculate the internal electrical loss heat equivalent based on the equivalent load factor. The equivalent load factor is calculated according to the following formula: The internal electrical loss heat equivalent is calculated according to the following formula: In the formula, Equivalent load factor This is the effective value of the low-voltage side output current. This is the rated current of the transformer. For transformer no-load loss, This refers to the rated load loss of the transformer.

[0088] S406, calculate the external radiative heat equivalent. Similarly, the control module calculates the steady-state average solar irradiance. After performing the non-negative clamping check, it is equivalent to the heat source power applied to the outside of the transformer. The calculation process is as follows: In the formula, This is the external radiative heat equivalent, measured in watts. The extracted steady-state average solar irradiance; The effective light-receiving area of ​​a photovoltaic transformer is usually taken as half of the sum of the top area and the area of ​​the light-receiving sidewall in engineering specifications. The comprehensive absorption coefficient of the enclosure surface is typically between 0.55 and 0.75, depending on the material of the anti-corrosion coating.

[0089] S407 performs thermal equivalent fusion and normalization processing. The control module uses the internal electrical loss thermal equivalent calculated above. Equivalent to external radiation heat Algebraic summation is performed to obtain the total equivalent heat generation power for the steady-state range. To avoid abnormal gradient updates in the neural network due to directly inputting large values, the control module normalizes this total power by dividing it by a preset system limit heat generation benchmark value.

[0090] The normalized total equivalent heat capacity characteristic is calculated according to the following formula: In the formula, It is a dimensionless total equivalent heat characteristic. This is the system's limiting heating baseline value. In implementations that only use the dimensionless total equivalent heat equivalent characteristic as model input, this dimensionless total equivalent heat equivalent characteristic... This serves as the single-node input feature for the subsequent feedforward neural network model's input layer. The baseline value is set as the sum of the transformer's rated total loss and the equivalent of peak summer radiative heat. After the above processing, the original two-dimensional electrical and environmental conditions are reduced in dimension to a dimensionless one-dimensional input feature.

[0091] In some alternative implementations, the input feature is specifically designated as a single-node feature of the input layer of a subsequent feedforward neural network (or multilayer perceptron), while the corresponding steady-state winding temperature is used as the physical target label of the output layer. Together, they constitute a supervised learning sample pair for model weight updates.

[0092] In another optional implementation, the input features of the feedforward neural network model further include at least one of the following: steady-state temperature of the transformer substation volume, ambient temperature, or initial historical winding temperature, to improve the predictive adaptability of the reference winding temperature prediction model under different ambient temperatures or different initial thermal states. As for the underlying floating-point multiply-accumulate operations and register access mechanisms of the microcontroller, those skilled in the art can develop them based on conventional microcontroller instruction sets; these are well-known technologies in the field and will not be elaborated upon here.

[0093] See attached document Figure 7During long-term outdoor operation, photovoltaic transformer substations face physical degradation issues such as dust filter clogging and fan bearing aging, leading to a gradual decline in the overall system's heat dissipation capacity. Based on the characteristic that heat dissipation degradation will cause temperature deviation under the same heat input conditions, the control module calculates the difference between the actual winding temperature and the predicted reference winding temperature under initial healthy conditions to quantify the degradation characteristics. The degradation assessment method based on the reference winding temperature prediction model in this embodiment can be used as an alternative to the method in S40 mentioned above, which calculates the first heat dissipation degradation index based on the forced operating range. It can also be used in conjunction with the first heat dissipation degradation index to improve the reliability of judging filter clogging or fan performance degradation.

[0094] To enable early warning of such gradual faults, the control module establishes a reference winding temperature prediction model based on historical data from the initial stage of operation, and uses the deviation between this model and real-time measurements to quantitatively assess the heat dissipation degradation trend.

[0095] S408 constructs and trains a reference winding temperature prediction model. The control module integrates a feedforward neural network model, preferably a three-layer multilayer perceptron architecture.

[0096] In another implementation, the feedforward neural network model can be trained by an external terminal or a higher-level monitoring station based on initial commissioning data, and the trained model weight parameters are sent to the control module. The control module only performs forward propagation prediction calculations. The specific network hierarchy and connections include: an input layer, used to receive the dimensionless total equivalent thermal equivalent features output from the preceding steps, and fully connected to the hidden layer; in implementations where the model input only uses the dimensionless total equivalent thermal equivalent features, the input layer is a single-node input layer; in implementations where the model input also includes at least one of the following: the steady-state temperature of the transformer substation volume, the external ambient temperature, or the initial value of the historical winding temperature, the input layer is a multi-node input layer; a fully connected hidden layer, typically with 3 to 10 neurons in engineering implementations, using the ReLU linear rectified function as the activation function; and a single-node output layer, fully connected to the hidden layer, used to output the predicted reference winding temperature under the corresponding operating condition.

[0097] During the model training phase, the control module collects steady-state datasets from the initial commissioning phase of the photovoltaic transformer (typically the first three months of healthy operation). Preprocessed equivalent thermal features are extracted as input training samples, and the actual winding temperatures measured during the same period are used as output labels for supervised learning. The control module uses mean squared error as the loss function and employs backpropagation for iterative weight training until the maximum number of iterations is reached or the loss function value falls below a preset lower limit, thereby learning and solidifying the baseline thermal response characteristics of the equipment under healthy conditions. If the model has not yet completed training or the number of effective training samples is below a preset sample count threshold, the control module does not activate the baseline winding temperature prediction model but instead uses the aforementioned first heat dissipation degradation index based on the forced operating range for filter status assessment.

[0098] S409 performs real-time winding temperature prediction and temperature residual extraction. During routine monitoring, once a valid steady-state observation interval is confirmed and the current thermal equivalent characteristics are calculated, the control module inputs it into the trained and fixed prediction model. Through forward propagation, the predicted reference winding temperature under ideal healthy conditions is obtained. To prevent the model from outputting results that violate physical principles due to generalization errors when encountering extreme inputs, the control module has a physical lower limit interception mechanism after the output node: if the calculated predicted reference winding temperature is lower than the preset physical lower limit, it is forcibly corrected to the preset physical lower limit.

[0099] The control module synchronously reads the actual winding temperature at the end of the current steady-state range and calculates the temperature residual according to the following formula: In the formula, This represents the temperature residual in the current steady-state range. This represents the steady-state winding temperature actually collected by the temperature sensor. This is the predicted baseline winding temperature output by the neural network model. This residual value reflects the deviation between the current actual heat dissipation capacity and the initial healthy state.

[0100] S410 calculates the heat dissipation degradation index and performs trend analysis. To smooth out the interference of single measurement noise, the control module calculates the arithmetic mean of all effective temperature residuals within the most recent statistical period (usually set to seven to fifteen days) and uses this as the second heat dissipation degradation index for trend assessment. The second heat dissipation degradation index is calculated according to the following formula: In the formula, The second heat dissipation degradation index, This represents the number of valid steady-state samples within the current statistical period. For the first The temperature residual corresponding to each effective steady-state sample.

[0101] In particular, to avoid division-by-zero errors caused by the inability to extract effective steady-state samples under special operating conditions such as continuous rain, the control module performs a sample size check before calculating the mean: if the number of effective residual samples in the current period is less than the preset minimum sample threshold (e.g., 3), the control module skips the current calculation and directly maintains the exponential value of the previous period.

[0102] The control module has a preset degradation alarm threshold, which is calibrated based on the transformer insulation withstand margin and is typically set between 5 and 8 degrees Celsius. The control module executes the following logic: if the current second heat dissipation degradation index exceeds the set degradation alarm threshold multiple times consecutively, it confirms that the system has experienced substantial heat dissipation capacity degradation and sends an alarm message to the superior monitoring station requesting dust filter cleaning or fan performance maintenance. When the system simultaneously uses both the first and second heat dissipation degradation indices, the control module can output a warning when either exceeds its corresponding threshold, or a confirmation alarm when both exceed their respective thresholds. For the underlying matrix operations of the neural network forward propagation, those skilled in the art can implement it using conventional C language arrays and loop structures; the specific code organization is well-known in the field and will not be elaborated upon here.

[0103] Specific application examples: The following describes the specific application examples and experimental verification results of the aforementioned control logic, using specific numerical values ​​and corresponding diagrams in the instruction manual.

[0104] See attached document Figure 8 To facilitate a comprehensive understanding of the multidimensional control features of this invention, the operating data and comparison curves of this embodiment and the verification section are uniformly integrated in the appendix. Figure 8 The diagram contains multiple sub-diagrams reflecting the logical characteristics of different operating conditions.

[0105] This embodiment sets the target as a unit with a rated capacity. In this embodiment of the photovoltaic dry-type transformer, the collected load current is the effective value of the low-voltage side output current; therefore, the transformer's rated current is... Specifically, the rated current on the low-voltage side is taken as follows: The transformer's factory thermal parameters are as follows: no-load loss Rated load loss The timer interrupt period of the control module is set to... .

[0106] Taking the rapid increase in photovoltaic inverter output current as an example, the execution process of the wind turbine feedforward precooling control is explained. The control module presets a basic load threshold. The threshold is rising rapidly. The starting temperature threshold for traditional winding fans is set to 7.5 A / s. The control module continuously samples and digitally filters the data to obtain the current steady-state current. The current load ramp-up slope is calculated. .like Figure 8 The upper part of the diagram represents the load ramping condition. At this time, the transformer winding temperature obtained by the sensor acquisition module is only 65℃, which is lower than the start-up temperature threshold of 85℃. The control module determines that the steady-state current is greater than the base load threshold and the load ramping slope is greater than the extreme ramping threshold. At the corresponding timing node, it generates the step fan start signal shown in the diagram, closes the power supply contact, and establishes forced convection in advance.

[0107] This section explains the execution process of integral control for heat accumulation in the casing fan volume, addressing the short-term temperature fluctuations caused by changes in ambient wind speed. The control module presets a base reference temperature. Volumetric heat accumulation trigger threshold Set to 3000℃ When the temperature of the transformer substation rises to 42℃ due to external disturbance and remains there for 40 seconds, the control module calculates the effective temperature rise difference. The current temperature rise time integral cumulative value is Since the accumulated value is less than the set volumetric heat accumulation trigger threshold, the control module maintains the contact disconnected state and does not output a casing fan start signal, thus avoiding frequent fan start-stop caused by short-term temperature fluctuations.

[0108] The process of evaluating the heat dissipation degradation index is explained under the condition of gradual filter clogging. Within a deterministic steady-state observation interval that satisfies the joint constraints, the control module extracts the average apparent load of 2500 kVA and the average solar irradiance of 800 W / m² within the steady-state interval. 2 The calculated internal electrical loss heat equivalent is approximately 18.2 kW, and the external radiative heat equivalent is 4.8 kW (assuming an effective light-receiving area of ​​10 m²). 2 (Absorption coefficient 0.6). The control module inputs the normalized total equivalent heat equivalent characteristics into the pre-trained reference winding temperature prediction model, and the model outputs a predicted reference winding temperature of 82℃. If the current actual winding temperature is 89℃, the single temperature residual is 7℃. Figure 8 The lower half shows the subplot for assessing heat dissipation degradation. The plot displays the scatter plot distribution of temperature residuals and the moving average curve over a long period. After statistical period calculation, the second heat dissipation degradation index stabilizes at 6.8℃, which is greater than the preset maintenance limit of 5℃. The system then generates a filter cleaning command.

[0109] To verify the actual effect of the control strategy in the embodiments of the present invention, a control experiment was conducted using two identical 3150kVA photovoltaic transformer substations. The experimental data were collected, processed, and plotted synchronously. Figure 8In this embodiment, transformer substation No. 1 adopts a conventional ventilation control strategy based on a fixed temperature threshold, while transformer substation No. 2 adopts the ventilation control method of this invention.

[0110] The test was first conducted under a load ramp-up condition where the load rate increased from 20% to full load within 3 minutes. Combined with... Figure 8 As can be seen from the dynamic temperature comparison tracking curves in the upper part, the winding temperature of transformer No. 1 exhibits a significant response delay, with the highest temperature overshooting to 96℃ during the test. In contrast, transformer No. 2, when its winding temperature reaches 70℃, triggers a feedforward action due to meeting the slope threshold, effectively limiting the highest temperature during the test to 83℃. The comparison results indicate that the maximum dynamic temperature rise of transformer No. 2 is effectively reduced.

[0111] Secondly, under cloudy weather conditions with significant fluctuations in solar irradiance and load power, the frequency of operation of the temperature control contacts of the two transformer substations was statistically analyzed. Test data showed that substation No. 1, lacking state machine anti-jitter and integral logic, experienced 47 operations per day; substation No. 2, based on volumetric heat accumulation integral logic and contact timing interlocking logic, reduced its daily operation frequency to 12. The tests demonstrate that this solution significantly reduces the frequency of contact operations and the risk of contact wear.

[0112] Finally, damping material was installed at the air inlet of the transformer substation to simulate the gradual clogging process of the dust filter, and long-term testing was conducted. Combined with... Figure 8 The lower half of the data, showing the heat dissipation degradation index and experimental comparison, reveals that the No. 1 transformer substation triggered an over-temperature trip on the 52nd day of testing due to the actual winding temperature reaching 105℃. For the No. 2 transformer substation, based on the calculated moving average curve of the second heat dissipation degradation index, the curve exceeded the 5℃ maintenance limit reference line on the 25th day of testing, issuing a filter cleaning warning 27 days in advance. The experimental results demonstrate that the method of this invention can effectively eliminate environmental variable interference and accurately quantify and provide early warning of the declining trend in the system's internal physical heat dissipation capacity.

[0113] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A photovoltaic transformer substation ventilation system with extended capacity temperature control protection, characterized in that, include: The sensing and acquisition module is used to acquire current and temperature data of the photovoltaic transformer box, wherein the temperature data includes at least the winding temperature and the box volume temperature. The control module is connected to the sensing and acquisition module; The execution module, connected to the control module, includes at least a winding fan and a casing fan; The control module is used for: The load ramp-up slope is obtained by calculating the time derivative of the current data; When the absolute value of the current data is greater than the preset basic load threshold and the load ramping slope is greater than the preset extreme ramping threshold, a winding fan start signal is generated and output before the winding temperature reaches the preset winding fan start temperature threshold, so as to establish forced convection in advance and improve the dynamic overload capacity of the photovoltaic transformer. Calculate the effective temperature rise difference between the transformer substation volume temperature and the preset base reference temperature, and perform cumulative integration on the effective temperature rise difference to obtain the temperature rise time integral value. Based on the integral value of the temperature rise time, a start-up determination is made for the casing fan and a start-up signal for the casing fan is output; The actuator is controlled to drive the corresponding fan according to the winding fan start signal or the housing fan start signal.

2. The photovoltaic transformer ventilation system with expanded capacity temperature control protection according to claim 1, characterized in that, The control module is also used to perform digital filtering on the current data and the temperature data to filter out high-frequency interference signals and generate time series parameters, wherein the time series parameters include at least a steady-state current time series and a steady-state temperature time series. The control module is specifically used for: Steady-state parameters are extracted from the steady-state current time series and the steady-state temperature time series and used as the current data and temperature data, respectively, to participate in various threshold determinations and slope calculations.

3. A photovoltaic transformer substation ventilation system with expanded capacity temperature control protection according to claim 1, characterized in that, The control module is also used to perform backup protection determination: If the absolute value of the current data is not greater than the basic load threshold and the load ramping slope is not greater than the extreme ramping threshold at the same time, the winding fan start signal is generated and output if the absolute value of the current data is greater than the preset high load protection threshold, or if the winding temperature is greater than the winding fan start temperature threshold.

4. A photovoltaic transformer substation ventilation system with expanded capacity temperature control protection according to claim 1, characterized in that, The specific logic for the control module to perform the start-up determination for the casing fan based on the integral value of the temperature rise time is as follows: When the integral value of the temperature rise time reaches the preset volumetric heat accumulation trigger threshold, the casing fan start signal is output. When the temperature of the transformer substation is detected to be less than or equal to the base reference temperature, the effective temperature rise difference is set to zero, and the temperature rise time integral value is simultaneously cleared to zero.

5. A photovoltaic transformer substation ventilation system with expanded capacity temperature control protection according to claim 1, characterized in that, The execution module also includes a temperature controller and a contact expansion unit connected to the control module. The output contacts of the temperature controller are connected to the control terminal of the contact expansion unit, and the main contacts of the contact expansion unit are connected in series with the power supply circuit of the corresponding fan. After generating the corresponding fan start signal, the control module controls the temperature controller to close the corresponding output contact, thereby driving the contact expansion unit to close the corresponding main contact, and starts the forced run timer. During the activation interval of the forced run timer, the shutdown logic is suspended to keep the output contact and its corresponding main contact closed. When the accumulated value of the forced operation timer reaches the preset forced operation time threshold, and the shutdown condition of the fan corresponding to the main contact is met, the control module controls the temperature controller to disconnect the output contact, thereby driving the contact expansion unit to disconnect the corresponding main contact and start the forced cooling timer. During the activation interval of the forced cooling timer, the start request is suspended to keep the output contact and its corresponding main contact disconnected. Wherein, when the corresponding fan is the winding fan, the shutdown conditions include at least the winding temperature being lower than a preset winding shutdown temperature threshold, the steady-state current generated after the current data is digitally filtered being lower than a preset shutdown load threshold, and the load ramping slope being lower than a preset reset slope threshold. When the corresponding fan is the casing fan, the shutdown conditions include at least the transformer substation volume temperature being lower than the preset casing fan shutdown temperature threshold or lower than the base reference temperature.

6. A photovoltaic transformer substation ventilation system with expanded capacity temperature control protection according to claim 5, characterized in that, The control module establishes a state machine for each wind turbine control channel. The state machine includes at least the shutdown lockout state, the shutdown standby state, the forced operation state, the operation pending judgment state, and the emergency protection state. The emergency protection state corresponds to the stage when a short circuit, severe overload, severe overheating, or fan failure is detected. When the winding temperature or the transformer substation volume temperature exceeds a preset emergency protection threshold, the control module releases the forced cooling isolation restriction constructed by the forced cooling timer.

7. A photovoltaic transformer substation ventilation system with expanded capacity temperature control protection according to claim 5, characterized in that, The control module is also used to intercept the current data and the transformer volume temperature within the activation interval of the forced operation timer corresponding to the casing fan; The weighted sum of the square integral of the current and the integral of the first term of the current within the interval is used to obtain the comprehensive thermodynamic input equivalent, and the effective cooling range between the temperature at the beginning and the end of the interval is obtained. When the effective cooling rate is less than the preset minimum temperature difference, the preset minimum temperature difference is used in subsequent calculations. The first heat dissipation degradation index is calculated based on the comprehensive thermodynamic input equivalent and the effective cooling range. When the first heat dissipation degradation index of multiple cycles exceeds the set maintenance limit after sliding average, a filter cleaning command is output to the external terminal.

8. A photovoltaic transformer substation ventilation system with expanded capacity temperature control protection according to claim 1, characterized in that, The sensing and acquisition module also includes a solar irradiance sensor for acquiring external energy input and a load power acquisition unit for acquiring internal energy dissipation. The control module acquires the solar irradiance sequence, load parameter sequence, and winding temperature sequence generated based on the winding temperature within a fixed sampling period, and calculates the winding temperature change rate of the winding temperature sequence. A deterministic steady-state observation interval is confirmed only when the standard deviation of the fluctuation of the solar irradiance sequence is lower than a preset solar irradiance stability threshold, the standard deviation of the fluctuation of the load parameter sequence is lower than a preset load parameter stability threshold, the absolute value of the winding temperature change rate is lower than a preset thermal balance micro-change threshold, and the above conditions are met for a period of time exceeding a preset shortest steady-state interval threshold.

9. A photovoltaic transformer substation ventilation system with expanded capacity temperature control protection according to claim 8, characterized in that, After extracting the deterministic steady-state observation interval, the control module calculates the average load parameters within the deterministic steady-state observation interval, and calculates the internal electrical loss thermal equivalent by combining the transformer no-load loss and rated load loss. Calculate the average solar irradiance within the deterministic steady-state observation interval, and calculate the external radiative heat equivalent by combining the effective light-receiving area of ​​the photovoltaic transformer and the comprehensive absorption coefficient of the transformer surface; The internal electrical loss heat equivalent and the external radiation heat equivalent are algebraically added together and then divided by a preset system limit heating benchmark value for normalization, generating a dimensionless total equivalent heat equivalent feature.

10. A photovoltaic transformer substation ventilation system with expanded capacity temperature control protection according to claim 9, characterized in that, The control module integrates a trained and fixed feedforward neural network model. The control module inputs the dimensionless total equivalent thermal equivalent feature into the feedforward neural network model and outputs the predicted reference winding temperature. The winding temperature at the end of the deterministic steady-state observation interval is calculated as the actual winding temperature, and the difference between the actual winding temperature and the predicted reference winding temperature is calculated to generate a temperature residual. The arithmetic mean of the temperature residuals within the statistical period is calculated to obtain the second heat dissipation decay index. When the second heat dissipation decay index exceeds the preset decay alarm threshold, an alarm message is output.

Images