A transformer power distribution system
By integrating data acquisition, prediction, and compensation units into the transformer power distribution system and combining them with a neural network model, the reactive power compensation strategy is dynamically adjusted. This solves the problem that traditional systems cannot predict voltage changes, enables flexible responses to distributed photovoltaic power fluctuations, avoids voltage over-limits and capacitor oscillations, and improves the stability and safety of the system.
Patent Information
- Application Number
- CN202611180290.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-05
- Publication Date
- 2026-11-17
AI Technical Summary
Traditional transformer power distribution systems cannot predict voltage change trends and are difficult to adapt to the drastic fluctuations in distributed photovoltaic power, resulting in short-term over-limit voltage of the distribution bus or frequent switching oscillations of capacitors.
The system employs a data acquisition unit, a rate of change calculation module, a prediction unit, a decision-making unit, a hybrid compensation unit, a threshold adaptive module, and a constraint verification module. By combining a neural network model and the voltage change rate, it dynamically adjusts the reactive power compensation strategy, predicts the risk of voltage exceeding limits, and optimizes the switching dead zone.
This effectively avoids short-term voltage overruns caused by photovoltaic power fluctuations, reduces frequent capacitor switching oscillations, and improves the reliability and operational safety of system control.
Smart Images

Figure CN123121677A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power distribution systems, and more particularly to a transformer power distribution system. Background Technology
[0002] With the large-scale grid connection of distributed photovoltaic power in urban and rural distribution areas and industrial and commercial parks, photovoltaic power output exhibits significant intermittent and fluctuating characteristics. Traditional transformer distribution systems generally adopt a passive reactive power compensation control method with fixed voltage thresholds and fixed switching dead zones.
[0003] However, traditional passive control methods have a slow response, cannot predict voltage change trends, are difficult to adapt to the drastic fluctuations in distributed photovoltaic power, and are prone to short-term voltage overruns on the distribution bus (such as overvoltage caused by a surge in photovoltaic power generation at noon), or frequent switching oscillations of capacitors due to improper dead zone settings. Summary of the Invention
[0004] The present invention aims to provide a transformer power distribution system to solve the problems mentioned in the background art.
[0005] To achieve the above objectives, the present invention provides the following technical solution: A transformer power distribution system, comprising: The data acquisition unit is used to collect the three-phase voltage value, three-phase load current value, and real-time power generation of the distributed photovoltaic system on the low-voltage side of the distribution transformer in real time. The rate of change calculation module, whose input terminal is connected to the data acquisition unit, is used to calculate the voltage rate of change; The prediction unit, whose input terminals are connected to the data acquisition unit and the rate of change calculation module respectively, has a built-in neural network model to predict the risk of voltage exceeding the limit within a future time window based on the current voltage value, voltage change rate and photovoltaic power generation change rate, and calculate the estimated value of reactive power compensation capacity in combination with voltage deviation. The decision unit, whose input is connected to the prediction unit, is used to generate instructions based on the estimated reactive power compensation capacity, the instructions including capacitor connection instructions or capacitor disconnection instructions. The hybrid compensation unit includes a thyristor-switched capacitor and a static var generator; A threshold adaptive module, whose input is connected to the data acquisition unit, is used to dynamically adjust the switching dead zone of the hybrid compensation unit according to the photovoltaic power fluctuation rate. The switching dead zone refers to the voltage range that does not trigger any capacitor switching operation. The voltage range is narrowed when the photovoltaic power fluctuation rate increases to improve sensitivity, and widened when the photovoltaic power fluctuation rate decreases to avoid oscillation. A constraint verification module, whose input is connected to the decision unit, whose output is connected to the hybrid compensation unit, and whose verification input is connected to the rate of change calculation module, is used to verify the command based on the voltage change trend. If the current voltage is in a decreasing state, the capacitor connection command is rejected; If the current voltage is rising, the capacitor disconnect command is rejected; If the current voltage is rising and the received instruction is a capacitor connection instruction, or if the current voltage is falling and the received instruction is a capacitor disconnection instruction, then the instruction is allowed to be sent to the hybrid compensation unit. The system is configured such that when the prediction unit predicts that the voltage will exceed the upper limit threshold, the decision unit issues an instruction, which is verified by the constraint verification module, and the hybrid compensation unit performs hierarchical compensation according to the switching dead zone adjusted by the threshold adaptive module.
[0006] Preferably, the rate of change calculation module calculates the voltage change rate according to the following formula:
[0007] in, This is the voltage value at the current moment. This is the voltage value at the previous sampling time. The sampling interval; the sampling interval It takes 100ms.
[0008] Preferably, the future time window in the prediction unit is 1 to 5 seconds.
[0009] Preferably, the neural network model in the prediction unit is a BP neural network, whose input layer nodes include: current voltage value, voltage change rate, photovoltaic power generation, and photovoltaic power generation change rate; and the output layer node is: the estimated value of the required reactive power compensation capacity.
[0010] Preferably, the upper limit threshold is 235V to 242V.
[0011] Preferably, the threshold adaptive module adjusts the switching dead zone, which is adjustable from 0.5V to 5V; when the photovoltaic power fluctuation rate is greater than 10% / min, the dead zone is narrowed to 0.5V-1V; when the photovoltaic power fluctuation rate is less than 2% / min, the dead zone is widened to 3V-5V.
[0012] Preferably, when the hybrid compensation unit performs graded compensation: the thyristor-switched capacitor is used to undertake large-step reactive power compensation, with a single input capacity not less than 60% of the system's reactive power deficit; the static var generator is used to undertake the remaining small-step reactive power compensation, achieving stepless smooth adjustment.
[0013] Preferably, the system further includes a state recognition module, which is connected to the data acquisition unit and the prediction unit respectively. The state recognition module identifies three operating conditions—pure grid power supply, photovoltaic grid connection, and sudden load change—based on the acquired three-phase voltage values, three-phase load current values, and photovoltaic power generation. It then issues an operating condition label to the prediction unit, which switches the corresponding neural network weight parameters according to the different operating conditions.
[0014] Preferably, the system further includes a fault alarm module, which is connected to the constraint verification module and the hybrid compensation unit respectively. When the constraint verification module rejects the decision unit instruction three times in a row, or the hybrid compensation unit fails to switch, the fault alarm module generates an alarm signal and uploads it to the background operation and maintenance terminal.
[0015] Preferably, the hybrid compensation unit is connected in parallel with the low-voltage side bus of the distribution transformer, and the thyristor-switched capacitor and the static var generator are arranged independently, sharing the same set of current and voltage sampling circuits.
[0016] The beneficial effects of this technical solution compared to existing technologies are as follows: (1) This scheme uses the neural network model built into the prediction unit to predict the risk of voltage exceeding the limit in the future time window by combining the current voltage value, voltage change rate and photovoltaic power change rate. Before the voltage exceeds the limit, the decision unit issues a compensation instruction in advance, which solves the problem of the traditional passive control response being lagging and unable to predict the voltage change trend. It effectively avoids the short-term voltage exceeding the limit of the distribution bus caused by the violent fluctuation of photovoltaic power, such as the overvoltage phenomenon caused by the large photovoltaic power generation at noon.
[0017] (2) The threshold adaptive module dynamically adjusts the switching dead zone width according to the photovoltaic power fluctuation rate: when the fluctuation rate increases, the dead zone is narrowed to improve the response sensitivity; when the fluctuation rate decreases, the dead zone is widened to avoid invalid switching actions. This solves the problem of frequent switching oscillation of capacitors caused by unreasonable fixed dead zone settings and extends the service life of capacitors and switching equipment.
[0018] (3) By using the constraint verification module, the decision instructions are physically verified in combination with the voltage rise and fall trend. Instructions that contradict the voltage change trend are rejected, such as incorrect instructions like putting in the capacitor when the voltage drops or cutting off the capacitor when the voltage rises. This avoids the malfunction problem that is easy to occur in pure artificial intelligence control and further improves the reliability and operational safety of the system control. Attached Figure Description
[0019] Figure 1 The system framework diagram provided for this invention. Detailed Implementation
[0020] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments: like Figure 1 The transformer power distribution system shown includes: The data acquisition unit is used to collect the three-phase voltage values, three-phase load current values, and real-time power generation of the distributed photovoltaic system on the low-voltage side of the distribution transformer in real time. In this embodiment, the data acquisition unit is integrated into the transformer substation intelligent fusion terminal (TTU). The TTU has a built-in analog signal acquisition module with a sampling accuracy of not less than 0.5 and a fixed sampling frequency of 10kHz. After digital filtering and downsampling processing, it outputs valid data with a period of 100ms to the rate of change calculation module to meet the real-time control requirements of the system.
[0021] The rate of change calculation module, whose input terminal is connected to the data acquisition unit, is used to calculate the voltage rate of change; The rate of change calculation module calculates the voltage change rate according to the following formula:
[0022] in, This is the voltage value at the current moment. This is the voltage value at the previous sampling time. Sampling interval; Sampling interval It takes 100ms.
[0023] The prediction unit, whose input terminals are connected to the data acquisition unit and the rate of change calculation module respectively, has a built-in neural network model. It is used to predict the risk of voltage exceeding the limit within a future time window based on the current voltage value, voltage change rate and photovoltaic power generation change rate, and calculate the estimated reactive power compensation capacity in combination with the voltage deviation. The photovoltaic power generation change rate is used as the instantaneous change input to the neural network.
[0024] The specific method by which the prediction unit predicts voltage over-limit risk is as follows: (1) The voltage prediction value at the end of the future time window is output by the neural network. Then Compared with the upper threshold (238V), if If the voltage is V, it is determined that there is a risk of voltage exceeding the limit.
[0025] (2) The calculation method for the estimated reactive power compensation capacity is as follows: The voltage prediction value determined in step (1) With upper limit threshold Calculate the difference: Then multiply by a scaling factor :
[0026] in, kvar / V. This calculation method is based on the physical law that "the required compensation is proportional to the predicted voltage deviation".
[0027] (3) Execution cycle: The above process is executed once every 100ms, and the execution time of a single execution is 5ms.
[0028] The rate of change of photovoltaic power generation is calculated according to the following formula:
[0029] in, This represents the photovoltaic power generation value at the current moment, in kilowatts (kW). The value represents the photovoltaic power generation at the previous sampling time, in kilowatts (kW). The sampling interval is 100 milliseconds (ms), which is equivalent to 0.1 seconds (s).
[0030] The future time window in the prediction unit is 1 to 5 seconds. This time window can be dynamically adjusted according to the system inertia of the low-voltage side of the distribution transformer: for urban transformer substations with shorter power supply lines, it can be set to 1 to 2 seconds; for rural transformer substations with longer power supply lines, it can be set to 3 to 5 seconds, to adapt to the response characteristics of different systems.
[0031] The neural network model in the prediction unit is a backpropagation (BP) neural network. Its input layer nodes include: current voltage value, voltage change rate, photovoltaic power generation, and photovoltaic power generation change rate; the output layer node is: the estimated required reactive power compensation capacity. This BP neural network has a three-layer structure: the input layer has four nodes corresponding to the current voltage value, voltage change rate, photovoltaic power generation, and photovoltaic power generation change rate; the hidden layer has ten nodes with the sigmoid function as the activation function; and the output layer has one node, outputting the predicted voltage value at the end of the future time window. A total of 62 weight parameters were obtained through offline training. The training samples were historical operating data from 30 consecutive days, and the training algorithm was the Levenberg-Marquardt algorithm.
[0032] The decision unit, whose input is connected to the prediction unit, is used to generate instructions based on the estimated reactive power compensation capacity. These instructions include capacitor connection instructions or capacitor disconnection instructions. The decision unit has a set of quantitative thresholds: when the estimated reactive power compensation capacity is greater than 10% of the system's reactive power deficit, a capacitor connection instruction is generated; when the estimated reactive power compensation capacity is less than -10% of the system's reactive power deficit, a capacitor disconnection instruction is generated; when it is between ±10%, no instruction is generated to maintain the current state.
[0033] The hybrid compensation unit includes a thyristor-switched capacitor and a static var generator; The threshold adaptive module, whose input is connected to the data acquisition unit, is used to dynamically adjust the switching dead zone of the hybrid compensation unit according to the photovoltaic power fluctuation rate. The switching dead zone refers to the voltage range that does not trigger any capacitor switching operation. This voltage range is narrowed when the photovoltaic power fluctuation rate increases to improve sensitivity, and widened when the photovoltaic power fluctuation rate decreases to avoid oscillation. The photovoltaic power fluctuation rate is used as a minute-level statistic for adjusting the dead zone.
[0034] The threshold adaptive module adjusts the switching dead zone, which is adjustable from 0.5V to 5V. When the photovoltaic power fluctuation rate is greater than 10% / min, the dead zone narrows to 0.5V-1V; when the photovoltaic power fluctuation rate is less than 2% / min, the dead zone widens to 3V-5V. At system startup, the initial value of the switching dead zone is set to 2V. When the fluctuation rate is between 2% / min and 10% / min, the switching dead zone remains unchanged from the previous adjustment value.
[0035] The photovoltaic power volatility is calculated as follows: within a minute-by-minute time window, the difference between the maximum and minimum photovoltaic power generation is calculated, divided by the average power within that time window, and expressed as a percentage. The volatility calculation period is synchronized with the switching dead zone adjustment period, both being 1 minute.
[0036] The constraint verification module has its input connected to the decision unit, its output connected to the hybrid compensation unit, and its verification input connected to the rate of change calculation module; it is used to verify commands based on voltage change trends. If the current voltage is in a decreasing state, the capacitor connection command is rejected; If the current voltage is rising, the capacitor disconnect command is rejected; If the current voltage is rising and the received instruction is a capacitor connection instruction, or if the current voltage is falling and the received instruction is a capacitor disconnection instruction, then the instruction is allowed to be sent to the hybrid compensation unit. The essence of this verification rule is that the expected effect of the instruction should be consistent with the current voltage change trend. Specifically, the expected effect of a capacitor connection instruction is to reduce the voltage; therefore, connecting the capacitor when the voltage rises is reasonable, while connecting it when the voltage falls is unreasonable. Conversely, the expected effect of a capacitor disconnection instruction is to increase the voltage; therefore, disconnecting the capacitor when the voltage falls is reasonable, while disconnecting it when the voltage rises is unreasonable. The constraint verification module achieves rapid verification of the physical rules by comparing the sign of the voltage change rate with the instruction type, with a processing time of no more than 5ms for a single verification cycle.
[0037] The system is configured such that when the prediction unit predicts the voltage will exceed the upper limit threshold, the decision unit issues a command. After verification by the constraint verification module, the hybrid compensation unit performs tiered compensation based on the switching dead zone adjusted by the threshold adaptive module. The upper limit threshold is 235V to 242V. This upper limit threshold can be set according to the voltage level and power quality requirements of the distribution network: for a 220V nominal voltage system, the upper limit threshold is typically set to 235V to 242V; for a 380V nominal voltage system, it is multiplied by √3, i.e., 407V to 419V. This system allows maintenance personnel to remotely modify this upper limit threshold via a backend terminal.
[0038] When the hybrid compensation unit performs tiered compensation: thyristor-switched capacitors (TSCs) are used to handle large-step reactive power compensation, with a single input capacity not less than 60% of the system's reactive power deficit; static var generators (SVGs) are used to handle the remaining small-step reactive power compensation, achieving stepless smooth adjustment. The strategy of using a hybrid compensation of TSCs and SVGs combines economy and speed: TSCs have lower costs and are suitable for compensating for large fixed reactive power deficits, but their response speed is slower (approximately 20ms to 40ms); SVGs have a fast response speed (approximately 5ms to 10ms), but their cost is higher. This system combines the two, with TSCs handling the main reactive power deficit and SVGs for fine-tuning, reducing system costs while ensuring dynamic response performance.
[0039] The system also includes a status identification module, which is connected to both the data acquisition unit and the prediction unit. Based on the acquired three-phase voltage values, three-phase load current values, and photovoltaic power generation, the status identification module identifies three operating conditions: pure grid supply, photovoltaic grid connection, and sudden load change. It then sends an operating condition flag to the prediction unit, which switches the corresponding neural network weight parameters according to the different operating conditions. The operating condition identification rules are as follows: when the photovoltaic power generation is less than 5% of the rated power and the load current change rate is less than 10% / min, it is determined to be a pure grid supply condition; when the photovoltaic power generation is greater than 10% of the rated power, it is determined to be a photovoltaic grid connection condition; when the load current change rate is greater than 30% / min, it is determined to be a sudden load change condition. Different operating conditions correspond to different neural network weight parameters, which are pre-stored in the prediction unit's memory, with a switching time of no more than 1ms.
[0040] It also includes a fault alarm module, which is connected to both the constraint verification module and the hybrid compensation unit. When the constraint verification module rejects the decision unit's command three times consecutively, or when the hybrid compensation unit's switching action fails, the fault alarm module generates an alarm signal and uploads it to the backend maintenance terminal. The alarm signal includes the alarm type, occurrence time, and associated module identifier. After receiving the alarm, maintenance personnel can remotely view the system logs to locate the cause of the fault and, if necessary, dispatch personnel for on-site repair. The threshold for triggering an alarm after three consecutive rejections can be adjusted according to on-site maintenance needs, ranging from 1 to 10 times.
[0041] The hybrid compensation unit is connected in parallel with the low-voltage busbar of the distribution transformer, and the thyristor-switched capacitors and static var generators are arranged independently, sharing the same set of current and voltage sampling circuits. Sharing the sampling circuits avoids compensation deviations caused by asynchronous sampling and reduces hardware costs. The sampling circuits use high-precision current transformers with a sampling accuracy of 0.2% and a sampling frequency of 10kHz, meeting the real-time control requirements of the TSC and SVG.
[0042] The specific implementation process is as follows: This embodiment provides a transformer power distribution system applied to a 10kV / 0.4kV distribution transformer area in an industrial park. The rated voltage of the low-voltage side of the distribution transformer in this area is 380V / 220V, the distributed photovoltaic installed capacity is 500kWp, and the photovoltaic penetration rate is approximately 40%.
[0043] I. System Configuration In this embodiment, the data acquisition unit is integrated into the distribution transformer intelligent fusion terminal (TTU). The TTU has a built-in analog acquisition module with a sampling accuracy of 0.5 class and a sampling frequency of 10kHz. The acquisition module collects the three-phase voltage values, three-phase load current values, and real-time power generation of the distributed photovoltaic system on the low-voltage side of the distribution transformer in real time. After digital filtering and downsampling processing, a set of valid data is output every 100ms to the rate of change calculation module, prediction unit, threshold adaptive module, and status recognition module.
[0044] The rate of change calculation module calculates the voltage change rate every 100ms. The calculation formula is as follows:
[0045] in, This is the voltage value at the current moment. The voltage value is 100ms ago, and the voltage change rate is expressed in V / s.
[0046] The prediction unit incorporates a backpropagation (BP) neural network model. This neural network has a three-layer structure: an input layer with four nodes corresponding to the current voltage value, voltage change rate, photovoltaic power generation, and photovoltaic power generation change rate, respectively; a hidden layer with ten nodes using the sigmoid function as the activation function; and an output layer with one node, outputting the predicted voltage value at the end of the future time window. A total of 62 weight parameters were obtained through offline training. The training samples consisted of 30 consecutive days of historical operating data for the transformer area, and the training algorithm was the Levenberg-Marquardt algorithm.
[0047] In this embodiment, the power supply line is 300m long and belongs to the urban area, so the future time window is set to 1 second.
[0048] The rate of change of photovoltaic power generation is calculated using the following formula:
[0049] in, This represents the current photovoltaic power generation capacity (kW). The value of photovoltaic power generation (kW) is 100ms ago, and the rate of change of photovoltaic power generation is in kW / s.
[0050] The prediction unit performs voltage over-limit risk prediction according to the following process: (1) The voltage prediction value at the end time 1 second from the output of the neural network. .
[0051] (2) Compare with the upper limit threshold (set to 238V in this embodiment). If If the voltage is V, it is determined that there is a risk of voltage exceeding the limit.
[0052] (3) When it is determined that there is a risk of voltage exceeding the limit, the estimated value of reactive power compensation capacity shall be calculated according to the following formula. :
[0053] Among them, the proportionality coefficient kvar / V.
[0054] (4) The above process is executed once every 100ms, and the execution time of a single execution is 5ms.
[0055] The decision-making unit receives the estimated reactive power compensation capacity output from the forecasting unit. Generate instructions according to the following rules: when When the reactive power deficit exceeds 10% of the system's reactive power deficit, a capacitor connection command is generated. when When the reactive power deficit is less than -10% of the system's reactive power deficit, a capacitor disconnection command is generated. when When the value is between ±10%, no instructions are generated.
[0056] The hybrid compensation unit includes a thyristor switched capacitor (TSC) and a static var generator (SVG), which are connected in parallel to the low-voltage side bus of the distribution transformer.
[0057] The threshold adaptive module dynamically adjusts the switching dead zone of the hybrid compensation unit based on the photovoltaic power fluctuation rate. The photovoltaic power fluctuation rate is calculated every minute, and the calculation formula is as follows:
[0058] in, , , These represent the maximum, minimum, and average photovoltaic power generation within a 1-minute time window.
[0059] In this embodiment, the initial value of the switching dead zone is set to 2V. The adjustment rule is as follows: When the photovoltaic power fluctuation rate is greater than 10% / min, the switching dead zone is narrowed to 0.5V~1V; When the photovoltaic power fluctuation rate is less than 2% / min, the switching dead zone is relaxed to 3V~5V; When the photovoltaic power fluctuation rate is between 2% / min and 10% / min, the switching dead zone remains unchanged from the previous adjustment value.
[0060] The constraint verification module receives instructions from the decision unit and simultaneously obtains the current voltage change trend from the rate of change calculation module, and performs verification according to the following rules: If the current voltage is decreasing and a capacitor connection command is received, then the command is rejected. If the current voltage is rising and a capacitor cutoff command is received, then the command is rejected. In other cases, the instruction is sent to the hybrid compensation unit.
[0061] In this embodiment, the processing time for a single verification cycle is 3ms.
[0062] The system is configured such that when the prediction unit predicts that the voltage will exceed 238V, the decision unit issues a command. After the constraint verification module verifies the command, the hybrid compensation unit performs tiered compensation based on the switching dead zone adjusted by the threshold adaptive module. During tiered compensation, thyristor-switched capacitors undertake large-step reactive power compensation, with a single input capacity not less than 60% of the system's reactive power deficit; the static var generator undertakes the remaining small-step reactive power compensation, achieving stepless smooth adjustment.
[0063] The status identification module identifies the operating condition based on the collected three-phase voltage values, three-phase load current values, and photovoltaic power generation, and issues an operating condition flag to the prediction unit. In this embodiment, the operating condition identification rule is as follows: When the photovoltaic power generation is less than 25kW (5% of the rated power) and the load current change rate is less than 10% / min, it is determined to be a pure grid power supply condition. When the photovoltaic power generation capacity is greater than 50kW (10% of the rated power), it is judged as a photovoltaic grid-connected operating condition; When the load current change rate is greater than 30% / min, it is determined to be a sudden load change condition.
[0064] When the operating parameters are within the boundary range of the above-mentioned operating condition judgment, the system maintains the operating condition label of the previous moment to avoid frequent switching of neural network weights. The prediction unit switches the corresponding neural network weight parameters according to different operating conditions, and the switching time does not exceed 1ms.
[0065] The fault alarm module is connected to both the constraint verification module and the hybrid compensation unit. When the constraint verification module rejects the decision unit's command three times consecutively, or when the hybrid compensation unit's switching action fails, the fault alarm module generates an alarm signal and uploads it to the backend maintenance terminal. The alarm signal includes the alarm type, occurrence time, and associated module identifier. In this embodiment, the threshold for triggering an alarm with consecutive rejection commands is set to 3 times. Maintenance personnel can remotely adjust this threshold via the backend terminal, with an adjustment range of 1 to 10 times.
[0066] The hybrid compensation unit is connected in parallel to the low-voltage busbar of the distribution transformer, and the thyristor-switched capacitors and the static var generator are arranged independently, sharing the same set of current and voltage sampling circuits. Sharing the sampling circuits avoids compensation deviations caused by asynchronous sampling and reduces hardware costs. The sampling circuits use high-precision current transformers with a sampling accuracy of 0.2% and a sampling frequency of 10kHz, meeting the real-time control requirements of the thyristor-switched capacitors and the static var generator.
[0067] II. System Operation Examples Scenario: At noon one day, the weather changed from cloudy to sunny, and the photovoltaic power generation rapidly increased from 50kW to 450kW within 30 seconds. In this scenario, the photovoltaic power fluctuation rate is greater than 10% / min. The system automatically narrows the switching dead zone to 0.5V~1V to improve response sensitivity.
[0068]
[0069] Effect comparison:
[0070] In this embodiment, under the scenario of severe fluctuations in photovoltaic power, the system predicts and issues a compensation command about 3 seconds before the voltage actually exceeds the limit based on real-time monitoring with a 100ms cycle, successfully stabilizing the voltage within the set threshold and completely avoiding voltage over-limit problems and abnormal grid disconnection of photovoltaics. At the same time, relying on the adaptive switching dead zone and hierarchical compensation strategy, it effectively suppresses frequent switching oscillations of capacitors, making the equipment operation more stable.
[0071] The above descriptions are merely embodiments of the present invention, and common knowledge such as specific technical solutions and / or characteristics are not described in detail here. It should be noted that those skilled in the art can make various modifications and improvements without departing from the technical solutions of the present invention, and these should also be considered within the scope of protection of the present invention. These modifications and improvements will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.
Claims
1. A transformer power distribution system, characterized in that, include: The data acquisition unit is used to collect the three-phase voltage value, three-phase load current value, and real-time power generation of the distributed photovoltaic system on the low-voltage side of the distribution transformer in real time. The rate of change calculation module, whose input terminal is connected to the data acquisition unit, is used to calculate the voltage rate of change; The prediction unit, whose input terminals are connected to the data acquisition unit and the rate of change calculation module respectively, has a built-in neural network model to predict the risk of voltage exceeding the limit within a future time window based on the current voltage value, voltage change rate and photovoltaic power generation change rate, and calculate the estimated value of reactive power compensation capacity in combination with voltage deviation. The decision unit, whose input is connected to the prediction unit, is used to generate instructions based on the estimated reactive power compensation capacity, the instructions including capacitor connection instructions or capacitor disconnection instructions. The hybrid compensation unit includes a thyristor-switched capacitor and a static var generator; A threshold adaptive module, whose input is connected to the data acquisition unit, is used to dynamically adjust the switching dead zone of the hybrid compensation unit according to the photovoltaic power fluctuation rate. The switching dead zone refers to the voltage range that does not trigger any capacitor switching operation. The voltage range is narrowed when the photovoltaic power fluctuation rate increases to improve sensitivity, and widened when the photovoltaic power fluctuation rate decreases to avoid oscillation. A constraint verification module, whose input is connected to the decision unit, whose output is connected to the hybrid compensation unit, and whose verification input is connected to the rate of change calculation module, is used to verify the command based on the voltage change trend. If the current voltage is in a decreasing state, the capacitor connection command is rejected; If the current voltage is rising, the capacitor disconnect command is rejected; If the current voltage is rising and the received instruction is a capacitor connection instruction, or if the current voltage is falling and the received instruction is a capacitor disconnection instruction, then the instruction is allowed to be sent to the hybrid compensation unit. The system is configured such that when the prediction unit predicts that the voltage will exceed the upper limit threshold, the decision unit issues an instruction, which is verified by the constraint verification module, and the hybrid compensation unit performs hierarchical compensation according to the switching dead zone adjusted by the threshold adaptive module.
2. The transformer power distribution system as described in claim 1, characterized in that, The rate of change calculation module calculates the voltage rate of change according to the following formula: in, This is the voltage value at the current moment. This is the voltage value at the previous sampling time. The sampling interval; It takes 100ms.
3. The transformer power distribution system as described in claim 1, characterized in that: The future time window in the prediction unit is 1 to 5 seconds.
4. The transformer power distribution system as described in claim 1, characterized in that: The neural network model in the prediction unit is a BP neural network, whose input layer nodes include: current voltage value, voltage change rate, photovoltaic power generation, and photovoltaic power generation change rate; the output layer node is: the estimated value of the required reactive power compensation capacity.
5. A transformer power distribution system as described in claim 1, characterized in that: The upper limit threshold is 235V to 242V.
6. A transformer power distribution system as described in claim 1, characterized in that: The threshold adaptive module adjusts the switching dead zone, which is adjustable from 0.5V to 5V. When the photovoltaic power fluctuation rate is greater than 10% / min, the dead zone is narrowed to 0.5V-1V. When the photovoltaic power fluctuation rate is less than 2% / min, the dead zone is widened to 3V-5V.
7. A transformer power distribution system as described in claim 1, characterized in that, When the hybrid compensation unit performs graded compensation: the thyristor-switched capacitor is used to undertake large-step reactive power compensation, with a single input capacity not less than 60% of the system's reactive power deficit; the static var generator is used to undertake the remaining small-step reactive power compensation, achieving stepless smooth adjustment.
8. A transformer power distribution system as described in claim 1, characterized in that: It also includes a status identification module, which is connected to the data acquisition unit and the prediction unit respectively. The status identification module identifies three operating conditions—pure grid power supply, photovoltaic grid connection, and sudden load change—based on the acquired three-phase voltage value, three-phase load current value, and photovoltaic power generation. It then issues an operating condition label to the prediction unit, which switches the corresponding neural network weight parameters according to the different operating conditions.
9. A transformer power distribution system as described in claim 1, characterized in that: It also includes a fault alarm module, which is connected to the constraint verification module and the hybrid compensation unit respectively. When the constraint verification module rejects the decision unit instruction three times in a row, or the hybrid compensation unit fails to switch, the fault alarm module generates an alarm signal and uploads it to the background operation and maintenance terminal.
10. A transformer power distribution system as described in claim 1, characterized in that: The hybrid compensation unit is connected in parallel with the low-voltage side bus of the distribution transformer, and the thyristor-switched capacitor and the static var generator are arranged independently, sharing the same set of current and voltage sampling circuits.
