Virtual reactive power plant user single-phase asynchronous motor reactive power regulation method

Abstract

The invention discloses a virtual reactive power plant user single-phase asynchronous motor reactive power regulation method, and relates to the technical field of electrical equipment control, and the method comprises the steps: collecting the real-time state data of a key node of a power distribution network, and building a short-term prediction model based on the state data; setting a judgment condition according to a prediction result of the model, generating a preliminary reactive power regulation instruction, and adjusting the priority and the regulation amplitude of the preliminary reactive power regulation instruction in real time based on dynamic updating of a prediction trend; issuing the adjusted reactive power adjustment instruction to a user terminal, and performing cooperative processing based on distributed coordination control to generate a final cooperative adjustment instruction; executing an action of cutting off the split-phase capacitor of the single-phase asynchronous motor according to the coordinated adjustment instruction; according to the invention, the split-phase capacitance of the single-phase asynchronous motor at the user side is regulated and controlled by using the virtual reactive power plant, so that low-cost, quick-response, continuous and accurate reactive power regulation is realized, and the problems of reactive power excess and high voltage of a power grid are effectively solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electrical equipment control, and more particularly to a virtual reactive power plant user single-phase asynchronous motor reactive power regulation method. BACKGROUND

[0002] The new power system has high requirements for grid reactive power compensation, and precise compensation is required in many cases. The traditional capacitor and reactor are intelligently graded and switched, and cannot be frequently switched due to the problem of impact current. The SVG, SVC and the like based on power electronic technology utilize semiconductor devices to form controllable reactors or static var generators, which can continuously and precisely compensate reactive power and do not have the problem of slow response to switching, but generally have the problems of high cost and high loss. Therefore, there is an urgent need for a reactive power compensation regulation method that is low in price, can continuously and precisely regulate, has a high response speed and is low in loss. Under the condition of high proportion of new energy distributed photovoltaic and the like connected to the grid, the problem of excessive grid reactive power and high voltage is very important to utilize existing equipment resources and absorb excess reactive power to control voltage.

[0003] In view of the above problems, the present application provides a solution. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a virtual reactive power plant user single-phase asynchronous motor reactive power regulation method, which utilizes the virtual reactive power plant to regulate and control the split-phase capacitor of the user side single-phase asynchronous motor, realizes low-cost, fast-response and continuous-precision reactive power regulation, and effectively solves the problems of excessive grid reactive power and high voltage.

[0005] To achieve the above purpose, the present application provides the following technical scheme: a virtual reactive power plant user single-phase asynchronous motor reactive power regulation method, which collects real-time state data of key nodes of a power distribution network, constructs a short-term prediction model based on the state data, predicts the reactive power change and voltage change trend in a future period of time, sets a determination condition according to the prediction result of the model, generates a preliminary reactive power regulation instruction, and dynamically updates the priority and regulation amplitude of the preliminary reactive power regulation instruction in real time based on the prediction trend, issues the adjusted reactive power regulation instruction to a user terminal, and performs collaborative processing based on distributed coordinated control to generate a final collaborative regulation instruction, and executes the action of cutting off the split-phase capacitor of the single-phase asynchronous motor according to the collaborative regulation instruction.

[0006] In a preferred embodiment, real-time state data of a key node of a power distribution network is collected, and a short-term prediction model is constructed based on the state data. Specifically, input feature weights are assigned to the state data based on an adaptive neural network, and a comprehensive reactive state feature is obtained through weighted fusion. The state data includes instantaneous reactive disturbance data, phase voltage nonlinear data, and load response delay data. A supervised learning data set is constructed based on the comprehensive reactive state feature and future short-term reactive power and voltage. A bidirectional LSTM network is constructed, and the supervised learning data set is input into the bidirectional LSTM network for training to obtain a trained short-term prediction model. The real-time comprehensive reactive state feature is input into the short-term prediction model, and the future short-term power grid reactive power and voltage prediction values are output.

[0007] In a preferred embodiment, input feature weights are assigned to the state data based on an adaptive neural network, and a comprehensive reactive state feature is obtained through weighted fusion. Specifically, a standardized feature matrix is generated based on the state data, and the power grid reactive power and voltage in the future short term are taken as historical prediction targets. A neural network is constructed, the standardized feature matrix and the corresponding historical prediction targets are input into the network, and the error between the prediction result and the actual target is obtained using a mean square error loss function. Through a back propagation algorithm, the network weights are iteratively updated according to the prediction error until the loss function converges. The feature weights are extracted to form an adaptive weight matrix, which is weighted and fused with the standardized feature matrix to generate the comprehensive reactive state feature.

[0008] In a preferred embodiment, a preliminary reactive regulation instruction is generated based on the prediction result of the model. Specifically, the real-time voltage and real-time reactive power of the power grid node are obtained. The real-time voltage is compared with a preset voltage upper limit, and the real-time reactive power is compared with a preset power threshold to obtain a first determination result. The voltage prediction value and the reactive power prediction value of the power grid node are obtained. The voltage prediction value is compared with the voltage upper limit, and the reactive power prediction value is compared with the power threshold to obtain a second determination result. Logical determination is performed based on the first determination result and the second determination result. When the first determination result indicates that there is a potential overvoltage and reactive surplus state, or the second determination result indicates that there is an overvoltage and reactive surplus state, a preliminary reactive regulation instruction is generated.

[0009] In a preferred embodiment, the priority and adjustment range of the preliminary reactive power adjustment instruction are adjusted in real time, specifically: first data after the preliminary reactive power adjustment instruction is generated is acquired, the reactive power adjustment weight associated with the first data is combined, an adjustment priority index is calculated, and the reactive power adjustment weight is dynamically adjusted based on a Bayesian optimization algorithm; the multiple preliminary reactive power adjustment instructions are sorted according to the adjustment priority index, and the instructions with high priority are executed in priority over the instructions with low priority; the current operating state of the controlled motor is acquired, and the preliminary adjustment range is calculated through a preset preliminary adjustment range calculation formula in combination with the first data; the preliminary adjustment range is optimized based on a robust control method to obtain an optimized adjustment range; the corresponding reactive power adjustment instruction is executed according to the execution order after sorting and the optimized adjustment range, and the deviation difference between the actual change value and the predicted change value is acquired by monitoring the power grid voltage and the reactive power change in real time; the adjustment priority and adjustment range of the subsequent reactive power adjustment instruction are dynamically adjusted according to the deviation difference, and a final reactive power adjustment instruction is formed.

[0010] In a preferred embodiment, the reactive power adjustment weight is dynamically adjusted based on the Bayesian optimization algorithm, specifically: the reactive power adjustment weight of the system is regarded as a to-be-optimized variable, and an optimization objective function is constructed; based on historical operation data, a Gaussian process prior model is constructed for the reactive power adjustment weight space; an initial sampling point is selected in the weight space, the adjustment priority index is calculated, and an adjustment operation is performed; the value of the corresponding objective function is calculated based on the adjustment result, and an acquisition function is constructed according to the predicted mean and variance of the Gaussian process prior model; the weight combination with the maximum acquisition function value is selected as the next sampling point, the corresponding objective function value is calculated, and the Gaussian process model is updated; the above selection, calculation and updating process is iteratively executed until a preset iteration termination condition is met; the reactive power adjustment weight combination that minimizes the optimization objective function value is selected from all iteration samples as the final optimization result.

[0011] In a preferred embodiment, the preliminary adjustment range is optimized based on a robust control method to obtain an optimized adjustment range, specifically: second data is acquired and modeled as system uncertainty, the deviation of the actual voltage value measured from the target voltage reference value is taken as an index, a robust optimization objective function and a safety constraint are constructed; the robust optimization objective function is solved by an optimization algorithm to meet the voltage and reactive power safety constraints under consideration of system uncertainty, and the optimized adjustment range is obtained.

[0012] In a preferred embodiment, the adjusted reactive power regulation instruction is issued to the user terminal, and is collaboratively processed based on distributed coordinated control to generate a final collaborative regulation instruction, specifically: after receiving the final reactive power regulation instruction, the user terminal acquires the reactive power regulation capability of the current motor; each user terminal sends the adjustable capability to the neighbor node through a preset communication protocol, and receives the capability information of the neighbor node; the user terminal corrects the execution proportion of the reactive power regulation instruction based on the adjustable capability of itself and the neighbor node information using a consensus iteration algorithm; after the iterative calculation converges, each user terminal generates the final collaborative regulation instruction according to the convergence result.

[0013] In a preferred embodiment, the execution proportion of the reactive power regulation instruction is corrected using a consensus iteration algorithm; after the iterative calculation converges, each user terminal generates the final collaborative regulation instruction according to the convergence result, specifically: each user terminal calculates an initial regulation proportion according to the received initial reactive power regulation instruction and the adjustable capability of itself, and exchanges the regulation proportion information with the neighbor node through a communication network; each user terminal iteratively updates the regulation proportion according to the local regulation proportion and the regulation proportion information of the neighbor node according to the consensus iteration rule; after each iteration, the change amount of the local regulation proportion is calculated, and the change amount information is exchanged with the neighbor node, when the regulation proportion change amount of all neighbor nodes is less than a preset change amount threshold, it is determined that the iteration converges; the regulation proportion after the iteration converges and the adjustable capability of itself are used to generate the final collaborative regulation instruction.

[0014] In a preferred embodiment, according to the collaborative regulation instruction, the action of cutting off the split-phase capacitor of the single-phase asynchronous motor is performed, specifically: the real-time running state parameters of the user terminal according to the motor are acquired and conditional judgment is performed; 1) the change rate of the motor speed in a preset time window is less than a set threshold; 2) whether the motor temperature rise is within a preset allowable range; 3) whether the deviation of the motor stator voltage and current relative to the rated value exceeds a preset threshold; when the above conditions are all met, it is determined that the motor can safely perform the split-phase capacitor cutting operation; under the condition that the motor running condition is met and the collaborative regulation instruction is required, the action of cutting off the split-phase capacitor of the single-phase asynchronous motor is performed; when the real-time state or the predicted state does not meet the overvoltage and reactive power surplus conditions, the user terminal restores the split-phase capacitor access state.

[0015] The one or more technical solutions provided in the embodiments of the application have at least the following technical effects or advantages: 1. Through the combination of short-term prediction model, adaptive neural network feature weighting, bidirectional LSTM time series prediction, Bayesian optimization weight adjustment and robust control amplitude optimization, the accurate and dynamic adjustment of grid reactive power and voltage is realized. The advantages are as follows: the current state and short-term future trend of the grid can be considered simultaneously to realize active and forward-looking reactive power regulation; the adaptive weight and Bayesian optimization are used to dynamically allocate the priority and amplitude of regulation to improve the accuracy and adaptability of the regulation strategy; the robust control is combined to optimize the regulation amplitude under the uncertainty of system measurement error, prediction deviation and external disturbance, so as to ensure the stable operation of voltage and power factor within the safety constraint range; the overall scheme forms a closed-loop, adaptive and intelligent reactive power regulation mechanism, which not only improves the power factor and voltage stability of the grid, but also enhances the reliability and system robustness of the regulation, providing a solid guarantee for the safe and efficient operation of the grid.

[0016] 2. By issuing the adjusted reactive power regulation instruction to the user terminal and combining the distributed coordination control and consistency iteration algorithm, the collaborative optimization regulation of each terminal under its own safety constraint is realized, so that the grid reactive power and voltage level are effectively controlled. At the same time, the dynamic removal and restoration of the split-phase capacitor of the single-phase asynchronous motor are used to convert the internal controllable resources of the load equipment into virtual reactive power, realizing fast, continuous and accurate reactive power regulation, and avoiding the cost of adding additional capacitors or electronic compensation devices. This scheme takes into account real-time response, prediction foresight, safety constraint and grid coordination, which not only improves the reliability and robustness of the regulation, but also significantly enhances the power factor optimization ability and voltage stability of the grid, achieving the efficient operation effect of the virtual reactive power plant. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 A virtual reactive power plant user single-phase asynchronous motor reactive power regulation method flowchart is provided for the embodiments of the present application.

[0018] Figure 2 A short-term grid reactive power and voltage prediction comparison chart is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0020] Referring to Figure 1 The virtual reactive power plant user single-phase asynchronous motor reactive power regulation method flowchart provided by the present application includes the following steps: S1, collecting real-time state data of key nodes of the power distribution network, and constructing a short-term prediction model of reactive power and voltage variation of the power grid.

[0021] In this embodiment, real-time state data of key nodes of the power distribution network is collected, and a short-term prediction model is constructed based on the state data to predict the trend of reactive power variation and voltage variation in the future period of time, specifically: Collecting real-time state data of key nodes of the power distribution network, the state data including instantaneous reactive disturbance data, phase voltage nonlinear data and load response delay data; Based on the adaptive neural network, input feature weights are assigned to the state data, and weighted fusion is performed to obtain comprehensive reactive state features; Based on the preset short-term prediction window, the comprehensive reactive state features are taken as time series input, and the future short-term grid reactive power and voltage are taken as output sequence to construct a supervised learning data set; A bidirectional LSTM network is constructed, including an input layer, a bidirectional LSTM hidden layer and an output layer. The number of nodes of the input layer is consistent with the dimension of the comprehensive reactive state features, and the number of nodes of the hidden layer is determined according to cross-validation or experience. The bidirectional LSTM hidden layer simultaneously captures past and future feature dependencies through forward and reverse time sequence paths, improving the short-term fluctuation prediction capability. The output layer uses a linear activation function to continuously predict future reactive power and voltage; As shown in Figure 2 The time series input and the output sequence are input into the bidirectional LSTM network, and the mean square error loss function is used to perform iterative training by gradient descent method until the preset convergence number is reached, obtaining the trained short-term prediction model; The real-time comprehensive reactive state features are input into the short-term prediction model, and the future short-term grid reactive power and voltage prediction values are output.

[0022] It should be noted that the short-term prediction model can be constructed by the bidirectional LSTM network to simultaneously capture the past trend and potential future dependency of the grid reactive state features, achieving high-precision prediction of short-term reactive power and voltage fluctuations. Combined with the adaptive neural network weighted fusion comprehensive feature input, the model can dynamically identify the importance of each feature to the prediction result, improving the stability and robustness of the prediction, thereby providing timely and reliable decision basis for grid reactive power regulation, significantly enhancing the safety and power factor optimization capability of the grid operation.

[0023] Further, based on the adaptive neural network, input feature weights are assigned to the grid state data, and weighted fusion is performed to obtain comprehensive reactive state features, specifically: The grid state data is standardized to generate a standardized feature matrix; The reactive power and voltage of the power grid in the near future are taken as historical prediction targets to provide labels for model training A neural network with an input layer, a hidden layer and an output layer is constructed, wherein the input layer nodes correspond to each standardized feature matrix, the hidden layer activation function adopts ReLU, the output layer is linear activation, and the network weight is initialized as mean 0; The standardized feature matrix and the corresponding historical prediction target are input into the network, and the error between the prediction result and the actual target is obtained by using the mean square error loss function; Through the back propagation algorithm, the network weight is iteratively updated according to the prediction error until the loss function converges; The feature weight is extracted from the connection weight from the input layer to the output layer of the trained network to form an adaptive weight matrix; The adaptive weight matrix and the standardized feature matrix are weighted and fused to generate a comprehensive reactive state feature, which is used in the short-term prediction model to realize high-precision prediction of the future reactive power and voltage of the power grid; In the actual operation process of the model, the prediction error is fed back to the adaptive neural network in real time, and the weight matrix is dynamically adjusted according to the latest error to ensure that the contribution of each feature to the prediction result is adaptively optimized with the change of the power grid operation state.

[0024] The instantaneous reactive disturbance data is an index quantifying the amplitude of reactive power fluctuation in a short time. It reflects the strength of the instantaneous reactive disturbance of the power grid and can capture short-term power fluctuations caused by load mutation, local device switching or instantaneous reactive power injection. By obtaining the instantaneous reactive disturbance data and taking it as the input of the short-term prediction model, the short-term fluctuation characteristics of the reactive power of the power grid can be quantified in real time, and the instantaneous disturbance trend can be captured. Based on this quantitative index, a prediction model is constructed, which helps to identify the risk of reactive excess and voltage overshoot in the next few minutes in advance, so as to realize active adjustment and optimization control, reduce power grid voltage fluctuation, improve power factor stability, and reduce equipment loss and power quality problems caused by reactive excess. The specific acquisition steps of the instantaneous reactive disturbance data are as follows: Real-time acquisition of reactive power of each phase ; Calculate the change rate of reactive power in the preset time window; The specific calculation formula of the change rate of reactive power is as follows:

[0025] In the formula, is the change rate, is the sampling interval, is a small constant to avoid division by zero, is the last sampling interval Instantaneous reactive power at the moment.

[0026] The instantaneous reactive power disturbance data is obtained by averaging the continuous N sampling points of the collected reactive power.

[0027]

[0028] In the formula, is the instantaneous reactive power disturbance data.

[0029] It should be noted that the higher the instantaneous reactive power disturbance data, the more severe the short-term reactive power fluctuation of the power grid, and there is a risk of excess or sudden change.

[0030] The phase voltage nonlinear data is an index for measuring the degree of deviation of the voltage waveform from the ideal sine wave, which is represented by the ratio of the high-order harmonic component to the fundamental component. By obtaining the phase voltage nonlinear data and taking it as the input of the short-term prediction model, the high-order harmonic component and distortion degree of the voltage waveform can be quantified, reflecting the possible reactive abnormality and local voltage rise of the power grid in a short time. Based on this index, a prediction model is constructed, which helps to identify the potential risks of reactive excess and voltage high of the power grid in advance, realizes more accurate active adjustment, improves the stability of power grid operation and power quality, and reduces the loss and failure probability caused by voltage waveform distortion of equipment. The specific acquisition steps of the phase voltage nonlinear data are as follows: Real-time acquisition of phase voltage , the sampling waveform is subjected to fast Fourier transform to obtain the fundamental component and the high-order harmonic component ; The phase voltage nonlinear data is calculated by the ratio of the high-order harmonic component to the fundamental component.

[0031] The specific calculation formula of the phase voltage nonlinear data is as follows:

[0032] In the formula, is the phase voltage nonlinear data.

[0033] It should be noted that the higher the phase voltage nonlinear data, the more serious the voltage waveform distortion, accompanied by local reactive excess and voltage abnormality.

[0034] The load response delay data is an index for measuring the response lag time of the user side load to voltage or reactive power change. By obtaining the load response delay data and taking it as the input of the short-term prediction model, the response lag characteristics of the user side load to voltage and reactive power change can be quantified, and the trend of reactive power accumulation or voltage rise that may occur in the power grid in a short time can be revealed. Based on this index, the prediction model is constructed, which helps to discover the risk of excessive reactive power and voltage anomaly in the power grid in advance, realize active adjustment, improve the dynamic response capability of the power grid, and optimize the power factor control. The specific acquisition steps of the load response delay data are as follows: collecting the grid voltage and the load current , calculating the cross-correlation function of voltage change and load change ; The specific calculation formula of the cross-correlation function of voltage change and load change is as follows:

[0035] In the formula, is the average value of voltage, is the average value of current, is the lag time of the current sequence relative to the voltage sequence.

[0036] traverse different values, find the maximum lag time , and obtain the load response delay data.

[0037] The specific calculation formula of the load response delay data is as follows:

[0038] It should be noted that the larger the load response delay data is, the more obvious the load response lag is, and the power grid will have excessive reactive power or high voltage in a short time.

[0039] S2, according to the prediction result of the model, set the determination condition, generate the preliminary reactive power adjustment instruction, and based on the dynamic update of the prediction trend, real-time adjust the priority and adjustment amplitude of the preliminary reactive power adjustment instruction.

[0040] In this embodiment, according to the prediction result of the model, the determination condition is set, and the preliminary reactive power adjustment instruction is generated, which is specifically: obtaining the future short-term grid reactive power prediction value and voltage prediction value according to the short-term prediction model; obtaining real-time voltage and real-time reactive power; comparing the real-time voltage with the preset voltage upper limit, and comparing the real-time reactive power with the preset power threshold; If the real-time voltage is greater than the preset upper limit of voltage and the real-time reactive power is greater than or equal to the preset power threshold, it is determined that the current real-time state is a potential overvoltage and reactive power surplus state, and recorded as a first determination result; The voltage prediction value is compared with the preset upper limit of voltage, and the reactive power prediction value is compared with the preset power threshold; If the voltage prediction value is greater than the preset upper limit of voltage and the reactive power prediction value is greater than or equal to the preset power threshold, it is determined that the prediction state is an overvoltage and reactive power surplus state, and recorded as a second determination result; Logical determination is performed according to the first determination result and the second determination result; When the first determination result indicates that there is the potential overvoltage and reactive power surplus risk state, or the second determination result indicates that there is the overvoltage and reactive power surplus state, a preliminary reactive power regulation instruction is generated.

[0041] It should be noted that the purpose of determining the real-time power grid state and the prediction power grid state respectively is to take into account the current actual operation and the future short-term trend, so as to realize more safe, proactive and accurate reactive power regulation: the real-time state ensures that the immediate overvoltage and reactive power surplus can be responded immediately to ensure the current safety; the prediction state identifies the possible future abnormalities in advance to provide forward-looking guidance for regulation actions, so that the regulation range is smooth and the sequence is optimized, and the impact on the motor and the power grid is reduced; the combination of the two forms an intelligent regulation strategy that can respond immediately and predictively drive, improving system stability and regulation accuracy.

[0042] Further, the priority and regulation range of the preliminary reactive power regulation instruction are adjusted in real time, specifically: First data after generating the preliminary reactive power regulation instruction is obtained, the first data including the amplitude of the prediction voltage exceeding the upper limit (the difference between the prediction voltage and the preset upper limit of voltage), the high degree of the reactive power prediction value (the difference between the reactive power prediction value and the preset power threshold), and the safety margin of the single-phase asynchronous motor; The regulation priority index is calculated according to the first data and the corresponding reactive power regulation weight, and the reactive power regulation weight is dynamically adjusted based on the Bayesian optimization algorithm; The specific calculation formula of the regulation priority index is as follows:

[0043] In the formula, is the regulation priority index, is the amplitude of the prediction voltage exceeding the upper limit, is the high degree of the reactive power prediction value, is the safety margin of the single-phase asynchronous motor, is the reactive power regulation weight of the amplitude of the prediction voltage exceeding the upper limit, a reactive power regulation weight of a degree of overestimation of the reactive power prediction value, a reactive power regulation weight of a safety margin of the single-phase asynchronous motor.

[0044] sorting the plurality of preliminary reactive power regulation instructions according to the regulation priority indicators (from large to small), and the instruction with a high priority is prior to the instruction with a low priority in the execution order; obtaining a current operating state of the controlled motor and combining the first data to calculate a preliminary regulation amplitude by using a preset preliminary regulation amplitude calculation formula; The preliminary regulation amplitude calculation formula is specifically as follows:

[0045] In the formula, is the preliminary regulation amplitude, , is a regulation weight coefficient, is a motor state coefficient, which is 1 when the motor is in a steady state, and is a coefficient less than 1 otherwise.

[0046] optimizing the preliminary regulation amplitude based on a robust control method to obtain an optimized regulation amplitude; executing the corresponding reactive power regulation instructions according to the execution order after sorting and the optimized regulation amplitude, and monitoring the power grid voltage and the reactive power in real time to obtain a deviation difference value between an actual change value and a predicted change value; dynamically adjusting the regulation priority and the regulation amplitude of a subsequent reactive power regulation instruction according to the deviation difference value to form a final reactive power regulation instruction, the final reactive power regulation instruction including a target regulation object, an execution order, a regulation amplitude, and an execution condition.

[0047] The dynamically adjusted regulation priority and regulation amplitude are specifically calculated by using the following formula:

[0048]

[0049] In the formula, is the dynamically adjusted regulation priority, , is an adjustment coefficient, is the dynamically adjusted regulation amplitude, is the deviation difference value.

[0050] Further, the reactive power regulation weight is dynamically adjusted based on a Bayesian optimization algorithm, specifically as follows: The reactive power regulation weight (W) is adjusted based on the Bayesian optimization algorithm, and the adjustment is specifically as follows: , , ) as the optimization variable, and an optimization objective function is constructed; the optimization objective function , and the specific calculation formula is as follows:

[0051] In the formula, V is an actual value of voltage, is a target voltage reference value, PF is an actual value of reactive power, is a target reactive power reference value, , is an adjustment performance weight.

[0052] Based on historical operation data, a weight space of reactive power adjustment , , A Gaussian process prior model is constructed to describe the uncertainty and potential distribution of the optimization objective function ; An initial sampling point is selected in the weight space , is substituted into the adjustment priority index for calculation, and adjustment is performed according to the priority to obtain an adjustment effect, which refers to the degree to which the state of the power grid approaches the ideal target reference value after the execution of the instruction; Based on the adjustment effect, a value of the corresponding objective function is calculated as a preliminary observation sample to update the Gaussian process model; According to a predicted mean and a variance of the Gaussian process model, a collection function is constructed, and the collection function is expected improvement (EI); A weight combination with the maximum value of the collection function is selected as the next sampling point, a value of the corresponding objective function is calculated, and the Gaussian process model is updated, and the above selection, calculation and updating process is iteratively executed until a preset maximum iteration number or the objective function converges; From all the iteration samples, a weight combination of reactive power adjustment is selected, which makes the optimization objective function have the minimum value.

[0053] It should be noted that the reactive power adjustment weight is dynamically adjusted through Bayesian optimization, so that the weight combination can adaptively match the power grid operation characteristics and adjustment targets, thereby maximizing the adjustment effect while ensuring that the voltage and power factor are close to the target reference values. The specific benefits include: on the one hand, the self-adaptability and accuracy of the reactive power adjustment strategy are improved, so that the adjustment priority and amplitude under different operating states are more reasonable; on the other hand, through Gaussian process prediction and expected improvement collection function, efficient exploration and optimization of the weight space are realized, the dependence on manual parameter adjustment is reduced, and the reliability and stability of the strategy under complex power grid conditions are improved, thereby providing a reliable foundation for subsequent real-time reactive power adjustment and intelligent control.

[0054] Further, the preliminary adjustment amplitude is optimized based on a robust control method to obtain an optimized adjustment amplitude, specifically: Obtain second data and model as system uncertainty The second data is power grid measurement error, prediction error and external disturbance; The deviation between the actual value of the current measured voltage and the target voltage reference value is used as an index to construct a robust optimization objective function and voltage and reactive power safety constraints, That is , ; The robust optimization objective function is calculated as follows:

[0055] In the formula, is the robust adjustment amplitude, that is, the optimal amplitude of reactive power adjustment considering system uncertainty (measurement error, prediction error and external disturbance), is the value of system uncertainty Δ, and u is the uncertainty set, is under the action of system uncertainty influenced voltage change, is the norm operation, is the adjustment amplitude that can minimize the voltage deviation under the premise of considering the worst-case uncertainty is the core idea of robust optimization.

[0056] The robust optimization objective function is solved by an optimization algorithm to satisfy the voltage and reactive power safety constraints considering system uncertainty, and the optimized adjustment amplitude is obtained.

[0057] It should be noted that by obtaining the preliminary adjustment amplitude and priority, and constructing the robust adjustment objective function under the system uncertainty model, the optimal solution and output of the reactive power adjustment amplitude are realized. The advantage is that no matter how the power grid measurement error, prediction error or external disturbance changes, the adjustment amplitude can be adaptively optimized under the premise of satisfying the voltage and power factor safety constraints, thereby ensuring the stability of the power grid voltage, the power factor close to the target value, improving the reliability and execution accuracy of the reactive power adjustment strategy, providing a solid foundation for subsequent dynamic adjustment and intelligent control, and reducing the risk of power grid overvoltage or power factor deviation caused by uncertainty.

[0058] S3, the adjusted reactive power adjustment instruction is issued to the user terminal, and is processed based on distributed coordinated control to generate a final coordinated adjustment instruction.

[0059] ​In this embodiment, the adjusted reactive power regulation instruction is issued to the user terminal, and is cooperatively processed based on distributed coordination control to generate a final cooperative regulation instruction, specifically: After receiving the final reactive power regulation instruction, the user terminal acquires the current motor reactive power regulation capability, which includes the maximum absorbable or releasable reactive power and the corresponding regulation amplitude range; Each user terminal sends the local adjustable capability to the neighbor node through a preset communication protocol, and receives the capability information of the neighbor node, thereby realizing information intercommunication between the local and the neighbor. Based on the self-adjustable capability and the neighbor node information, the user terminal corrects the execution proportion of the reactive power regulation instruction using a consistency iteration algorithm, and generates the final cooperative regulation instruction according to the convergence result after the iteration calculation converges.

[0060] Further, the consistency iteration algorithm is used to correct the execution proportion of the reactive power regulation instruction, and the final cooperative regulation instruction is generated according to the convergence result after the iteration calculation converges, specifically: Each user terminal calculates an initial regulation proportion according to the received reactive power regulation instruction and the self-adjustable capability, and exchanges the regulation proportion information with the neighbor node through a communication network.

[0061] In the formula, is the initial regulation proportion, is the maximum adjustable reactive power of terminal i, is the minimum adjustable reactive power of terminal i, is the reactive power regulation target of the adjusted reactive power regulation instruction, and K is the total number of user terminals participating in the regulation. is the total adjustable capability of all participating terminals in the entire network.

[0062] According to the neighbor information and the local initial regulation proportion, the user terminal iteratively updates the regulation proportion according to the consistency iteration rule, specifically:

[0063] In the formula, is the regulation proportion of user terminal i at the g+1th iteration, is the regulation proportion of user terminal i at the gth iteration, is an iteration step coefficient for controlling the adjustment amplitude of each iteration, is the neighbor node set of user terminal i, is the weight coefficient of neighbor node j to node i, is the regulation proportion of neighbor node j of node i at the gth iteration.

[0064] After each iteration, the change of the local adjustment ratio is calculated, and when the change of all neighbor nodes is less than a preset change threshold, it is determined that the iteration converges. The adjustment ratio after the iteration converges is combined with the self-adjustable ability to generate a final coordinated adjustment instruction.

[0065] It should be noted that by adopting the distributed coordination control mode, single point failure and communication bottleneck caused by centralized control can be effectively avoided, so that each user terminal can not only meet its own operation safety constraints, but also optimize the reactive power and voltage level of the whole network. At the same time, the real-time and robustness of the adjustment are improved through local rapid calculation and neighbor iteration convergence mechanism, and the adjustment conflict is avoided. Combined with the dynamic feedback update link, the adaptive ability of the power grid in fluctuating environment is enhanced, so as to realize the safety, coordination and stability of the whole network reactive power adjustment.

[0066] S4, performing the action of cutting off the split-phase capacitor of the single-phase asynchronous motor according to the coordinated adjustment instruction.

[0067] In this embodiment, the action of cutting off the split-phase capacitor of the single-phase asynchronous motor according to the coordinated adjustment instruction is specifically: The user terminal acquires real-time operation state parameters of the motor to which it belongs and performs conditional judgment, and the real-time operation state parameters include motor speed, motor temperature rise, and motor stator voltage and current. The operation state parameters of the motor are judged, and specifically: 1) The change rate of the motor speed in a preset time window is less than a set threshold; 2) Whether the motor temperature rise is within a preset allowable range; 3) Whether the deviation of the motor stator voltage and current from the rated value is not more than a preset threshold; When the operation state judgment result meets the above conditions, it is determined that the motor can safely perform the split-phase capacitor cutting operation; Under the condition that the motor operation condition is met and the coordinated adjustment instruction is required, the action of cutting off the split-phase capacitor of the single-phase asynchronous motor is performed, which is realized by disconnecting the capacitor access line, so that the motor power factor is reduced, thereby actively absorbing the excess reactive power of the power grid. When the real-time state or the predicted state does not meet the overvoltage and excess reactive power conditions, the user terminal restores the split-phase capacitor access state.

[0068] ​It should be noted that the single-phase asynchronous motor has a large amount of users in low-voltage electrical equipment, and the single-phase asynchronous motor starts by relying on the split-phase capacitor to generate starting torque, and after starting, it plays a role in compensating for reactive power and improving the power factor of the motor. After starting, the power factor is not considered, and the split-phase capacitor is disconnected, which basically does not affect the performance of the motor. The application uses this feature to propose the concept of virtual reactive power plant, and through the centralized reactive power control system, the command is wirelessly issued. When the local motor meets certain operating conditions, such as non-starting state, normal current, voltage and temperature rise, it has the ability to remove the split-phase capacitor to absorb excess reactive power. At this time, the application issues a command through a wireless network to remove the split-phase capacitor, reduces the power factor of the motor, and plays a role in optimizing the reactive power of the whole network, realizing the effect of virtual reactive power plant. Most reactive power regulation methods need to increase capacitors, reactors or electronic compensation equipment. The application ingeniously activates the controllable resources of the internal capacitor of the load equipment, and through in-depth regulation of the split-phase capacitor of the user motor, the reactive power can be changed and continuously changed accurately in a few seconds.

[0069] The above formulas are dimensionless values, and the formulas are obtained by collecting a large amount of data to simulate the latest real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0070] The above embodiments can be realized wholly or partially by software, hardware, firmware or any combination thereof. When realized by software, the above embodiments can be realized in the form of a computer program product wholly or partially.

[0071] Those skilled in the art can realize that the modules and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0072] In addition, the functional modules in each embodiment of the present application can be integrated in one processing module, or each module can exist physically alone, or two or more modules can be integrated in one module.

[0073] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0074] Finally: the above only for the preferred embodiments of the present application, and not for limiting the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application, should be included in the scope of protection of the present application.

Claims

1. A method for reactive power regulation of a single-phase asynchronous motor in a virtual reactive power plant user, characterized in that, include: Collect real-time status data of key nodes in the power distribution network, and build a short-term prediction model based on the status data to predict the trend of reactive power change and voltage change in the future. Based on the model's prediction results, the judgment conditions are set, a preliminary reactive power adjustment command is generated, and the priority and adjustment range of the preliminary reactive power adjustment command are adjusted in real time based on the dynamic update of the prediction trend. The adjusted reactive power regulation command is sent to the user terminal and processed collaboratively based on distributed coordination control to generate the final collaborative regulation command. According to the coordinated adjustment command, the action of cutting off the split-phase capacitor of the single-phase asynchronous motor is executed.

2. The reactive power regulation method for single-phase asynchronous motors in virtual reactive power plants according to claim 1, characterized in that, The process of collecting real-time status data from key nodes of the distribution network and constructing a short-term prediction model based on this data is as follows: The state data is assigned input feature weights based on an adaptive neural network and then weighted and fused to obtain comprehensive reactive power state characteristics; the state data includes instantaneous reactive power disturbance data, phase voltage nonlinear data, and load response delay data. A supervised learning dataset is constructed based on the comprehensive reactive power state characteristics and the future short-term reactive power and voltage. Construct a bidirectional LSTM network and input the supervised learning dataset into the bidirectional LSTM network for training to obtain a trained short-term prediction model. The real-time comprehensive reactive power state characteristics are input into the short-term prediction model, and the predicted values ​​of reactive power and voltage of the power grid in the short term are output.

3. The reactive power regulation method for single-phase asynchronous motors in virtual reactive power plants according to claim 2, characterized in that, The method of assigning input feature weights to the state data based on an adaptive neural network and then performing weighted fusion to obtain comprehensive reactive power state features is as follows: A standardized feature matrix is ​​generated based on state data, and the power grid reactive power and voltage in the near future are used as historical prediction targets. A neural network is constructed by inputting a standardized feature matrix and the corresponding historical prediction target into the network, and the error between the prediction result and the actual target is obtained by using the mean squared error loss function. The network weights are iteratively updated based on the prediction error using the backpropagation algorithm until the loss function converges. Feature weights are extracted to form an adaptive weight matrix, which is then weighted and fused with the standardized feature matrix to generate comprehensive reactive power state features.

4. The reactive power regulation method for single-phase asynchronous motors in virtual reactive power plants according to claim 3, characterized in that, The step of setting judgment conditions based on the model's prediction results and generating preliminary reactive power adjustment instructions specifically involves: Obtain the real-time voltage and real-time reactive power of the power grid nodes; The real-time voltage is compared with the preset voltage upper limit, and the real-time reactive power is compared with the preset power threshold to obtain the first judgment result; Obtain the predicted voltage and reactive power values ​​of the power grid nodes; The voltage prediction value is compared with the voltage upper limit, and the reactive power prediction value is compared with the power threshold to obtain a second determination result; Logical determination is performed based on the first determination result and the second determination result; When the first determination result indicates the existence of potential overvoltage and excess reactive power, or when the second determination result indicates the existence of overvoltage and excess reactive power, a preliminary reactive power adjustment command is generated.

5. The reactive power regulation method for single-phase asynchronous motors in virtual reactive power plants according to claim 4, characterized in that, The priority and adjustment range of the real-time adjustment of the initial reactive power regulation command are specifically as follows: The first data after the initial reactive power regulation instruction is generated is obtained, and the regulation priority index is calculated by combining the reactive power regulation weight associated with the first data. The reactive power regulation weight is then dynamically adjusted based on the Bayesian optimization algorithm. Multiple preliminary reactive power adjustment commands are sorted according to the adjustment priority index, and commands with higher priority are executed before commands with lower priority. The current operating status of the controlled motor is obtained, and the initial adjustment range is calculated by combining the first data with the preset initial adjustment range calculation formula. The initial adjustment range is optimized based on a robust control method to obtain the optimized adjustment range. Execute the corresponding reactive power regulation commands according to the sorted execution order and optimized adjustment range, and monitor the changes in grid voltage and reactive power in real time to obtain the deviation difference between the actual change value and the predicted change value. The adjustment priority and adjustment range of subsequent reactive power adjustment commands are dynamically adjusted based on the deviation difference to form the final reactive power adjustment command.

6. The reactive power regulation method for a single-phase asynchronous motor in a virtual reactive power plant according to claim 5, characterized in that, The dynamic adjustment of reactive power regulation weights based on the Bayesian optimization algorithm is specifically as follows: The reactive power regulation weight of the system is regarded as the variable to be optimized, and an optimization objective function is constructed. Based on historical operating data, a Gaussian process prior model is constructed for the reactive power regulation weight space; Select initial sampling points in the weight space, calculate the adjustment priority index, and perform adjustment operations; The value of the corresponding objective function is calculated based on the adjustment results, and the acquisition function is constructed based on the predicted mean and variance of the Gaussian process prior model. Select the weight combination with the largest sampling function value as the next sampling point, calculate the corresponding objective function value, and update the Gaussian process model; The above selection, calculation and update process is executed iteratively until the preset iteration termination condition is met; Select the reactive power adjustment weight combination that minimizes the objective function value from all iterative samples as the final optimization result.

7. The reactive power regulation method for single-phase asynchronous motors in virtual reactive power plants according to claim 5, characterized in that, The initial adjustment range is optimized using a robust control method to obtain the optimized adjustment range, specifically as follows: The second data is acquired and modeled as system uncertainty. The deviation between the current measured voltage value and the target voltage reference value is used as the index to construct a robust optimization objective function and safety constraints. The robust optimization objective function is solved using an optimization algorithm to ensure that it meets the safety constraints of voltage and reactive power while considering system uncertainties, thus obtaining the optimized regulation range.

8. The reactive power regulation method for a single-phase asynchronous motor in a virtual reactive power plant according to claim 7, characterized in that, The process of sending the adjusted reactive power regulation command to the user terminal and performing collaborative processing based on distributed coordination control to generate the final collaborative regulation command specifically involves: After receiving the final reactive power adjustment command, the user terminal obtains the current reactive power adjustment capability of the motor. Each user terminal sends its adjustable capabilities to neighboring nodes via a preset communication protocol, and simultaneously receives capability information from neighboring nodes. Based on its own adjustability and neighbor node information, the user terminal uses a consensus iterative algorithm to correct the execution ratio of the reactive power adjustment command; when the iterative calculation converges, each user terminal generates the final coordinated adjustment command according to the convergence result.

9. The reactive power regulation method for a single-phase asynchronous motor in a virtual reactive power plant according to claim 8, characterized in that, The consistency iterative algorithm is used to correct the execution ratio of the reactive power regulation command; after the iterative calculation converges, each user terminal generates the final coordinated regulation command based on the convergence result, specifically: Each user terminal calculates the initial adjustment ratio based on the received initial reactive power adjustment command and its own adjustment capability, and exchanges the adjustment ratio information with neighboring nodes through the communication network. Each user terminal updates its adjustment ratio iteratively according to the consistency iteration rules, based on the local adjustment ratio and the adjustment ratio information of neighboring nodes. After each iteration, the change in the local adjustment ratio is calculated and the change information is exchanged with neighboring nodes. When the change in the adjustment ratio of all neighboring nodes is less than the preset change threshold, the iteration is considered to have converged. The final coordinated adjustment command is generated by combining the adjustment ratio after iterative convergence with its own adjustability.

10. The method for reactive power regulation of a single-phase asynchronous motor in a virtual reactive power plant according to claim 9, characterized in that, The action of cutting off the split-phase capacitor of the single-phase asynchronous motor according to the coordinated adjustment command is specifically as follows: Obtain the real-time operating status parameters of the user terminal based on the motor it belongs to and make conditional judgments; 1) The rate of change of motor speed within a preset time window is less than a set threshold; 2) Whether the motor temperature rise is within the preset allowable range; 3) Whether the deviation of the motor stator voltage and current from their rated values ​​does not exceed the preset threshold; When all of the above conditions are met, it is determined that the motor can safely perform the split capacitor cut-off operation; Under the condition that the motor operating conditions are met and the requirements of the coordinated adjustment command are met, the action of cutting off the split-phase capacitor of the single-phase asynchronous motor is executed. When the real-time or predicted status does not meet the overvoltage and reactive power excess conditions, the user terminal restores the split-phase capacitor connection status.

Images