Three-phase unbalance degree monitoring and adjusting integrated method fused with edge calculation

Abstract

The invention relates to the technical field of power monitoring and adjustment, and discloses a three-phase unbalance degree monitoring and adjustment integrated method fused with edge computing, comprising the following steps: S1, deploying edge computing nodes; s2, acquiring three-phase voltage, current and active power data of the power distribution network through the power distribution terminal; s3, calculating a three-phase unbalance degree; s4, carrying out edge side real-time analysis and threshold judgment; s5, generating a dynamic adjustment strategy; and S6, adjusting instruction execution and feedback. According to the technical architecture of edge calculation enabling, dynamic adaptation optimization and hierarchical cooperative control, multi-dimensional improvement of monitoring precision, adjustment efficiency, operation economy and system collaboration is achieved for the pain points of three-phase unbalance monitoring lagging, extensive adjustment, low resource utilization rate and the like of a traditional power distribution network.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power monitoring and regulation, more particularly, it relates to a three-phase imbalance monitoring and regulation integrated method fusing edge computing. BACKGROUND

[0002] As a key link connecting the power generation side and the user side, the operation stability and power supply quality of the distribution network directly affect the continuity of industrial production and the experience of residential power consumption. In the actual operation of the distribution network, three-phase imbalance is a typical problem that exists universally, and its causes mainly include two aspects: one is the randomness and asymmetry of load characteristics. A large number of single-phase loads such as household appliances and commercial lighting equipment in the distribution network have strong randomness, and the load distribution often presents the characteristics of "single-phase concentration and three-phase imbalance", such as a single-phase connecting multiple residential power consumption and the other two-phase light load; the other is the large-scale access of distributed power sources. In recent years, photovoltaic and wind power and other distributed power sources are mostly connected to the grid in a single-phase or three-phase unbalanced manner, and their output is greatly affected by natural factors such as light and wind speed, which further aggravates the imbalance of three-phase voltage and current in the distribution network.

[0003] The traditional three-phase imbalance monitoring and regulation of the distribution network mostly adopts a centralized architecture of distribution terminal-cloud: after the distribution terminal collects three-phase voltage, current and other data, it needs to be transmitted to the cloud platform through the public network or the power dedicated network for a long distance, and the cloud platform completes the imbalance calculation, threshold judgment and regulation instruction generation. This architecture has two limitations: one is the significant transmission delay. The 10kV feeder line of the distribution network covers a wide range, and a large amount of electric parameter data transmitted to the cloud is easily limited by the communication bandwidth, resulting in a delay of more than 100ms for the data to reach the cloud; the other is the overload of cloud computing. After the data of all the distribution terminals in the network are converged, the cloud needs to simultaneously process the imbalance analysis tasks of multiple regions and multiple feeders, and the allocation of computing resources is strained, which further leads to the lag of regulation instruction generation. The above problems make the existing technology unable to cope with the "instantaneous three-phase imbalance" caused by rapid load fluctuation, and the regulation response delay often exceeds 500ms, missing the best regulation window. SUMMARY

[0004] To solve the above technical problems, the present application provides a three-phase imbalance monitoring and regulation integrated method fusing edge computing, which realizes the multi-dimensional improvement of monitoring accuracy, regulation efficiency, operation economy and system synergy through the technical architecture of edge computing empowerment, dynamic adaptive optimization and hierarchical collaborative control.

[0005] To achieve the above purpose, the technical scheme of the present application is as follows:

[0006] A three-phase imbalance monitoring and regulation integrated method fusing edge computing, comprising the following steps:

[0007] S1, edge computing node deployment; deploy edge computing nodes along the 10kV feeder line of the distribution network and at the low-voltage side of the distribution transformer, and the edge computing nodes establish real-time communication connection with the distribution terminal through Ethernet or 5G private network;

[0008] S2, collecting three-phase voltage 、 、 , three-phase current 、 、 and active power 、 、 data through the distribution terminal, the collection frequency is dynamically adjustable based on load characteristics , and the time synchronization between the edge computing node and the distribution terminal is realized through a time synchronization protocol;

[0009] S3, three-phase imbalance degree calculation; the edge computing node constructs a three-phase imbalance degree calculation model based on the symmetrical component method, and calculates the voltage imbalance degree and the current imbalance degree respectively;

[0010] S4, real-time analysis and threshold judgment on the edge side; the edge computing node compares the calculated imbalance degree with the preset threshold value, and if the threshold value is continuously exceeded, the dynamic adjustment process is triggered;

[0011] S5, dynamic adjustment strategy generation; the edge computing node constructs a double-objective optimization model based on real-time load data 、 、 and generates adjustment instructions;

[0012] S6, adjustment instruction execution and feedback; the edge computing node issues adjustment instructions to the SVG static var generator or three-phase load commutating switch, executes the adjustment, and then collects three-phase electrical parameters and calculates the imbalance degree again. If the imbalance degree is less than the threshold value and , one adjustment is completed; otherwise, return to S5 to regenerate the adjustment strategy.

[0013] As a preferred scheme of the present application, the dynamically adjustable collection frequency in S2 is adaptively adjusted based on load fluctuation rate, load type characteristics and seasonal period characteristics during real-time collection of three-phase electrical parameters, specifically including:

[0014] (1) real-time calculation of load fluctuation rate: the edge computing node continuously calculates the fluctuation rate of three-phase total active power within the collection period:

[0015]

[0016] wherein is the fluctuation rate of three-phase total active power Total active power at the moment; is the length of the previous data collection period, and the current sampling rate corresponds;

[0017] (2) Sampling rate dynamic switching logic: preset high fluctuation rate threshold and low fluctuation rate threshold , wherein > At the same time, a load type correction coefficient and a seasonal period correction coefficient are introduced, and after each collection is completed, according to the comparison results of the current and , , the adjustment is triggered:

[0018] When , the sampling rate is switched to the highest sampling rate ;

[0019] When , the sampling rate is switched to the lowest sampling rate , and ;

[0020] When , the linear interpolation formula is used for adjustment:

[0021]

[0022] In the formula, k is a proportional coefficient, and the adjusted takes effect in the next collection period.

[0023] As a preferred scheme of the present application, the number of deployment of the edge computing node in S1 is determined in advance based on the maximum sampling rate demand corresponding to the historical load fluctuation characteristics of the target area, and the function is:

[0024]

[0025] In the formula:

[0026] is the maximum possible sampling rate based on the statistical load data of the past three years;

[0027] is the maximum data processing capacity of a single edge node under ;

[0028] is the load density of the target area power distribution network, ​Total length of 10kV feeder, Average area covered by a single feeder;

[0029] Annual load growth rate of target area; Design service life of edge node, i.e. reserve Processing margin of annual load growth; Redundancy correction coefficient; Upward rounding function.

[0030] As a preferred scheme of the present application, the three-phase unbalance degree calculation model in S3 introduces a load weight factor, and the improved current unbalance degree calculation formula is:

[0031]

[0032] In the formula:

[0033] Average value of three-phase current;

[0034] , , Load weight factors of phase A, B and C respectively;

[0035] The load weight factor is positively correlated with the active power fluctuation rate, and the fluctuation rate is calculated based on the collected data under the current sampling rate:

[0036] Reference weight, Weight coefficient; when Approaches , It is automatically increased, so that the threshold judgment of S4 is more focused on the balance control of the high fluctuation phase.

[0037] As a preferred scheme of the present application, the edge computing node in S4 calculates the total data processing delay during real-time analysis, and the delay is related to the current sampling rate:

[0038]

[0039] Among them, Distribution terminal data acquisition delay, Transmission delay of data from distribution terminal to edge node, Unbalance degree calculation delay of edge node, Collection time consumption under the current sampling rate, Calculation time consumption under the current sampling rate; if Exceeds the preset threshold​​​ , then the sampling rate adjustment mechanism of S2 is called to reduce the sampling rate before triggering the regulation procedure in S4 , ensuring the real-time performance of S5 to generate the regulation instruction is not affected by the delay.

[0040] As a preferred embodiment of the present application, the dynamic regulation strategy generation in S5 adopts a model predictive control (MPC) algorithm, and the objective function and constraint conditions of the MPC are as follows:

[0041] Objective function:

[0042]

[0043] Constraint conditions:

[0044] Regulation amount constraint: , where is the reactive power compensation amount of the SVG or the switching frequency of the phase-change switch;

[0045] Imbalance constraint: , ;

[0046] In the formula: is the prediction time domain, which is 5-10 sampling periods; is the control time domain, which is 2-5 sampling periods; is the target imbalance, which is 1%-2%; is the regulation energy consumption weight coefficient, which is 0.1-0.3; is the change amount of the regulation amount at adjacent time points;

[0047] Prediction time domain of the model predictive control (MPC) algorithm Based on the current sampling rate of S2 Dynamic adaptation:

[0048]

[0049] When is higher, the edge computing node can obtain denser load data samples, and automatically expand to improve the prediction accuracy; when is lower, shorten to reduce the amount of calculation, ensuring that the strategy generation efficiency of S5 matches the sampling efficiency of S2.

[0050] As a preferred embodiment of the present application, in S6, the calculation of the regulation efficiency index introduces a sampling rate correction term based on the sampling rate when the regulation instruction is executed:

[0051]

[0052] In the formula, is the reference efficiency based on the difference of unbalance degree before and after adjustment, is a correction coefficient, 0 ≤0.3; when the adjustment is completed at a high sampling rate and meets the standard, the correction term makes increase, for optimizing the sampling rate adjustment strategy of S2.

[0053] As a preferred scheme of the present application, it further comprises S7, edge-cloud collaborative optimization; the edge computing node uploads the statistical data of the three-phase unbalance degree in the region every interval , wherein 5min≤ ≤15min; the average , maximum , and adjustment times of the regional three-phase unbalance degree are uploaded to the power distribution network cloud platform; when the cloud platform detects that a certain region has three consecutive periods of unbalance degree exceeding the threshold, global optimization is triggered, and the threshold parameters of the edge nodes in the region are adjusted by a particle swarm optimization (PSO) algorithm , and the weight of the MPC target function ; the fitness function of the PSO is:

[0054]

[0055] In the formula: is the regional average current unbalance degree; is the regional average adjustment time consumption; is the regional adjustment times per unit time; , , are weight coefficients, satisfying , and .

[0056] As a preferred scheme of the present application, the dynamic adjustment strategy generation in S5 is further combined with short-term load prediction results; the load prediction adopts an ARIMA( , , ) model, and the prediction formula is:

[0057]

[0058] In the formula: is the predicted value of the load of a phase at a future time;

[0059] is the number of autoregression terms, is the order of difference, is the number of moving average terms; is an autoregressive coefficient, is a moving average coefficient; is a prediction error; the edge computing node adjusts the reserved amount of the adjustment instruction in advance according to the load prediction value, to avoid imbalance degree rebound caused by load mutation.

[0060] As a preferred scheme of the present application, the double-target optimization model in S5 takes minimizing three-phase imbalance degree and minimizing adjustment energy consumption as targets, and the objective function is:

[0061] In the formula, is a current current imbalance degree, is an energy consumption index of the adjustment device, is a weight coefficient, , and increases with the increase of the sampling rate in S2.

[0062] The beneficial technical effects of the present application are:

[0063] The present application realizes multi-dimensional improvement of monitoring accuracy, adjustment efficiency, operation economy and system collaboration through the technical architecture of edge computing empowerment, dynamic adaptive optimization and hierarchical collaborative control, aiming at the pain points of traditional power distribution network three-phase imbalance degree monitoring lag, adjustment extensive, low resource utilization rate, etc.

[0064] Localized processing reduces delay: the edge node is close to the power distribution terminal, realizes data acquisition, calculation and analysis localization through Ethernet or 5G private network, avoids transmission and computing power congestion of the “terminal-cloud” architecture, shortens the whole link time consumption; the number of computing nodes takes into account the load characteristics and node processing capacity, considers the future load growth trend and the redundancy demand of edge nodes, to ensure the real-time performance of the adjustment instruction.

[0065] Dynamic sampling balances accuracy and resources: the sampling rate is adjusted according to the load fluctuation rate, load type and seasonal period; the highest sampling rate is used to accurately capture abnormalities in high fluctuation, and the lowest sampling rate is used to save resources in low fluctuation; the current imbalance degree introduces a load weight factor that is positively correlated with active power fluctuation rate, focuses on high-risk phases and reduces false and missed judgments; through the time synchronization protocol, the collection data time stamp is aligned to improve the imbalance degree calculation accuracy.

[0066] The MPC algorithm combines ARIMA load prediction to predict imbalance degree and load change in advance, adjust the adjustment reserve amount, and reduce repeated adjustment; takes minimizing imbalance degree and minimizing adjustment energy consumption as targets, and the weight coefficient increases with the increase of the sampling rate, balances the accuracy and energy saving demand; sets the adjustment amount constraint to avoid device overload or loss.​​

[0067] Regulation efficiency index Includes a sampling rate correction term, which improves performance when the high sampling rate meets the target. Optimize subsequent sampling rate adjustments; dynamically adjust parameters such as MPC prediction time domain and load weight factor to adapt to different load characteristics and reduce manual intervention.

[0068] Dynamic sampling reduces resource consumption, and precise node deployment avoids hardware waste and reduces construction costs; it also reduces ineffective energy consumption of regulating equipment, meets energy-saving requirements, and lowers the operating costs of the power distribution network. Attached Figure Description

[0069] Figure 1 This is a flowchart illustrating the present invention. Detailed Implementation

[0070] In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0071] Combination Figure 1 The present invention provides the following embodiments:

[0072] An integrated method for monitoring and regulating three-phase imbalance using edge computing includes the following steps:

[0073] S1. Edge computing node deployment: Edge computing nodes are deployed along the 10kV feeder of the power distribution network and on the low-voltage side of the power distribution transformer. The edge computing nodes establish real-time communication connections with the power distribution terminal through Ethernet or 5G private network.

[0074] S2. Collect the three-phase voltage of the power distribution network through the power distribution terminal. , , Three-phase current , , and active power , , The data collection frequency is dynamically adjustable based on load characteristics. And time synchronization between edge computing nodes and power distribution terminals is achieved through a time synchronization protocol;

[0075] S3. Three-phase unbalance calculation; the edge computing nodes construct a three-phase unbalance calculation model based on the symmetrical component method, and calculate the voltage unbalance and current unbalance respectively, where:

[0076] The formula for calculating voltage unbalance is:

[0077]

[0078] wherein, is the positive sequence voltage effective value, is the negative sequence voltage effective value, obtained by Fourier transform decomposition of three-phase voltage sampling values , ,

[0079] The current unbalance degree calculation formula is:

[0080]

[0081] wherein, is the positive sequence current effective value, is the negative sequence current effective value, obtained by Fourier transform decomposition of three-phase current sampling values , ,

[0082] S4, edge side real-time analysis and threshold judgment; the edge computing node compares the calculated unbalance degree , with a preset threshold , If it continuously exceeds the threshold, the dynamic adjustment process is triggered.

[0083] S5, dynamic adjustment strategy generation; the edge computing node constructs a double-objective optimization model based on real-time load data , , to generate adjustment instructions.

[0084] S6, adjustment instruction execution and feedback; the edge computing node issues the adjustment instructions to the SVG static var generator or three-phase load commutating switch, executes the adjustment, and then collects three-phase electrical parameters again and calculates the unbalance degree. If the unbalance degree after adjustment and , one adjustment is completed; otherwise, return to S5 to regenerate the adjustment strategy.

[0085] Further, the dynamically adjustable sampling frequency in S2 is adjusted adaptively based on load fluctuation rate, load type characteristics, and seasonal period characteristics during real-time collection of three-phase electrical parameters, specifically including:

[0086] (1) Real-time calculation of load fluctuation rate: the edge computing node continuously calculates the fluctuation rate of three-phase total active power within the collection period:

[0087]

[0088] ​​In the formula, is the total active power of three-phase at the moment; is the length of the previous data collection period, and the current sampling rate corresponds.

[0089] (2) Dynamic switching logic of sampling rate: preset high fluctuation rate threshold and low fluctuation rate threshold , wherein > At the same time, the load type correction coefficient and the seasonal period correction coefficient are introduced, and after each collection is completed, according to the comparison results of the current and , , the adjustment is triggered:

[0090] When , the sampling rate is switched to the highest sampling rate ;

[0091] When , the sampling rate is switched to the lowest sampling rate , and ;

[0092] When , the linear interpolation formula is used for adjustment:

[0093]

[0094] In the formula, is the proportional coefficient, and the adjusted takes effect in the next collection period.

[0095] The load type correction coefficient value range and corresponding scene: industrial load , commercial load , residential load , used to quantify the fluctuation sensitivity of different load types; the seasonal period correction coefficient value range and corresponding scene: summer / winter , spring / autumn , peak time (7:00-22:00) , valley time (22:00-7:00 the next day) , used to quantify the influence of season and period on load fluctuation.

[0096] Through adaptive matching of sampling rate, the total active power fluctuation rate is used to reflect the load stability. When the fluctuation is high, such as the sudden increase of resident electricity consumption during the morning peak, the highest sampling rate is switched to accurately capture the instantaneous unbalanced data and avoid missed judgment. When the fluctuation is low, such as the low load during the night, the lowest sampling rate is switched to reduce the data processing amount and energy consumption of the edge node. When the fluctuation is moderate, linear interpolation is used for transition, which ensures the monitoring accuracy and avoids resource waste, and realizes dynamic balance.

[0097] Further, the number of deployed edge computing nodes in S1 is determined in advance based on the maximum sampling rate demand corresponding to the historical load fluctuation characteristics of the target area, and the function is:

[0098]

[0099] In the formula:

[0100] is the maximum possible sampling rate based on the statistical load data of the past three years;

[0101] is the maximum data processing capacity of a single edge node in ;

[0102] is the load density of the target area power distribution network, is the total length of the 10kV feeder, is the average area covered by a single feeder;

[0103] is the annual load growth rate of the target area (based on regional power grid planning data, the value range is 0.05-0.1), which is used to quantify the demand for node processing capacity due to future load growth;

[0104] is the designed service life of the edge node (the value range is 5-10 years), that is, the processing margin for years of load growth is reserved;

[0105] is the redundancy correction coefficient (the value range is 1.2-1.5), which is used to cope with the data takeover demand when the node fails (when a node fails, other nodes need to take over 10%-30% of its data processing amount);

[0106] is the ceiling function, which ensures that the deployed edge node can still meet the real-time processing demand under the actual sampling, load growth and failure scenarios in the future.

[0107] By inferring the maximum sampling rate requirement from historical load data and calculating the number of nodes to deploy based on node processing capacity, we can avoid the problems of hardware redundancy due to too many nodes or insufficient nodes to handle high sampling rate data. Rounding up ensures that even when the maximum sampling rate is triggered, edge nodes can still process data in real time, achieving precise matching between node resources and monitoring needs, and reducing the construction cost of the distribution network edge side.

[0108] Furthermore, the three-phase unbalance calculation model described in S3 introduces a load weighting factor, and the improved current unbalance calculation formula is as follows:

[0109]

[0110] In the formula:

[0111] This is the average value of the three-phase current;

[0112] , , These are the load weighting factors for phases A, B, and C, respectively.

[0113] Load weighting factor and active power volatility Positive correlation, and the volatility is based on the current sampling rate. The following data collection calculations: ;

[0114] The baseline weight is ≥1. Weighting coefficients (0 < ≤1); when near (correspond When it is relatively large, The threshold of S4 is automatically increased, making the threshold judgment of S4 more focused on the balance control of the high fluctuation phase.

[0115] By focusing on high-risk phases and phases with high active power fluctuations through load weighting factors, such as single phases connected to multiple households, the weight is automatically increased. When calculating the imbalance, more emphasis is placed on the deviation of this phase, avoiding the omission of imbalance in high-fluctuation phases due to the flattening of data in low-fluctuation phases. This improves the pertinence and accuracy of imbalance calculation and reduces erroneous adjustments.

[0116] Furthermore, in S4, the edge computing nodes calculate the total data processing latency during real-time analysis, and this latency is related to the current sampling rate:

[0117]

[0118] in, Due to the delay in data acquisition from the power distribution terminal, For the data transmission delay from the power distribution terminal to the edge node, Calculate the delay for the imbalance of edge nodes. The acquisition time at the current sampling rate (and) (inversely proportional) The computation time at the current sampling rate (and) (proportional); if Exceeding the preset threshold Before S4 triggers the adjustment process, the sampling rate adjustment mechanism of S2 is called first to reduce the sampling rate. This ensures that the real-time performance of the adjustment commands generated by S5 is not affected by delays.

[0119] By dynamically controlling the sampling rate through total latency calculation, the problem of high sampling rate causing the superposition of calculation and acquisition delays and missing the adjustment window is solved. When the latency exceeds the threshold, the sampling rate is reduced first to ensure the real-time generation of adjustment instructions. This avoids the defect of data congestion causing adjustment lag in traditional architecture and improves the response speed of instantaneous imbalance.

[0120] Furthermore, the dynamic adjustment strategy generation described in S5 employs a Model Predictive Control (MPC) algorithm. The objective function and constraints of MPC are as follows:

[0121] Objective function:

[0122]

[0123] Constraints:

[0124] Adjustment constraint: ,in This refers to the reactive power compensation of the SVG or the number of switching operations of the commutator.

[0125] Imbalance constraints: , ;

[0126] In the formula: For prediction in the time domain, its value ranges from 5 to 10 sampling periods; To control the time domain, its value is set to 2 to 5 sampling periods; The target imbalance is defined as 1% to 2%. To adjust the energy consumption weighting coefficient, its value ranges from 0.1 to 0.3; The change in adjustment amount between adjacent time points;

[0127] Prediction time domain of Model Predictive Control (MPC) algorithm Based on the current sampling rate of S2 Dynamic adaptation:

[0128]

[0129] when When the sampling rate is higher, the edge computing node can obtain denser load data samples, automatically expand to improve prediction accuracy; when the sampling rate is lower, shorten to reduce the amount of calculation, ensure that the strategy generation efficiency of S5 matches the sampling efficiency of S2.

[0130] When the sampling rate is low shorten, reduce the amount of calculation to avoid resource waste, realize the adaptation of prediction accuracy, calculation efficiency and sampling rate, and improve the rationality and generation speed of the adjustment strategy.

[0131] Further, in S6, the calculation of the adjustment efficiency index introduces a sampling rate correction term, which is based on the sampling rate when the adjustment is performed

[0132]

[0133] In the formula, is the reference efficiency based on the difference in unbalance degree before and after adjustment, is the correction coefficient (0 ≤0.3); when the adjustment is completed at a high sampling rate (such as ) and meets the standard, the correction term improves, which is used to optimize the sampling rate adjustment strategy of S2.

[0134] Optimize efficiency evaluation through sampling rate correction term: when the sampling rate is high and meets the standard, improvement can provide feedback incentives for subsequent high sampling rate accurate adjustment, and provide basis for S2 sampling rate adjustment, forming a closed loop of sampling-adjustment-efficiency feedback, and improving the overall adjustment efficiency of the system.

[0135] Further, it also includes S7, edge-cloud collaborative optimization; the edge computing node every interval , where 5min≤ ≤15min; the statistical data of three-phase unbalance degree in the region is averaged , maximum , and the number of adjustments is uploaded to the power distribution network cloud platform; when the cloud platform detects that a certain region has for 3 consecutive periods, trigger global optimization, adjust the threshold parameters (a, b) of the edge node in the region and the weight of the MPC objective function through particle swarm optimization (PSO), the fitness function of PSO is: ,

[0136]

[0137] In the formula:​​ is the regional average current imbalance; is the regional average regulation time consumption; is the number of regulation times per unit time in the region; , , are weight coefficients, satisfying , and , and priority is given to ensuring the optimization of the imbalance. To achieve the global balance control of the regional distribution network and improve the overall operation stability.

[0138] Furthermore, the generation of the dynamic regulation strategy in S5 also combines the short-term load forecasting results. The load forecasting uses the ARIMA(p,d,q) model, and the forecasting formula is:

[0139]

[0140] In the formula: is the predicted value of the load of a certain phase at the future (5min ≤ T_{} ≤ 10min) moment;

[0141] is the number of autoregressive terms, is the order of differencing, is the number of moving average terms, determined by the AIC criterion, usually , , ; is the autoregressive coefficient, is the moving average coefficient; is the prediction error, following a normal distribution ; The edge computing node adjusts the reserve of the regulation instruction in advance according to the load prediction value. For example, the SVG reactive power compensation amount increases by 5% - 10% to avoid the rebound of the imbalance caused by the load mutation.

[0142] Furthermore, the double-objective optimization model in S5 aims to minimize the three-phase imbalance and minimize the regulation energy consumption. The objective function is: ;

[0143] In the formula, is the current current imbalance, is the energy consumption index of the regulation equipment, , are weight coefficients, , and increases with the increase of the sampling rate in S2, and the balance accuracy is preferentially ensured at high sampling rates.

[0144] Dynamically match the sampling scenario through the weight coefficient. At high sampling rates Increase, priority to ensure that the unbalance degree is reduced, the focusing accuracy; under low sampling rate Relative increase, taking into account the regulation energy consumption control, focusing on energy saving, balance in different sampling scenarios, improve the economic efficiency of power distribution network operation.

[0145] The following is an embodiment step of the present application:

[0146] (I) edge computing node deployment

[0147] The maximum preset sampling rate is calculated: based on the historical 3-year load data, the maximum load fluctuation rate in summer evening peak is 15%, and the corresponding maximum preset sampling rate is .

[0148] The number of computing node deployment is calculated:

[0149]

[0150] Substitute parameters:

[0151]

[0152] Deployment location: deploy one edge computing node at the midpoint and terminal distribution transformer low voltage side of the feeder, communicate with the power distribution terminal through the 5G private network, and the time delay is measured to be less than 20ms.

[0153] (II) dynamic collection of three-phase electric parameters

[0154] Load fluctuation rate calculation: at a certain time , the three-phase active power , , is collected, the previous collection period (corresponding to the current sampling rate ), The total active power at time .

[0155] Total active power:

[0156]

[0157] Load fluctuation rate:

[0158]

[0159] Sampling rate dynamic switching: preset high fluctuation rate threshold , low fluctuation rate threshold , current , Then:

[0160]

[0161] Because (5.04%<4.44% not true, actually ), switch to the lowest sampling rate Here is the correction: actually calculated , so:

[0162]

[0163] Time synchronization: use IEEE1588 PTP protocol, edge node and power distribution terminal time synchronization error <1ms.

[0164] (Three), three-phase unbalance degree calculation

[0165] Collect three-phase current: , , , three-phase current average:

[0166]

[0167] Calculate the active power fluctuation rate of each phase:

[0168]

[0169] Calculate the load weight factor: , :

[0170]

[0171]

[0172] Improved current unbalance degree:

[0173]

[0174] Substitute data:

[0175]

[0176]

[0177] (Four), real-time analysis and threshold judgment on the edge side

[0178] Calculate the total delay of data processing: current , acquisition delay , transmission delay , calculation delay :

[0179]

[0180] Pre-set delay threshold , , Without adjusting the sampling rate.

[0181] Threshold comparison: current , and 3 consecutive sampling periods (0.3s) exceed the threshold, triggering the dynamic adjustment process.

[0182] (Five) S5: Dynamic adjustment strategy generation

[0183] ARIMA load prediction: ARIMA(2,1,1) model is used to predict future each phase load:

[0184]

[0185] Substitute the coefficient , , , , the predicted , , , reserve 10% SVG compensation.

[0186] MPC target function and constraints:

[0187] Prediction time domain : current , , :

[0188]

[0189] Control time domain , target imbalance , energy consumption weight :

[0190]

[0191] Constraint condition: SVG reactive power compensation , ; , .

[0192] Dual-objective optimization: (current is low, and energy consumption is considered), and :

[0193]

[0194] The solution is to adjust the command: SVG compensates for phase B , and the commutating switch switches 3 residential loads from phase A to phase B.

[0195] (vi) Execution and feedback of adjustment instructions

[0196] Command issuance and execution: The edge node issues commands to the LX-B03 transformer SVG (model: SVG-0.4 / 200) and load phase switching switch via the 5G private network, with an execution time of 0.08s.

[0197] Secondary data acquisition and calculation: after adjustment , , , ; , , , The standard has been met.

[0198] Regulation efficiency calculation: baseline efficiency (Based on the reduction in imbalance), correction coefficient :

[0199]

[0200] (vii) Edge-cloud collaborative optimization

[0201] Data upload: Each edge node Upload data to the Laixi District power distribution network cloud platform: average within the period ,maximum Adjustment times = 2.

[0202] Global optimization trigger judgment: 3 consecutive Within the period No need to trigger PSO global optimization.

[0203] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An integrated method for monitoring and regulating three-phase imbalance using edge computing, characterized in that, Includes the following steps: S1. Edge computing node deployment: Edge computing nodes are deployed along the 10kV feeder of the power distribution network and on the low-voltage side of the power distribution transformer. The edge computing nodes establish real-time communication connections with the power distribution terminal through Ethernet or 5G private network. S2. Collect the three-phase voltage of the power distribution network through the power distribution terminal. , , Three-phase current , , and active power , , The data collection frequency is dynamically adjustable based on load characteristics. And time synchronization between edge computing nodes and power distribution terminals is achieved through a time synchronization protocol; S3, Three-phase unbalance calculation; The edge computing nodes construct a three-phase unbalance calculation model based on the symmetrical component method to calculate the voltage unbalance and current unbalance respectively; S4. Real-time edge analysis and threshold determination; The edge computing nodes will calculate the imbalance. , With preset threshold , In contrast, if the threshold is exceeded continuously, a dynamic adjustment process is triggered. S5, dynamic adjustment strategy generation; edge computing nodes based on real-time load data. , , Construct a dual-objective optimization model and generate adjustment instructions; S6. Execution and Feedback of Adjustment Commands: The edge computing node sends adjustment commands to the SVG static var generator or three-phase load commutation switch. After the adjustment is executed, the three-phase electrical parameters are collected again and the imbalance is calculated. If the adjustment is successful... and If the adjustment is successful, then one adjustment is completed; otherwise, return to S5 to regenerate the adjustment strategy.

2. The integrated method for monitoring and adjusting three-phase imbalance using edge computing as described in claim 1, characterized in that, The dynamically adjustable sampling frequency described in S2 During the real-time acquisition of three-phase electrical parameters, adaptive adjustments are made based on load fluctuation rate, load type characteristics, and seasonal characteristics, specifically including: (1) Real-time calculation of load fluctuation rate: The edge computing node continuously calculates the fluctuation rate of the total active power of the three phases during the acquisition period: In the formula, for Total active power of three phases at all times; The duration of the previous data collection period, and the current sampling rate. correspond; (2) Dynamic switching logic for sampling rate: preset high volatility threshold and low volatility threshold ,in > At the same time, a load type correction factor is introduced. Seasonal adjustment factor After each data collection is completed, based on the current... and , The comparison results triggered Adjustment: when When the sampling rate is switched to the highest sampling rate... ; when At that time, the sampling rate is switched to the lowest sampling rate. ,and ; when When adjusting, a linear interpolation formula is used: In the formula, It is a proportionality coefficient, and the adjusted... It will take effect in the next data collection cycle.

3. The integrated method for monitoring and adjusting three-phase imbalance using edge computing as described in claim 1, characterized in that, The number of edge computing nodes deployed as described in S1 The maximum sampling rate requirement corresponding to the historical load fluctuation characteristics of the target area is pre-calculated and determined by the function: In the formula: This represents the maximum possible sampling rate based on three years of historical load data. For a single edge node in Maximum data processing capacity; For the target area distribution network load density, This refers to the total length of the 10kV feeder. The average area covered by a single feeder; The target area's annual load growth rate; Design the service life of edge nodes, i.e., reserve time for them. Annual load growth in processing capacity; This is a redundancy correction factor; This is the floor function.

4. The integrated method for monitoring and adjusting three-phase imbalance using edge computing as described in claim 1, characterized in that, The three-phase unbalance calculation model described in S3 introduces a load weighting factor, and the improved current unbalance calculation formula is as follows: In the formula: This is the average value of the three-phase current; , , These are the load weighting factors for phases A, B, and C, respectively. Load weighting factor and active power volatility Positive correlation, and the volatility is based on the current sampling rate. The following data collection calculations: ; As the benchmark weight, For weighting coefficients; when near hour, The threshold of S4 is automatically increased, making the threshold judgment of S4 more focused on the balance control of the high fluctuation phase.

5. The integrated method for monitoring and adjusting three-phase imbalance using edge computing as described in claim 1, characterized in that, In S4, edge computing nodes calculate the total data processing latency during real-time analysis, and this latency is related to the current sampling rate: in, Due to the delay in data acquisition from the power distribution terminal, For the data transmission delay from the power distribution terminal to the edge node, Calculate the delay for the imbalance of edge nodes. This represents the acquisition time at the current sampling rate. This represents the computation time at the current sampling rate; if Exceeding the preset threshold Before S4 triggers the adjustment process, the sampling rate adjustment mechanism of S2 is called first to reduce the sampling rate. This ensures that the real-time performance of the adjustment commands generated by S5 is not affected by delays.

6. The integrated method for monitoring and adjusting three-phase imbalance using edge computing as described in claim 1, characterized in that, The dynamic adjustment strategy described in S5 is generated using the Model Predictive Control (MPC) algorithm. The objective function and constraints of MPC are as follows: Objective function: Constraints: Adjustment constraint: ,in This refers to the reactive power compensation of the SVG or the number of switching operations of the phase commutator. Imbalance constraints: , ; In the formula: For prediction in the time domain, its value ranges from 5 to 10 sampling periods; To control the time domain, its value is set to 2 to 5 sampling periods; The target imbalance is defined as 1% to 2%. To adjust the energy consumption weighting coefficient, its value ranges from 0.1 to 0.3; The change in adjustment amount between adjacent time points; Prediction time domain of Model Predictive Control (MPC) algorithm Based on the current sampling rate of S2 Dynamic adaptation: when At higher speeds, edge computing nodes can acquire denser samples of load data. Automatic expansion to improve prediction accuracy; when At lower levels, Shorten the timeframe to reduce computational load and ensure that the policy generation efficiency of S5 matches the sampling efficiency of S2.

7. The integrated method for monitoring and adjusting three-phase imbalance using edge computing as described in claim 1, characterized in that, In S6, the calculation of the adjustment efficiency index introduces a sampling rate correction term, which is based on the sampling rate at the time of adjustment command execution. : In the formula, The baseline efficiency is based on the difference in imbalance before and after adjustment. For correction factors, 0 < ≤0.3; When the adjustment is completed and meets the target at a high sampling rate, the correction term makes... Improvements are made to optimize the sampling rate adjustment strategy for S2.

8. The integrated method for monitoring and adjusting three-phase imbalance using edge computing as described in claim 1, characterized in that, It also includes S7, edge-cloud collaborative optimization; edge computing nodes per interval , where 5min≤ ≤15min; average the statistical data of three-phase imbalance in the region. ,maximum The number of adjustments is uploaded to the distribution network cloud platform; when the cloud platform detects three consecutive adjustments in a certain area... Within the period When this occurs, global optimization is triggered, and the threshold parameters of the edge nodes in that region are adjusted using the Particle Swarm Optimization (PSO) algorithm. , Weights of the MPC objective function The fitness function of PSO is: In the formula: The regional average current imbalance; The average adjustment time for the region; The number of adjustments per unit time for the region; , , For the weighting coefficients, satisfying ,and .

9. The integrated method for monitoring and adjusting three-phase imbalance using edge computing as described in claim 6, characterized in that, The dynamic adjustment strategy described in S5 also incorporates short-term load forecasting results, with load forecasting employing ARIMA (…). , , The model's prediction formula is: In the formula: For the future The predicted load value for a certain phase at a given time; The number of autoregressive terms. It is the difference order. The number of terms in the moving average; These are the autoregressive coefficients. The moving average coefficient; To mitigate prediction errors, edge computing nodes adjust the reserve amount of adjustment commands in advance based on load forecast values ​​to avoid imbalance rebound caused by sudden load changes.

10. The integrated method for monitoring and adjusting three-phase imbalance using edge computing as described in claim 6, characterized in that, The dual-objective optimization model described in S5 aims to minimize the three-phase imbalance and the regulation energy consumption. The objective function is: ; In the formula, The current current imbalance. To adjust the energy consumption index of the equipment, , These are the weighting coefficients. ,and With the sampling rate in S2 It increases with the rise in temperature.

Images