Power distribution network optimization method based on hierarchical robust control and dynamic decision

Abstract

The invention provides a power distribution network optimization method based on hierarchical robust control and dynamic decision, which realizes multi-time scale coordination control and real-time adaptive control by constructing a hierarchical architecture of a minute-level equipment layer, an hour-level region layer and a day-preceding-level system layer and combining a dynamic information decision mechanism. The method comprises the steps of preprocessing measurement information in a grading manner, establishing a multi-target robust optimization model, decomposing the model into a local quick response and global coordination problem, constructing a time-varying uncertainty set and embedding mixed integer robust optimization, and realizing the collaboration of millisecond-level correction and hour-level scheduling by adopting a second-order cone relaxation technology. And the weight is dynamically adjusted and optimized through the information decision module, so that the robustness and the operation efficiency of the system are improved.

Description

Technical Field

[0001] The present invention relates to a distribution network optimization method based on hierarchical robust control and dynamic information decision-making. Background Art

[0002] With the sustainable development of society, building a new power system with clean energy as the main body is the future development trend. The power industry must actively transform the energy production and consumption modes, optimize the energy structure, and increase the development intensity of renewable energy. With the high proportion access of renewable energy and the wide deployment of power electronic devices, modern distribution networks are facing severe challenges such as a sharp increase in uncertainty, the complexity of control dimensions, and multi-time scale coupling.

[0003] Specifically, the following technical bottlenecks are gradually exposed when traditional distribution network optimization methods cope with these new situations: 1) The response lag problem of the centralized optimization architecture. The solution time of the whole network model often reaches dozens of minutes, and it cannot respond to scenarios such as the second-level fluctuation of photovoltaic output and the minute-level change of electric vehicle charging load. The timeliness difference between the millisecond-level measurement data (voltage / current) of the device layer (inverter / energy storage) and the hourly-level dispatching instruction of the system layer is not distinguished, resulting in the mismatch between the control instruction and the actual working condition; 2) The insufficient adaptability of single-time scale optimization. Only focusing on local voltage stability may deviate from the economic dispatching goal of the system layer, leading to an increase in network loss. The execution ability constraint of the device layer is not considered, resulting in an infeasible optimization result; 3) The photovoltaic output fluctuation can reach ±30% when the noon irradiation is strong, while it may suddenly drop to ±5% in rainy weather. Fixed boundary sets are prone to over-conservatism or under-robustness. The existing technologies do not effectively distinguish the fluctuation propagation characteristics of different power sources such as photovoltaic (inertialess) and wind power (weak inertia), and use a unified uncertainty set to reduce the model accuracy. To sum up, there are significant deficiencies in the architecture design, model adaptability, dynamic decision-making, etc. of the existing distribution network optimization technologies. Therefore, it is necessary to design a new method to meet the actual needs. Summary of the Invention

[0004] The present invention considers the uncertain factors in the distribution network under high-penetration conditions, classifies the measured quantities, and constructs an uncertainty set on the basis of the original deterministic optimization model of the distribution network, so that the adaptability of the dispatching scheme is higher.

[0005] The present invention is implemented by the following technical means: A distribution network optimization method based on hierarchical robust control and dynamic information decision-making, including the following steps:

[0006] Step 1: Considering the multi-time scale control requirements, data acquisition frequencies, and optimization goal differences, divide the distribution network into a device layer, a regional layer, and a system layer, respectively perform hierarchical preprocessing on millisecond-level measurement data (voltage / current), minute-level prediction data (photovoltaic output), and hourly-level dispatching instructions, and implement classification and fusion of the data based on the power source type (controllable / uncontrollable) and control level;

[0007] Step 2: Construct a cross - level training data set based on the real - time inverter status of the device layer, the operating data of energy storage SOC, the predicted values of distributed power generation output at the regional layer from the long - short - term memory (LSTM) prediction results combined with irradiance and wind speed, and the statistical data of load demand at the system layer. Establish a hierarchical multi - objective robust optimization model, where the robust model predictive control (RMPC) is used at the device layer and the alternating direction method of multipliers (ADMM) is used at the system layer;

[0008] Step 3: Improve the hierarchical optimization model using second - order cone relaxation technology. Define the time - varying boundaries of the control variables at the device layer (DG output of distributed generation, charge and discharge of energy storage) and the state variables at the system layer (node voltage, branch power flow). Embed the system model to coordinate the objectives of each layer, and dynamically adjust the optimization weight coefficients of each level through the information decision - making module to achieve the collaborative optimization of millisecond - level voltage correction and hour - level economic dispatch;

[0009] Step 4: Define the dynamic constraint boundaries of the control variables at each level (voltage amplitude at the device layer, network loss threshold at the system layer). Decompose the multi - layer robust optimization model into a local linear sub - problem at the device layer and a global convex optimization main - problem at the system layer, and express the cross - level power balance constraint through a coupling matrix for easy solution.

[0010] The beneficial effects of the present invention adopting the above technical solutions are as follows:

[0011] 1. Considering the measurement accuracy, time scale of multi - source data, and the synchronization of state prediction calculations, respectively pre - process the millisecond - level real - time measurement data, minute - level prediction data, and hour - level scheduling instructions in a hierarchical manner, and implement classification and fusion according to the power source type (controllable / uncontrollable) and data timeliness (such as Kalman filter noise reduction, LSTM interpolation, moving average filter), significantly improving the data quality and usability, and solving the problem of control instruction mismatch caused by data discontinuity or noise interference in traditional methods.

[0012] 2. Construct a multi - objective robust optimization model, enhance the dynamic coupling relationship between the control variables at the device layer (such as DG output, charge and discharge of energy storage) and the state variables at the system layer (such as node voltage, branch power flow) through a single - hidden - layer non - linear structure, and accurately depict the photovoltaic output fluctuation and load mutation scenarios in combination with the time - varying uncertainty set, greatly improving the accuracy and practicality of the prediction model, and realizing the collaborative optimization of cross - level objectives.

[0013] 3. Define the dynamic upper and lower limits of state variables such as the voltage amplitude at the device layer and the system-level network loss threshold, decompose the non-convex hierarchical optimization model into a device-layer linear sub-problem and a system-layer convex optimization main problem, and linearize the branch power flow constraint based on the improved second-order cone relaxation technique, while ensuring the model accuracy, the solution efficiency is increased by more than 30%, meeting the real-time requirements of millisecond-level voltage correction and hour-level economic dispatch.

[0014] 4. By constructing a time-varying uncertainty set to describe the fluctuation range of renewable energy output and the load mutation scenario, expand the traditional deterministic optimization model into a mixed-integer robust optimization model, and define the branch current relaxation boundary in combination with the rated operating point. On the premise of ensuring the feasibility of the power flow constraint, reduce the complexity of the non-convex model. The robustness of the obtained scheduling scheme is increased by 15%, the network loss is reduced by 8.767%, and the voltage deviation is reduced by 19.37%, significantly enhancing the system's adaptability to high-penetration new energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The above and / or additional aspects and advantages of the present invention will become obvious and easy to understand from the following description of the embodiments in conjunction with the drawings, where:

[0016] Figure 1 It is a schematic diagram of the work flow of a distribution network optimization method based on hierarchical robust control and dynamic information decision-making.

[0017] Figure 2 It is the voltage distribution of the entire network at 12:00 after the distribution network is optimized.

[0018] Figure 3 It is a comparison of the network losses between the hierarchical robust control and the second-order cone programming method.

[0019] Figure 4 It is a comparison of the SVC output operation curves between the hierarchical robust control and the SOCP method. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the appended Figures 1-4 in the embodiments of the present invention.

[0021] The above-mentioned distribution network optimization method based on hierarchical robust control and dynamic information decision-making includes the following processes:

[0022] Figure 1 It is a schematic diagram of the work flow of a distribution network optimization based on hierarchical robust control and dynamic information decision-making.

[0023] The specific method of hierarchical preprocessing and classification fusion described in Step 1 is as follows: When using millisecond-level measured voltage and current data at the device layer, dynamic noise filtering based on the Kalman filter algorithm and data interpolation based on LSTM time series prediction are required to address the data tomography problem caused by sudden irradiance changes or wind speed fluctuations in inertia-less units; when processing minute-level prediction data at the regional layer, the sampling frequency is allowed to be appropriately reduced, such as slightly greater than 30 seconds per time, and moving average filtering is used to smooth the power command fluctuations using its inertia characteristics; the real-time data at the device layer is sampled at 1 kHz, the LSTM rolling update prediction data at the regional layer, and the hourly scheduling data at the system layer are aligned by timestamp and marked with priorities (device layer > regional layer > system layer), and are dynamically weighted and fused in the information decision module according to the control level weights.

[0024] For the data at the device layer, the specific implementation details of preprocessing and fusion are further refined, such as Kalman filter noise reduction, data sampling period requirements, and data interpolation formulas.

[0025] The specific implementation details of refined preprocessing and fusion in Step 1, such as Kalman filter noise reduction, data sampling period requirements, and data interpolation formulas based on the LSTM method, etc., together with Claim 2 constitute a complete hierarchical preprocessing and classification fusion method.

[0026] 1) The millisecond-level real-time data (voltage, current) is denoised by the Kalman filter, and the filter parameters are adaptively adjusted according to the device noise characteristics. The update period of the noise covariance matrix is 1 second. The slope factor k1 in the priority weight function is dynamically adjusted according to the historical voltage over-limit frequency: when the number of over-limit times within 24 hours exceeds 5 times, k1 increases from 0.5 to 0.8 to improve the response speed. Hierarchical preprocessing priority weight function:

[0027]

[0028] where w dev (t) is the real-time weight coefficient of the device layer data, k1 is the weight adjustment slope, DV(t) is the real-time voltage deviation, and V th is the voltage deviation threshold.

[0029] 2) For controllable power sources, such as energy storage, the minute-level data sampling period should satisfy:

[0030]

[0031] where T ctrl is the data sampling period of the controllable power source, and t DG is the inertia time constant of the distributed power source.

[0032] 3) For uncontrollable power sources such as photovoltaics, LSTM is used for millisecond-level data interpolation. When using LSTM for millisecond-level data interpolation for uncontrollable power sources such as photovoltaics, the LSTM model utilizes its powerful sequence processing ability to learn the historical timing characteristics of the photovoltaic power output. When data loss is detected, LSTM will predict the photovoltaic output value at the missing time point according to the timing pattern learned previously, thus achieving high-precision millisecond-level data interpolation. This method can effectively handle the intermittency and volatility of photovoltaic output caused by factors such as weather changes, and improve the integrity and reliability of data.

[0033] LSTM controls the flow of information through forget gates, input gates, and output gates, and uses cell states to store and transfer long-term information, thus achieving effective processing and prediction of sequence data.

[0034] Forget gate:

[0035] f t = σ(W f × [h t-1 , x t + b f ) (3)

[0036] In the formula, f t represents the output of the forget gate; σ represents the sigmoid activation function, and the output value is between 0 and 1; W f is the weight matrix of the forget gate; h t-1 is the hidden state at the previous moment; x t is the input at the current moment; b f is the bias term of the forget gate.

[0037] Input gate:

[0038] i t = σ(W i × [h t-1 , x t + b i ) (4)

[0039] In the formula, i t is the output of the input gate; W i is the weight matrix of the input gate; b i is the bias term of the input gate.

[0040] Output gate:

[0041] o t = σ(W o × [h t-1 , x t + b o ) (5)

[0042] In the formula, ot is the output of the output gate; W o is the weight matrix of the output gate; b o is the bias term of the output gate.

[0043] The millisecond-level data interpolation formula is:

[0044]

[0045] In the formula, P PV is the photovoltaic output; k is a constant; is the interpolated photovoltaic output, M is the number of measured points in the interpolation window, T is the sampling period, and δ is the time decay coefficient, which is defaulted to 0.2 s.

[0046] The device layer RMPC objective function of the hierarchical optimization model described in step 2 is:

[0047]

[0048] In the formula, u k is the control input or prediction error at the k-th moment, which is used to adjust the system state to reach the desired voltage value; Q is a matrix, which is used to define the weight of the error; N p is the prediction horizon length, V k is the predicted node voltage value at the k-th moment, V ref is the voltage reference value.

[0049] The prediction horizon length N p needs to be set according to the device response speed (≤100 ms), and usually takes 5 - 10 steps.

[0050] The coupling constraint of the system layer ADMM global optimization model is:

[0051]

[0052] In the formula, A sys , B dev are the constraint matrices of the system layer and the device layer; x sys , x dev are the decision variables of the system layer and the device layer; b sys is the constant term on the right side of the constraint condition, which is used to define the specific value of the constraint.

[0053] Distributed solution is achieved through the augmented Lagrangian function:

[0054]

[0055] In the formula, ρ is the penalty factor, and λ is the Lagrange multiplier vector.

[0056] The penalty factor ρ needs to be dynamically adjusted according to the convergence speed and accuracy, and the initial value is recommended to be taken as 1-10.

[0057] The branch power flow constraint of the improved second-order cone relaxation model described in step 3 is:

[0058]

[0059] In the formula, P ij represents the active power flowing from node i to node j on the branch; Q ij represents the reactive power flowing from node i to node j on the branch; I i and I j are the currents of node i and node j respectively.

[0060] And introduce a slack variable ε ij to handle non-convex constraints:

[0061]

[0062] In the formula, ε ij is the slack variable.

[0063] The slack penalty term is added to the objective function:

[0064]

[0065] In the formula, ω ij is the slack penalty weight.

[0066] It should be noted that the slack penalty term needs to set a weight in the objective function, and its value is determined through sensitivity analysis.

[0067] The dynamic weight adjustment rule is:

[0068]

[0069] It is necessary to monitor the voltage deviation in real time, dynamically update the weight function, and ensure the balance between fast response and stability.

[0070] Figure 2 is the node voltage distribution diagram at the selected typical moment. It can be directly seen from the figure that the voltage deviation of the SOCP method is relatively large, and there is a risk of over-limit; the average voltage deviations of the SOCP method and the method proposed in this paper are 0.4998 and 0.4403 respectively. Compared with the SOCP method, the average voltage deviation of the method proposed in this paper is reduced by about 11.90%. Therefore, the method proposed in this paper can minimize the voltage deviation on the premise of ensuring the safe operation of the distribution network.

[0071] The method for constructing the sensitivity matrix of the device layer linearization sub-problem in step 4 is:

[0072]

[0073] Wherein, V is voltage; P is active power; J is the active power-voltage sensitivity matrix, and Y bus is the nodal admittance matrix; represents a diagonal matrix with the complex conjugate voltage as the diagonal elements; represents taking the real part of the complex matrix, because in an actual power system, voltage and power are usually real numbers.

[0074] The matrix expression of the cross-layer coupling constraint is:

[0075]

[0076] Wherein, I is the unit matrix of the device layer, and J cp1 is the cross-layer power interaction sensitivity matrix, and b cp1 , b sys are the coupling constraint and system layer constraint vectors.

[0077] The solution architecture of the hierarchical optimization model is as follows: Device layer RMPC problem: Use the CasADi + IPOPT solver, with a calculation period ≤ 100 ms; System layer ADMM problem: Call the CPLEX solver, with a calculation period ≤ 5 minutes; Cross-layer coupling: Achieve real-time data interaction through the OPC UA protocol, with a time delay ≤ 10 ms.

[0078] It should be noted that during actual deployment, the nodal admittance matrix and the slack penalty weight need to be adjusted according to the distribution network topology to ensure the adaptability of the model.

[0079] Figure 3 The network loss curves for different optimization methods are shown. It can be seen from the figure that the optimization effect is particularly obvious from 14:00 to 22:00. Among them, the network loss at 20:00 is reduced by 21.69 kW; the network losses of the SOCP method and the method of the present invention are 569.13 kW·h and 472.51 kW·h respectively. Compared with the SOCP method, the network loss of the method of the present invention is reduced by approximately 16.98%. Therefore, the method of the present invention performs better in reducing network loss.

[0080] Figure 4 The SVC output curves for different methods are shown. It can be analyzed that the SVC output of the method of the present invention is generally higher than that of the SOCP method, and it is more obvious during the period when the distributed power penetration rate is relatively high. At this time, the average SVC output reaches 0.89, which is increased by approximately 16.81% compared with the SOCP method; it plays a better reactive power compensation role.

[0081] Table 1 shows the comparison of various indicators between the traditional distribution network optimization SOCP method (Method 1) and the distribution network optimization method considering hierarchical robust control and dynamic information decision-making of the present invention (Method 2). The method of the present invention is comparable to the linear optimal power flow in terms of calculation speed and has a high solution efficiency; there is no significant difference in the new energy utilization rate compared with the traditional deterministic optimization method.

[0082] The power loss of the system is reduced by 8.767%; the average voltage deviation is reduced by 19.37%.

[0083] Table 1 Comparison of various indicators of the two methods

[0084]

[0085] The results show that the method proposed in the present invention can accurately predict the output interval of distributed power sources and the system voltage change interval, and obtain a good distribution network optimization effect.

Claims

1. A distribution network optimization method based on hierarchical robust control and dynamic information decision-making, characterized in that, The steps include the following: Step 1: Considering the multi-time-scale control requirements, data acquisition frequencies, and differences in optimization objectives, divide the distribution network into the device layer, regional layer, and system layer. Perform hierarchical preprocessing on millisecond-level measured voltage and current data, minute-level predicted PV output data, and hourly scheduling instructions respectively, and implement classification and fusion of data based on power source types (controllable / uncontrollable) and control levels. Step 2: Construct a cross-level training dataset through the real-time inverter status of the device layer, the operating data of the energy storage SOC, the predicted values of distributed power generation output in the regional layer combined with the long short-term memory (LSTM) prediction results of irradiance and wind speed, and the statistical data of load demand in the system layer. Establish a hierarchical multi-objective robust optimization model, where the robust model predictive control (RMPC) is used in the device layer and the alternating direction method of multipliers (ADMM) is used in the system layer. Step 3: Improve the hierarchical optimization model using the second-order cone relaxation technique, define the time-varying boundaries of the control variables in the device layer and the state variables in the system layer, and embed the system model to coordinate the objectives of each layer. Dynamically adjust the optimization weight coefficients of each level through the information decision-making module to achieve the collaborative optimization of millisecond-level voltage correction and hourly economic scheduling. Step 4: Define the dynamic constraint boundaries of the control variables at each level, decompose the multi-level robust optimization model into local linear sub-problems in the device layer and a global convex optimization main problem in the system layer, and express the cross-level power balance constraint through a coupling matrix for easy solution.

2. A distribution network optimization method based on hierarchical robust control and dynamic information decision-making according to claim 1, characterized in that: The methods of hierarchical preprocessing and classification fusion in Step 1 are as follows: When using millisecond-level measured voltage and current data in the device layer, dynamic noise filtering based on the Kalman filter algorithm and data interpolation based on LSTM time series prediction are required to address the data fault problems caused by sudden changes in irradiance or wind speed fluctuations of inertia-less units. When processing minute-level predicted data in the regional layer, the sampling frequency is allowed to be appropriately reduced (≥30 seconds / time) and moving average filtering is used to smooth the power command fluctuations using its inertia characteristics. Align the real-time data (1kHz sampling) in the device layer, the predicted data (LSTM rolling update) in the regional layer, and the scheduling data (hourly plan) in the system layer through timestamps and priority markings (device layer > regional layer > system layer), and perform dynamic weighted fusion according to the control level weights in the information decision-making module.

3. The optimization method for a distribution network based on hierarchical robust control and dynamic information decision-making according to claim 2, characterized in that: The methods of hierarchical preprocessing and classification fusion in Step 1 specifically include: 1) Millisecond-level real-time data (voltage, current) is denoised by the Kalman filter, and the filter parameters are adaptively adjusted according to the device noise characteristics. The update period of the noise covariance matrix is 1 second. The slope factor k1 in the priority weight function is dynamically adjusted according to the historical voltage overlimit frequency: When the number of overlimit times within 24 hours exceeds 5 times, k1 increases from 0.5 to 0.8 to improve the response speed. The priority weight function of hierarchical preprocessing: where w dev (t) is the real-time weight coefficient of the device layer data, k1 is the weight adjustment slope, DV(t) is the real-time voltage deviation, and V th is the voltage deviation threshold; 2) For controllable power sources such as energy storage, the minute-level data sampling period should satisfy: Where, T ctrl is the sampling period of the controllable power supply data, and t DG is the inertia time constant of the distributed power supply; 3) For uncontrollable power sources such as photovoltaic, LSTM is used for millisecond-level data interpolation: When using LSTM for millisecond-level data interpolation for uncontrollable power sources such as photovoltaic, when data loss is detected, LSTM will predict the photovoltaic output value at the missing time point according to the previously learned time series pattern, so as to achieve high-precision millisecond-level data interpolation; LSTM controls the flow of information through forget gates, input gates, and output gates, and uses cell states to store and transmit long-term information, so as to achieve effective processing and prediction of sequence data: Forget gate: f t = σ(W f × [h t-1 , x t + b f ) (3) where, f t represents the output of the forget gate; σ represents the sigmoid activation function, and the output value ranges from 0 to 1; W f is the weight matrix of the forget gate; h t-1 is the hidden state at the previous moment; x t is the input at the current moment; b f is the bias term of the forget gate; Input gate: i t = σ(W i × [h t-1 , x t + b i ) (4) where, i t is the output of the input gate; W i is the weight matrix of the input gate; b i is the bias term of the input gate; Output gate: o t = σ(W o × [h t-1 , x t + b o ) (5) where, o t is the output of the output gate; W o is the weight matrix of the output gate; b o is the bias term of the output gate; The millisecond-level data interpolation formula is: where P PV is the PV output; k is a constant; is the PV output after interpolation, M is the number of measured points in the interpolation window, T is the sampling period, and σ is the time decay coefficient, which is defaulted to 0.2 s.

4. A distribution network optimization method based on hierarchical robust control and dynamic information decision-making according to claim 1, characterized in that: The objective function of the robust model predictive control RMPC described in step 2 is: where \(u\) k is the control input or prediction error at the \(k\)-th moment, which is used to adjust the system state to achieve the desired voltage value; Q is a matrix used to define the weight of the error; N p is the prediction horizon length, V k is the predicted value of the node voltage at time k, V ref is the voltage reference value.

5. A distribution network optimization method based on hierarchical robust control and dynamic information decision-making according to claim 1, characterized in that: The coupling constraint of the global optimization model of the distributed alternating direction multiplier method in step 2 is: where b sys is the constant term on the right side of the constraint condition, used to define the specific value of the constraint, A sys , B dev are the constraint matrices of the system layer and the device layer, x sys , x dev are the decision variables of the system layer and the device layer; Distributed solution is achieved through the augmented Lagrangian function: In the formula, ρ is the penalty factor and λ is the Lagrangian multiplier vector.

6. A distribution network optimization method based on hierarchical robust control and dynamic information decision-making according to claim 1, characterized in that: The branch power flow constraint for improving the hierarchical optimization model by the second-order cone relaxation technique in step 3 is: Wherein, P ij represents the active power flowing from node i to node j on the branch; Q ij represents the reactive power flowing from node i to node j on the branch; I i and I j are the currents of node i and node j respectively; And introduce a slack variable ε ij Handle non-convex constraints: where ε ij is the slack variable; The relaxation penalty term is added to the objective function: where ω ij is the relaxation penalty weight.

7. A distribution network optimization method based on hierarchical robust control and dynamic information decision-making according to claim 1, characterized in that: The dynamic weight adjustment rule of the information decision module described in step 3 is:

8. A distribution network optimization method based on hierarchical robust control and dynamic information decision-making according to claim 1, characterized in that: Construction of the sensitivity matrix for the local linear sub-problem at the device layer in step 4: Where, V is the voltage; P is the active power; J is the active-power voltage sensitivity matrix, and Y bus is the nodal admittance matrix; denotes the diagonal matrix with the complex conjugate voltage as the diagonal elements; denotes taking the real part of the complex matrix, because in an actual power system, voltage and power are usually real numbers.

9. The optimization method for a distribution network based on hierarchical robust control and dynamic information decision-making according to claim 1, wherein: The matrix expression of the cross-level power balance constraint in step 4 is: where I is the unit matrix of the device layer, and J cp1 is the cross-layer power interaction sensitivity matrix, and b cp1 , b sys is the coupling constraint and system layer constraint vector.

10. A distribution network optimization method based on hierarchical robust control and dynamic information decision-making according to claim 1, characterized in that: The solution architecture of the hierarchical optimization model is: Robust model predictive control RMPC method: Using the CasADi + IPOPT solver, the calculation period ≤ 100 ms; Distributed alternating direction multiplier method ADMM method: Calling the CPLEX solver, the calculation period ≤ 5 minutes; Cross-layer coupling: Real-time data interaction is achieved through the OPC UA protocol, and the time delay ≤ 10 ms.

Images