A load intelligent regulation method and system of an electric energy metering box

Abstract

The application discloses a kind of load intelligent regulation and control method and system of electric energy metering box, comprising: the electrical quantity of each outgoing line loop in metering box and each node temperature data are collected;Using lumped parameter method, heat source and heat dissipation path in box are equivalent to multiple-node thermal network, and the electric-thermal coupling dynamic model in the form of state space is established;In each control cycle, the forward integral of model is obtained to obtain temperature prediction trajectory, with each node temperature not exceeding safety threshold as hard constraint, and optimal regulation sequence is solved using rolling horizon optimization and current time control action is executed;When power is restored, the heat coupling propagation between loops is simulated based on the non-diagonal element module of heat conduction matrix, the global temperature trajectory of the pseudo-power restoration sequence is preformed, and the power restoration sequence and waiting interval are iteratively adjusted;Actual temperature response is compared with the deviation of predicted trajectory, and model parameter online updating is triggered to form a closed loop.The application realizes the predictive, self-adaptive intelligent regulation and control of metering box load, and guarantees power utilization safety and power supply continuity.

Description

Technical Field

[0001] This application relates to the field of power distribution automation and intelligent safety control technology, and in particular to a method and system for intelligent load regulation of an energy metering box. Background Technology

[0002] Electricity metering boxes are key devices in low-voltage distribution networks, connecting power supply companies and end users. They are typically installed at the entrance of residential buildings or in distribution rooms, and contain a centralized main circuit breaker, miniature circuit breakers for each user's outgoing circuit, electricity meters, terminal blocks, and connecting busbars. With the continuous growth of residential electricity load, the number of outgoing circuits in a single electricity metering box has reached 6 to 24. The heat loss from copper in each circuit accumulates within the enclosed box, making the temperature rise problem of terminal blocks and circuit breaker contacts increasingly prominent.

[0003] Currently, the safety protection of electricity metering boxes mainly relies on the overcurrent tripping mechanism of the circuit breaker itself. When the current in a certain circuit exceeds the rated value, the circuit breaker of that circuit trips and cuts off the power supply. This protection method has the following shortcomings: First, it is a reactive response after the fact, and cannot intervene in advance before the temperature reaches a dangerous level; second, once it trips, all users in that circuit lose power, lacking the ability to classify and control different circuits under thermal safety constraints; third, when multiple circuits request power restoration simultaneously after a trip, the surge current of closing the circuit breaker and the residual heat inside the box are superimposed, which can easily trigger a secondary trip, forming a vicious cycle of repeated tripping.

[0004] In existing technologies, temperature sensors and controllers are installed inside the enclosure to cut off power to some circuits when the detected temperature exceeds a set threshold. This method, which uses single-point temperature detection and threshold comparison, fails to model the thermal coupling relationships between multiple circuits within the enclosure and cannot predict future temperature rise trends. Furthermore, by acquiring the power of each branch circuit and the global temperature, the target branch circuit is cut off when the temperature exceeds a warning threshold, and the remaining heat capacity margin is calculated based on the inherent heat capacity constant to guide the power restoration sequence. Although the concept of heat capacity margin is introduced, the temperature model is a simplified single-point model, lacking multi-node thermal network equations, and the control strategy is reactive, lacking feedforward predictive capabilities.

[0005] In summary, existing technologies lack a method for intelligent load control of power metering boxes that can be based on multi-node thermal coupling modeling, implement feedforward predictive control before the temperature reaches a safe threshold, and achieve model self-adaptation through closed-loop parameter identification. Summary of the Invention

[0006] In view of the technical problems in the existing technology of load protection of power metering boxes, which mainly relies on passive tripping after the fact, lacks feedforward predictive control methods, and fails to model and utilize the multi-circuit thermal coupling relationship, this application provides a method and system for intelligent load control of power metering boxes.

[0007] To achieve the above objectives, the present application provides a method for intelligent load control of an electricity metering box, which adopts the following technical solution: S1: Collect electrical quantity data of each outgoing circuit in the power metering box and temperature data of each node; S2: The lumped parameter method is used to convert the heat source and heat dissipation path in the power metering box into a multi-node thermal network. The real-time power of each loop is used as the heat source excitation input and the temperature of each node is used as the state variable to establish a state-space form of electrothermal coupling dynamic model. Among them, the off-diagonal elements of the thermal conductivity matrix represent the thermal coupling propagation relationship between different loops through the box wall, air layer or common mounting plate. S3: In each control cycle, with the current temperature of each node and the power of the loop as the initial conditions, the electrothermal coupling dynamic model is integrated forward in the prediction time domain to obtain the predicted temperature trajectory of each node; with the temperature of each node in the prediction time domain not exceeding the safety threshold as a hard constraint, the optimal control sequence is solved by rolling time domain optimization. S4: Based on the temperature prediction trajectory, before the temperature prediction at any node reaches the safety threshold, generate load control instructions for one or more loops, and perform on / off or power regulation on the corresponding loops. S5: Compare the actual temperature response after execution with the predicted temperature trajectory. When the deviation exceeds the preset allowable range, trigger the online update of the parameters of the electrothermal coupling dynamic model to form a closed-loop control.

[0008] Preferably, the step of establishing a state-space electrothermal coupling dynamic model using the real-time electrical power of each loop as the heat source excitation input and the temperature of each node as the state variable includes: Using the real-time electrical power of each loop as the heat source excitation input of the thermal network and the temperature of each node as the state variable, the state-space equation of the following form is established: in, Here is the node heat capacity matrix. This is a thermal conductivity matrix, where the diagonal elements represent the equivalent heat conduction of each node to the environment, and the off-diagonal elements represent the thermal conductivity. Representation Nodes With nodes The thermal coupling strength between them via the enclosure wall, air layer or common mounting plate; For the node temperature vector, For each circuit's power vector, For ambient temperature, This is the power-heat source mapping matrix. This is the environment coupling vector; Each component of the power vector is obtained by multiplying the square of the effective value of the current collected in real time in each circuit with the equivalent resistance of the corresponding conductor and terminal, and characterizes the copper loss heating power of each circuit.

[0009] Preferably, the rolling time-domain optimization solution for the optimal control sequence includes: In each control cycle, the electrothermal coupling dynamic model is solved within a preset prediction time domain, using the current temperature and loop power status of each node as initial conditions, to obtain the temperature prediction trajectory of each node. Construct the cost function: The first term is the weighted quadratic norm of the deviation between the temperature of each node and the safety threshold in the predicted time domain, and the second term is the number of control actions in the time domain. The third penalty is a weighted penalty for the impact on the user's electricity consumption; , These are the weighting coefficients; , This is a diagonal matrix representing the priority weights of each loop; This is the security threshold vector for each node.

[0010] Under the hard constraint that the temperature of each node does not exceed the safety threshold, the optimal control sequence of the cost function is solved, and the first set of control actions at the current moment is executed; in the next control cycle, the initial state is updated using the latest collected data, and the above rolling time-domain optimization process is repeated.

[0011] Preferably, the weight configuration method for each optimization objective in the cost function includes: The primary optimization objective is to minimize the total duration during which each circuit is disconnected or deloaded. The second optimization objective is to maximize the power supply continuity of high-priority user circuits. The third optimization objective is to minimize the impact on user electricity consumption. The first optimization objective is approximately optimized through the second term of the cost function, which is the penalty for the number of control actions. The second optimization objective corresponds to the priority weight configuration in the cost function, and the third optimization objective corresponds to the third term of the cost function, which is the penalty for the impact of user electricity consumption.

[0012] Preferably, the method further includes a peak-shaving power restoration strategy: Acquire the surge injection heat energy of each circuit to be restored; Based on the state-space equation of the electrothermal coupling dynamic model, a global multi-node temperature trajectory prediction is performed on the pseudo-re-energization sequence containing all circuits to be re-energized: through the forward integration of the state-space equation, the transient thermal process after the re-energization of each circuit to be re-energized is simulated in sequence. The temperature rise caused by the re-energization of each circuit is propagated to other nodes through the off-diagonal elements of the thermal conductivity matrix in the electrothermal coupling dynamic model, so that the remaining heat capacity margin of the node corresponding to the circuit to be re-energized later in the pseudo-re-energization sequence is affected by the cumulative thermal contribution of the circuits already re-energized earlier in the sequence. If the pre-simulation results show that the temperature of any node exceeds the safety threshold during the execution of the proposed power restoration sequence, then at least one of the following adjustments will be performed iteratively: adjust the power restoration sequence, insert a waiting interval between adjacent power restorations, or reduce the power restoration power of the circuit to be restored, until the pre-simulation is passed and the power restoration operation is performed according to the determined power restoration sequence.

[0013] Preferably, the peak-shaving power restoration strategy further includes: Based on the dynamic changes in the surge injection heat energy of each circuit to be restored and the remaining heat capacity margin of each node, a power restoration sorting queue is generated in real time. The surge injection heat energy is calculated by multiplying the square of the rated current of the circuit, the square of the inrush current multiple, the equivalent resistance of the circuit, and the closing duration. The effective value of the closing surge current is the product of the rated current and the inrush current multiple, or obtained through online learning of historical switching data. During the power restoration sequencing process, if the remaining heat capacity margin of any node is reduced to below the surge injection heat energy of the next circuit to be restored due to the input of new load or changes in ambient temperature, the queue execution is paused, the global temperature trajectory is re-performed based on the electrothermal coupling dynamic model, and the power restoration sequence is readjusted or the waiting interval is increased according to the re-performance results.

[0014] Preferably, the online parameter update includes: After executing a load control command that causes one or more circuits to switch on or off, collect transient temperature response data and corresponding power change data caused by the switching operation. The state-space equation of the electrothermal coupling dynamic model is discretized into a difference form. A linear regression equation is constructed with the temperature change of each node as the output and the power change of each loop and the ambient temperature change as the input. The parameters in the thermal capacity matrix, thermal conductivity matrix and environmental coupling vector of the electrothermal coupling dynamic model are updated online using the recursive least squares method with forgetting factor. The forgetting factor is used to adjust the relative weight of historical measurement data and current measurement data in the parameter recursive estimation, so that the model parameters are updated adaptively with the change of load conditions. The confidence criterion for setting parameter updates is as follows: when the data length of the transient process is less than the preset minimum number of samples or the signal-to-noise ratio is lower than the preset threshold, the corresponding parameters are not updated temporarily, and the parameter values ​​of the previous period are maintained.

[0015] Preferably, the method further includes a model self-evolution mechanism: The actual temperature response after regulation is compared with the predicted temperature trajectory on a node-by-node basis. When the prediction deviation of any node exceeds the preset allowable range for a certain number of consecutive times, a preset threshold is reached, and forced re-identification of the corresponding node parameters in the thermal conductivity matrix and thermal capacity matrix is ​​triggered. When the length of the consecutive same number segment of the deviation exceeds the preset proportion of the total number of samples and the absolute value of the first-order autocorrelation coefficient exceeds the preset threshold, it is determined that the number of nodes in the current thermal network model is insufficient to characterize the actual heat distribution inside the box. In the thermal network model, auxiliary heat dissipation nodes are added to improve the model order. The thermal network model is the node network structure composed of thermal resistance and thermal capacity in the electrothermal coupling dynamic model. When the number of circuits in the power metering box changes due to increases or decreases, the dimension of the state space equation is automatically reconstructed, and the corresponding heat source nodes and row and column elements of the thermal conductivity matrix are added or deleted.

[0016] Preferably, the rolling time-domain optimization further includes an adaptive adjustment method for the prediction time domain and the control time domain, including: The average deviation between the actual temperature response and the predicted temperature trajectory within the most recent preset number of control cycles is statistically analyzed. When the average deviation increases, the prediction time domain length is shortened to reduce the uncertainty of long-term prediction, while the control time domain length is shortened to increase the control frequency. When the average deviation decreases, the prediction time domain length is extended to increase the time margin of the feedforward prediction. The adjustment range of the prediction time domain length is limited to ensure that the temperature prediction uncertainty of the farthest prediction node does not exceed a preset confidence level; the preset confidence level is that the maximum value of the standard deviation of temperature prediction for each node in the prediction time domain does not exceed a preset threshold, and the preset threshold is determined based on the difference between the safety threshold and the control trigger margin.

[0017] This application also provides a load intelligent control system for an electricity metering box, including: a sensing module, used to collect electrical quantity data of each outgoing circuit and temperature data of each key node in the electricity metering box; The modeling and prediction module is used to build an electrothermal coupling dynamic model of a multi-node thermal network with lumped parameters based on the collected data. The off-diagonal elements of the thermal conductivity matrix represent the thermal coupling propagation relationship between different loops. The temperature prediction trajectory of each node is obtained by forward integration in each control cycle, and the optimal control sequence is solved by rolling time-domain optimization. The decision module is used to generate load control instructions before the temperature prediction trajectory of any node reaches the safety threshold, and to perform a global temperature trajectory pre-simulation of the proposed power restoration sequence based on the electrothermal coupling dynamic model during power restoration to determine the power restoration sequence and waiting interval. The execution module is used to perform on / off switching or power regulation on the corresponding outgoing circuit according to the load control command; The parameter online update module is used to compare the deviation between the actual temperature response and the temperature prediction trajectory. When the deviation exceeds the preset allowable range, the parameters of the electrothermal coupling dynamic model are updated online and fed back to the modeling and prediction module to form a closed-loop control.

[0018] Compared with the prior art, this application provides a method and system for intelligent load control of an electricity metering box, which has the following advantages: 1. Implement feedforward active control and multi-loop thermal coordination management based on an electrothermal coupling model. By establishing an electrothermal coupling dynamic model containing off-diagonal elements of the thermal conductivity matrix, the thermal coupling relationship between loops through the enclosure wall and common substrate is explicitly characterized. Rolling time-domain optimization is used to generate load reduction or circuit breaking commands in advance before the temperature exceeds the limit, turning passive tripping into active prevention and avoiding insulation aging and arcing faults. Combined with a multi-priority weighted objective function, refined load management of "ensuring critical loads and cutting secondary loads" is achieved under thermal safety constraints, ensuring the continuity of power supply for critical loads such as fire protection and elevators.

[0019] 2. A staggered power restoration strategy based on matching residual heat capacity margin is proposed to suppress cascading trips and secondary overheating. Based on an electrothermal coupling model, the global temperature trajectory is pre-simulated, and the dynamic residual heat capacity margin of each node is calculated and quantitatively matched with the surge injection heat energy of the circuit to be restored. When the residual heat capacity margin is greater than or equal to the surge heat energy, power is restored sequentially according to the heat energy from smallest to largest. This eliminates the thermal shock superposition of multiple circuits simultaneously restored from a mechanistic perspective, avoiding excessive secondary temperature rise and cascading trips, and improving power restoration safety.

[0020] 3. Construct a parameter-adaptive local edge autonomous architecture to ensure long-term model accuracy and millisecond-level response. A recursive least squares method with a forgetting factor is used to update thermal network parameters online, automatically correcting the model to adapt to slow-changing disturbances such as sealing strip aging and environmental changes, eliminating the need for manual calibration. All calculations are performed locally on the in-box edge processor, independent of communication links, and significantly smaller than the metering box's thermal inertia time constant, matching the thermal inertia time constant to ensure independent autonomy and rapid response capabilities under extreme conditions such as communication interruptions. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the steps of an intelligent load control method for an electricity metering box according to an embodiment of this application.

[0022] Figure 2This is a flowchart of the peak-shaving and power restoration strategy of a load intelligent control method for an electricity metering box according to an embodiment of this application.

[0023] Figure 3 This is a flowchart of an online parameter update method for a load intelligent control method for an energy metering box according to an embodiment of this application. Detailed Implementation

[0024] The following is in conjunction with the appendix Figure 1-3 The preferred embodiments of this application are described herein. It should be understood that the preferred embodiments described herein are for illustration and explanation of this application and are not intended to limit this application.

[0025] This application describes a low-voltage electricity metering box in a residential community as an example. The electricity metering box contains eight single-phase circuits with rated currents of: Circuit 1 for public lighting (16A), Circuit 2 for elevators (32A), and Circuits 3 through 8 for residential lines (40A). The box is made of 1.2mm cold-rolled steel plate, with dimensions of 600mm x 800mm x 200mm (width x height x depth). It is installed in an underground electrical distribution room where the ambient temperature is maintained between 20℃ and 38℃ year-round. It includes: S1. Collect electrical quantity data of each outgoing circuit in the power metering box and temperature data of each node; The electrical quantity data of each load circuit in the power metering box and the temperature data of each measuring point in the box are collected synchronously at a preset sampling period.

[0026] The electrical quantity data includes the RMS current, RMS voltage, and power factor of each circuit. Current acquisition uses a feedthrough Hall effect current sensor with a range of 0–100A, a measurement accuracy of at least ±0.5%FS, and a bandwidth of at least DC–5kHz. Voltage acquisition is achieved through a resistor divider network, and the voltage is conditioned by an isolated operational amplifier before being sent to the analog-to-digital converter.

[0027] The temperature data includes the temperature values ​​of each thermal node inside the enclosure. Temperature acquisition uses NTC thermistors with a nominal resistance of 10kΩ (25℃), a B value of 3950K, a measurement accuracy of ±1%, and a thermal response time constant of no more than 5s. Temperature sensors are located near the terminals of each circuit, at the center of the enclosure wall, and on the top of the enclosure, with no fewer than four measuring points.

[0028] The sampling period ranges from 100ms to 1s, and is determined based on the required control precision and the processor's computing power.

[0029] Because the thermal response time constant of an NTC thermistor is greater than the sampling period, the temperature measurement exhibits a first-order inertial lag. To compensate for the impact of this lag on prediction accuracy, a first-order inverse filter discretization formula is used for compensation before inputting the measured temperature into the model. The discretization formula for the first-order inverse filter is as follows: in, The compensated temperature value, Let be the measured temperature value in the k-th sampling period, and τ be the thermal response time constant of the NTC thermistor. Sampling period For example: Install one through-hole Hall current sensor at each of the outgoing terminals of loops 1 to 8. Place NTC thermistor temperature sensors at the following locations: Measurement point T1: Center position of the terminal block of circuits 3 to 5; Measurement point T2: Center position of the terminal block of circuits 6 to 8; Measuring point T3: Center of the rear wall of the enclosure; Measuring point T4: Inside the top of the enclosure.

[0030] An ambient temperature sensor is also installed on the outside of the enclosure.

[0031] All sensor signals are processed by a signal conditioning circuit and then acquired by an ARM Cortex-M4 microcontroller (168MHz clock speed, 12-bit analog-to-digital converter, maximum sampling rate of 2.4MSPS), with a sampling period of 100ms.

[0032] S2. The heat source and heat dissipation path in the power metering box are equivalent to a multi-node thermal network using the lumped parameter method. The real-time power of each circuit is used as the excitation input of the heat source and the temperature of each node is used as the state variable to establish a state-space form of electrothermal coupling dynamic model. Among them, the off-diagonal elements of the thermal conductivity matrix represent the thermal coupling propagation relationship between different circuits through the box wall, air layer or common mounting plate. The electricity metering box is regarded as a lumped parameter thermal network, which is equivalent to a thermal network model with M thermal nodes. The state-space equation between the internal temperature of the box and the power of each loop is established.

[0033] Specifically, taking the copper losses of each circuit conductor as heat source nodes and the box wall as heat dissipation nodes, the following state-space equations are established based on the law of conservation of energy: in: for The temperature vector of each thermal node, in units of ; for Diagonal heat capacity matrix, diagonal elements Indicates the first The equivalent heat capacity of each thermal node, in units of ; for Thermal conductivity matrix, diagonal elements Indicates the first The equivalent thermal conductance of each node to the environment, off-diagonal elements Indicates the first Node and the Thermal conductivity between nodes, in units of ; for The power vector of each load circuit, in units of ; for Power-heat source mapping matrix, elements Indicates the first The power loss of the first loop affects the power loss of the second loop. Contribution coefficient of each hot node; Ambient temperature, unit: ; for Environment Coupling Vector Power loss of each circuit according to Calculation, where For the first Effective value of loop current This is the equivalent resistance of the corresponding conductor and terminal.

[0034] Forward Euler stability condition: When using the forward Euler method for discretization, the sampling period should satisfy the stability condition: That is, it is less than the system's minimum thermal time constant. In this embodiment, the minimum thermal time constant is... = 850 / 1.85 ≈ 459s, which is much larger than the sampling period of 0.1s, so the discretization format is stable.

[0035] The specific steps of the offline identification experiment include: Step 1: Place the electricity metering box in an unloaded state and record the initial temperature of each node as a baseline; Step Two: For the first The circuit is subjected to a known constant power Continue heating until the temperature of all nodes reaches a steady state; Step 3: Calculate the thermal conductivity matrix elements from the steady-state temperature difference: in, For the first Steady-state temperature of the node Ambient temperature; Step 4: Cut off the first Loop power, recording the natural cooling transient response of each node temperature, and calculating the heat capacity from the time constant of the cooling curve: [The remaining text appears to be a fragmented and incomplete sentence, possibly due to OCR errors. A more accurate translation would require the full context.] The first-order time constant of the node cooling curve and the second-order time constant The product of the equivalent thermal conductance of each node to the environment is calculated.

[0036] Step 5: Repeat steps 2 to 4 for each loop to obtain the complete thermal conductivity matrix. and heat capacity matrix ; Step Six: Determine the matrix based on the power-heat source mapping relationship. ; For the Power applied to the circuit individually At that time, the steady-state temperature rise of each node satisfies: Therefore, the inverse solution yields... The List.

[0037] For example, based on the box structure and heat conduction path, the system is simplified to 4 thermal nodes, corresponding to 4 measuring points, and a 4-node thermal network model is established.

[0038] Initial model parameters were obtained through offline identification experiments: Heat capacity matrix ,unit: : Thermal conductivity matrix Unit: W / ℃ Power-heat source mapping matrix (4×8 dimensions), taking loop 3 as an example, its copper heat loss is mainly transferred to node T1. =0.72 The coupling coefficient of the remaining nodes is less than 0.1.

[0039] S3. In each control cycle, using the current node temperature and loop power as initial conditions, the electrothermal coupling dynamic model is integrated forward in the prediction time domain to obtain the predicted temperature trajectory of each node; with the node temperature not exceeding the safety threshold as a hard constraint, the optimal control sequence is solved by rolling time domain optimization, and only the first set of control actions at the current moment is executed; the state space equation established in step S2 is discretized, and the future temperature trajectory is recursively calculated based on the current temperature state and the current power of each loop within the prediction time domain step.

[0040] Discretization was performed using the forward Euler method, and the discretized state equations are as follows: in , , , It is an identity matrix.

[0041] In each control cycle, constrained by the predicted temperature value of the future prediction time step, the following multi-objective cost function is constructed: in: The first term is the weighted quadratic norm of the deviation between the temperature of each node and the safety threshold in the predicted time domain. The second item is a penalty for controlling the number of control actions within the time domain. The third item is a weighted penalty for the impact on user electricity consumption; , These are the weighting coefficients; , This is a diagonal matrix representing the priority weights of each loop; This is the security threshold vector for each node.

[0042] The method for obtaining the safety threshold is as follows: based on the rated operating temperature of each component in the power metering box, including circuit breakers, terminals, and wire insulation, take the lowest value among the rated operating temperatures of each component, and then subtract the preset safety margin to obtain the safety threshold.

[0043] By solving the above optimization problem, the optimal control sequence is obtained. Only the first set of control actions in the sequence is executed. In the next control cycle, the latest temperature data is collected as the initial state, and the above rolling time-domain optimization process is repeated.

[0044] The prediction time domain takes 10 to 60 steps, and the control time domain takes 3 to 10 steps. The specific values ​​are determined based on the thermal time constant and the control response speed requirements.

[0045] The quadratic programming solution employs the effective set method, specifically including: The on / off decision variables of each circuit are relaxed into continuous variables with a value range of [0,1], where 0 represents complete disconnection and 1 represents maintaining power supply. Transform the cost function into the standard QP form: The constraints are: The effective set method is used for iterative solution. The initial effective set is an empty set, and the effective set is updated in each iteration according to the constraint violation situation. For scenarios requiring integer decisions, i.e., binary decisions for on / off states, a rounding strategy is used to map continuous solutions to integer solutions: when the optimized solution is greater than 0.5, it is determined to maintain power supply; when it is less than or equal to 0.5, it is determined to disconnect. On the ARM Cortex-M4 microcontroller platform, the QP problem corresponding to the 4-node × 30-step prediction time domain has a dimension of 24. The solution time for a single effective set method is about 3.2ms, which meets the real-time requirement of a 500ms control cycle.

[0046] The temperature prediction uncertainty is evaluated based on the inherent parameter covariance matrix of the recursive least squares online identification algorithm with a forgetting factor. The recursive least squares algorithm with a forgetting factor maintains the parameter estimation covariance matrix during the recursive process. This matrix reflects the uncertainty of the current parameter estimation and is the standard output of the recursive least squares algorithm with a forgetting factor, requiring no additional computation.

[0047] The sensitivity of the predicted temperature to the model parameters is characterized by the Jacobian vector. In this embodiment, the Jacobian vector is calculated analytically: since the state-space equations are linear systems, the partial derivatives of the predicted temperature with respect to the parameters can be obtained by chain rule differentiation of the state transition matrix, without the need for numerical differentiation, and the computational complexity is the same as that of a first-order forward prediction.

[0048] No. The standard deviation of the predicted temperature is: The maximum prediction standard deviation within the prediction time domain is compared with a preset uncertainty threshold; where To predict temperature; It is a Jacobian vector.

[0049] If the maximum prediction standard deviation within the prediction time domain is less than or equal to the preset uncertainty threshold, the current prediction time domain is maintained. If the maximum prediction standard deviation within the prediction time domain is greater than the preset uncertainty threshold, the prediction time domain is gradually shortened in steps of 1 until the constraint is met. The lower limit of the prediction time domain is 3 control periods. If the constraint is still not met after shortening to the lower limit, the above-mentioned model self-evolution mechanism is triggered.

[0050] S4. Based on the temperature prediction trajectory, before the temperature at any node reaches the safety threshold, generate load control commands for one or more loops, and perform on / off or power regulation on the corresponding loops. Based on the optimization results of step S3, control commands for each load loop are generated. These control commands include three types: load shedding, load reduction, and load holding.

[0051] The following constraints must be followed when generating control commands: 1. A hard constraint is that the predicted temperature of all hot nodes does not exceed the safety threshold. 2. Minimize the total outage time of the disconnected or deloaded circuits while satisfying hard constraints; 3. The power consumption is sorted according to the preset priority, from high to low as follows: fire protection and emergency circuits, elevator power supply circuits, public lighting circuits, residential power circuits, and air conditioning and high-power socket circuits; 4. The number of circuits cut off in a single control action shall not exceed 30% of the total number of circuits to avoid large-scale power outages.

[0052] When a disconnected circuit needs to have its power restored, the following peak-shifting power restoration strategy is adopted: S41. Based on the dynamic changes in the surge injection heat energy of each circuit to be restored and the remaining heat capacity margin of each node, generate a power restoration sorting queue in real time. Specifically, obtain the electrical parameters of all circuits to be restored and the current thermal status of each node in the enclosure; Based on the formula for the remaining heat capacity margin of each hot node or an online learning model using historical switching data; The method of obtaining surge injection thermal energy through online learning of historical switching data specifically includes: The system continuously records the current waveform, transient temperature response of relevant nodes, and ambient temperature at the time of closing each circuit operation before and after the operation. It then filters out effective samples with significant temperature rise and no superimposed thermal disturbance to construct a historical switching sample set. For each valid sample, based on the measured temperature rise Inversely extrapolating the actual injected heat energy from the current heat network model As output labels, For equivalent heat capacity, This is a correction factor for heat transfer efficiency; Using the remaining heat capacity margin of each thermal node and the ambient temperature deviation as input features, incremental regression fitting is performed using the same recursive least squares method with a forgetting factor as the online update of the parameters. When the cumulative number of newly added valid samples reaches a preset threshold, the model is automatically updated, and the corrected surge injection thermal energy estimate is output. The estimated thermal energy value and the calculated value from the theoretical formula are weighted and fused according to the model confidence level to form the final surge injection thermal energy. In the cold start stage where historical samples are insufficient, the calculated value from the theoretical formula is used directly to ensure the availability of the power restoration strategy.

[0053] S42. The surge injection heat energy is calculated by multiplying the square of the rated current of the circuit, the square of the inrush current multiple, the equivalent resistance of the circuit, and the closing duration. The effective value of the closing surge current is the product of the rated current and the inrush current multiple, or obtained through online learning of historical switching data. Calculate the remaining heat capacity margin for each hot node: ,in For equivalent heat capacity, This is the current temperature.

[0054] Estimate the heat energy injected by the closing surge of the circuit to be restored: ,in This is the effective value of the closing surge current. The equivalent resistance of the circuit. The surge duration is defined as follows: The system generates an initial power restoration sorting queue in real time based on the magnitude of the heat energy injected by the closing surge of the circuit to be restored and the distribution of the remaining heat capacity margin of each hot node.

[0055] Comparative judgment: If the heat energy injected by the closing surge of the circuit to be restored is less than the remaining heat capacity margin of each hot node, then the circuit is allowed to be restored; otherwise, the restoration is postponed and reassessed after the temperature drops further.

[0056] S43. During the execution of the power restoration sequencing queue, if the remaining heat capacity margin of any node is reduced to below the surge injection heat energy of the next circuit to be restored due to the input of new load or changes in ambient temperature, the queue execution is suspended, the global temperature trajectory is re-performed based on the electrothermal coupling dynamic model, and the power restoration sequence is readjusted or the waiting interval is increased according to the pre-performance results. Specifically, when multiple circuits need to be restored, the restoration queue is arranged in order of surge injection heat energy from small to large. An interval of not less than the preset waiting time is set between two adjacent restoration operations, and the remaining heat capacity margin is reassessed before each restoration.

[0057] The inrush current multiple is determined based on the circuit load type: for resistive loads, the inrush current multiple is 1.0 to 1.5; for inductive loads, the inrush current multiple is 6 to 8; and for capacitive loads, the inrush current multiple is 3 to 5. When load type information is unavailable, a conservative inrush current multiple of 6 times the rated current is used.

[0058] For example, the execution layer uses intelligent miniature circuit breakers with communication interfaces to receive opening and closing commands from the microcontroller unit. For circuits requiring load reduction, pulse width modulation (PWM) power regulation is performed using solid-state relays connected in series in the circuit, with a duty cycle adjustment range of 0% to 100% and an adjustment step of 5%. When using solid-state relays for zero-crossing triggered PWM power regulation, the impact on the metering accuracy of the energy meter is negligible because the zero-crossing triggering method does not generate high-order harmonics. The load reduction mode is activated briefly when the temperature prediction approaches the safety threshold, with a duration not exceeding the prediction time domain length, and will not have a significant impact on user electricity consumption statistics.

[0059] S5. Compare the actual temperature response after execution with the temperature prediction trajectory. When the deviation exceeds the preset allowable range, trigger the online update of the parameters of the electrothermal coupling dynamic model to form a closed-loop control. The control command generated in step S4 is sent to the execution layer to control the intelligent miniature circuit breaker or solid-state relay of the corresponding circuit to perform on / off operations. After the control command is executed, the actual temperature response of each hot node is continuously collected, and the actual temperature is compared with the predicted temperature in step S3.

[0060] For disconnection commands, the circuit breaker of the corresponding circuit is controlled to open; for load reduction commands, pulse width modulation is performed through solid-state relays to reduce the actual power supply of the circuit; for hold commands, no operation is performed.

[0061] After the action is completed, record the action time, action type, and corresponding circuit number.

[0062] S51. After the execution of the load control command causes the switching operation of one or more circuits, the transient temperature response data and the corresponding power change data caused by the switching operation are collected. Specifically, starting from the moment of the on / off operation, the temperature sampling sequence of each node is continuously collected within a preset transient time window, and the power change sequence of each circuit and the ambient temperature sampling sequence are recorded simultaneously to form the transient sample dataset of this on / off operation.

[0063] S52. Discretize the state-space equation into a difference form, and construct a linear regression equation with the temperature change of each node as the output and the power change of each loop and the ambient temperature change as the input.

[0064] Specifically, the state-space equation described above is discretized with a sampling period to obtain the vector of node temperature changes, which is then used to construct a linear regression equation: in, This is the vector of node temperature changes; It is a regression vector composed of the power changes of each loop and the ambient temperature changes; It is a stacked vector of the parameters to be estimated in the thermal capacity matrix, thermal conductivity matrix, and environmental coupling vector; For measuring noise.

[0065] S53. The parameters in the heat capacity matrix, thermal conductivity matrix and environmental coupling vector are updated online using the recursive least squares method with a forgetting factor. The forgetting factor is used to adjust the relative weight of historical measurement data and current measurement data in the parameter recursive estimation, so that the model parameters are updated adaptively with changes in load conditions.

[0066] Specifically, the recursive update formula is: in, Forgetting factor, For the gain vector, Let covariance matrix be the variance matrix. It is the identity matrix. The current estimated value of the parameter vector to be identified. Let be the covariance matrix of the previous time step. The forgetting factor ranges from greater than 0.8 to less than or equal to 1: when the forgetting factor is close to 1, historical data is given higher weight, and parameter estimation is smoother, suitable for scenarios with slow load changes; when the forgetting factor is close to 0.8, recent data is given higher weight, and parameter tracking speed is faster, suitable for scenarios with frequent load switching. Preferably, the forgetting factor is between 0.95 and 0.999 to balance parameter tracking speed and estimation stability.

[0067] After the parameters are updated, the electrothermal coupling dynamic model is reconstructed using the updated thermal capacity matrix, thermal conductivity matrix, and environmental coupling vector. The temperature prediction trajectory for the next cycle is recalculated, and the subsequent control threshold is adjusted accordingly to achieve closed-loop adaptive control.

[0068] S54: Set the confidence criterion for parameter updates: When the data length of the transient process is less than the preset minimum number of samples or the signal-to-noise ratio is lower than the preset threshold, the corresponding parameters are not updated temporarily, and the parameter values ​​of the previous period are maintained.

[0069] Specifically, a statistical evaluation is performed on the transient sample dataset for each on / off operation: if the effective data length is less than the preset minimum number of samples, or the signal-to-noise ratio of the transient temperature response is lower than the preset threshold, then the data set will not participate in the recursive update, and the estimated value of the parameter vector to be identified at the previous moment will be maintained; only when the effective data length is greater than or equal to the preset minimum number of samples and the signal-to-noise ratio of the transient temperature response is greater than or equal to the preset threshold is both satisfied, the above recursive update formula will be adopted to update the result.

[0070] When the model prediction residual shows a systematic deviation between the measured temperature value of the k-th sampling period and the model prediction temperature value of the k-th period for multiple consecutive control periods, the process of adding an auxiliary heat dissipation node is triggered. In this embodiment, the determination of systematic deviation adopts the joint criterion of the following two tests: (a) The length of consecutive segments with the same sign in the residual sequence exceeds the preset proportion of the total number of samples, indicating that the bias is directional rather than random fluctuation; (b) The absolute value of the first-order autocorrelation coefficient of the residual sequence exceeds the preset threshold, indicating that there are dynamic components in the residual that have not been captured by the model.

[0071] If both of the above tests are met, the current model is deemed to be of insufficient order.

[0072] The specific steps for adding auxiliary heat dissipation nodes are as follows: Select the node with the largest absolute value of the mean of the predicted residuals. The physical region represented by the node is divided into two sub-regions on the thermal network. The original node is retained, and the new node represents the sub-region with the larger thermal gradient in the original region. A thermal resistance connection is established between the new node and the original node, and its initial parameters are determined based on the physical constraints of the region division: the sum of the heat capacities of the two sub-regions after division should be approximately equal to the heat capacity of the original region, and the thermal resistance is distributed inversely proportional to the heat dissipation area of ​​the sub-regions.

[0073] After the addition is completed, the thermal network model with the added auxiliary heat dissipation nodes is re-evaluated online using recursive least squares method with a forgetting factor, based on historical temperature data from the most recent preset period. After the re-evaluation, the prediction accuracy of the updated thermal network model is verified using measured data from subsequent preset verification periods: if the root mean square prediction error of the updated thermal network model decreases by more than a preset percentage compared to before the addition, then the updated thermal network model is officially adopted; otherwise, it reverts to the thermal network model before the addition.

[0074] In this embodiment, the model is allowed to add a maximum of 2 auxiliary heat dissipation nodes. Once this limit is exceeded, the addition process will no longer be triggered.

[0075] This application also provides a load intelligent control system for an electricity metering box, including: The sensing module includes a through-hole Hall current sensor array, an NTC thermistor temperature sensor array, and an ambient temperature sensor, used to collect electrical quantity data of each circuit and enclosure temperature data. The modeling and prediction module, including an edge computing processor, is used to build an electrothermal coupling dynamic model and perform rolling time-domain temperature prediction. The decision module, running in the edge computing processor, is used to perform multi-objective optimization solutions and staggered power restoration sorting. The execution module, including intelligent miniature circuit breakers and / or solid-state relays, is used to perform on / off and power regulation operations of load circuits; The online parameter update module runs in the edge computing processor and is used to perform online parameter identification and model updates.

[0076] The edge computing processor uses a microcontroller with an ARM Cortex-M4 or higher core, a main frequency of not less than 168MHz, and a floating-point arithmetic unit. The random access memory of the microcontroller is not less than 192KB, and the flash memory capacity is preferably not less than 1MB.

[0077] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for intelligent load control of an electricity metering box, characterized in that, Includes the following steps: S1: Collect electrical quantity data of each outgoing circuit in the power metering box and temperature data of each node; S2: The lumped parameter method is used to convert the heat source and heat dissipation path in the power metering box into a multi-node thermal network. The real-time power of each loop is used as the heat source excitation input and the temperature of each node is used as the state variable to establish a state-space form of electrothermal coupling dynamic model. Among them, the off-diagonal elements of the thermal conductivity matrix represent the thermal coupling propagation relationship between different loops through the box wall, air layer or common mounting plate. S3: In each control cycle, with the current temperature of each node and the power of the loop as the initial conditions, the electrothermal coupling dynamic model is integrated forward in the prediction time domain to obtain the predicted temperature trajectory of each node; with the temperature of each node in the prediction time domain not exceeding the safety threshold as a hard constraint, the optimal control sequence is solved by rolling time domain optimization. S4: Based on the temperature prediction trajectory, before the temperature prediction at any node reaches the safety threshold, generate load control instructions for one or more loops, and perform on / off or power regulation on the corresponding loops. S5: Compare the actual temperature response after execution with the predicted temperature trajectory. When the deviation exceeds the preset allowable range, trigger the online update of the parameters of the electrothermal coupling dynamic model to form a closed-loop control.

2. The intelligent load control method for an electricity metering box according to claim 1, characterized in that, The electrothermal coupling dynamic model, which uses the real-time electrical power of each loop as the heat source excitation input and the temperature of each node as the state variable, to establish a state-space form includes: Using the real-time electrical power of each loop as the heat source excitation input of the thermal network and the temperature of each node as the state variable, the state-space equation of the following form is established: in, Here is the node heat capacity matrix. This is a thermal conductivity matrix, where the diagonal elements represent the equivalent heat conduction of each node to the environment, and the off-diagonal elements represent the thermal conductivity. Representation Nodes With nodes The thermal coupling strength between them via the enclosure wall, air layer or common mounting plate; For the node temperature vector, For each circuit's power vector, For ambient temperature, This is the power-heat source mapping matrix. This is the environment coupling vector; Each component of the power vector is obtained by multiplying the square of the effective value of the current collected in real time in each circuit with the equivalent resistance of the corresponding conductor and terminal, and characterizes the copper loss heating power of each circuit.

3. The intelligent load control method for an electricity metering box according to claim 1, characterized in that, The rolling time-domain optimization solution for the optimal control sequence includes: In each control cycle, the electrothermal coupling dynamic model is solved within a preset prediction time domain, using the current temperature and loop power status of each node as initial conditions, to obtain the temperature prediction trajectory of each node. Construct the cost function: The first term is the weighted quadratic norm of the deviation between the temperature of each node and the safety threshold in the predicted time domain, and the second term is the number of control actions in the time domain. The third penalty is a weighted penalty for the impact on the user's electricity consumption; , These are the weighting coefficients; , This is a diagonal matrix representing the priority weights of each loop; This represents the security threshold vector for each node. Under the hard constraint that the temperature of each node does not exceed the safety threshold, the optimal control sequence of the cost function is solved, and the first set of control actions at the current moment is executed; in the next control cycle, the initial state is updated using the latest acquired data, and the rolling time-domain optimization process is repeated.

4. The intelligent load control method for an electricity metering box according to claim 3, characterized in that, The weight allocation methods for each optimization objective in the cost function include: The primary optimization objective is to minimize the total duration during which each circuit is disconnected or deloaded. The second optimization objective is to maximize the power supply continuity of high-priority user circuits. The third optimization objective is to minimize the impact on user electricity consumption. The first optimization objective is approximately optimized through the second term of the cost function, which is the penalty for the number of control actions. The second optimization objective corresponds to the priority weight configuration in the cost function, and the third optimization objective corresponds to the third term of the cost function, which is the penalty for the impact of user electricity consumption.

5. The intelligent load control method for an electricity metering box according to claim 2, characterized in that, The method also includes a peak-shaving power restoration strategy: Acquire the surge injection heat energy of each circuit to be restored; Based on the state-space equation of the electrothermal coupling dynamic model, a global multi-node temperature trajectory prediction is performed on the pseudo-re-energization sequence containing all circuits to be re-energized: through the forward integration of the state-space equation, the transient thermal process after the re-energization of each circuit to be re-energized is simulated in sequence. The temperature rise caused by the re-energization of each circuit is propagated to other nodes through the off-diagonal elements of the thermal conductivity matrix in the electrothermal coupling dynamic model, so that the remaining heat capacity margin of the node corresponding to the circuit to be re-energized later in the pseudo-re-energization sequence is affected by the cumulative thermal contribution of the circuits already re-energized earlier in the sequence. If the simulation results show that the temperature of any node exceeds the safety threshold during the execution of the proposed power restoration sequence, then at least one of the following adjustments will be performed iteratively: adjust the power restoration sequence, insert a waiting interval between adjacent power restorations, or reduce the power restoration power of the circuit to be restored, until the simulation passes and the power restoration operation is performed according to the determined power restoration sequence.

6. The intelligent load control method for an electricity metering box according to claim 5, characterized in that, The peak-shaving power restoration strategy includes: Based on the dynamic changes in the surge injection heat energy of each circuit to be restored and the remaining heat capacity margin of each node, a power restoration sorting queue is generated in real time. The surge injection heat energy is calculated by multiplying the square of the rated current of the circuit, the square of the inrush current multiple, the equivalent resistance of the circuit, and the closing duration. The effective value of the closing surge current is the product of the rated current and the closing inrush current multiple, or obtained through online learning of historical switching data. During the power restoration sequencing process, if the remaining heat capacity margin of any node is reduced to below the surge injection heat energy of the next circuit to be restored due to the input of new load or changes in ambient temperature, the queue execution is suspended, the global temperature trajectory is re-performed based on the electrothermal coupling dynamic model, and the power restoration sequence is readjusted or the waiting interval is increased according to the re-performance results.

7. The intelligent load control method for an electricity metering box according to claim 1, characterized in that, The online parameter updates include: After executing a load control command that causes one or more circuits to switch on or off, collect transient temperature response data and corresponding power change data caused by the switching operation. The state-space equation of the electrothermal coupling dynamic model is discretized into a difference form. A linear regression equation is constructed with the temperature change of each node as the output and the power change of each loop and the ambient temperature change as the input. The parameters in the thermal capacity matrix, thermal conductivity matrix and environmental coupling vector of the electrothermal coupling dynamic model are updated online using the recursive least squares method with forgetting factor. The forgetting factor is used to adjust the relative weight of historical measurement data and current measurement data in the parameter recursive estimation, so that the model parameters are updated adaptively with the change of load conditions. The confidence criterion for setting parameter updates is as follows: when the data length of the transient process is less than the preset minimum number of samples or the signal-to-noise ratio is lower than the preset threshold, the corresponding parameters are not updated temporarily, and the parameter values ​​of the previous period are maintained.

8. The intelligent load control method for an electricity metering box according to claim 7, characterized in that, The method also includes a model self-evolution mechanism: The actual temperature response after regulation is compared with the predicted temperature trajectory on a node-by-node basis. When the prediction deviation of any node exceeds the preset allowable range for a certain number of times, a preset threshold is reached, the forced re-identification of the corresponding node parameters in the thermal conductivity matrix and thermal capacity matrix is ​​triggered. When the length of the consecutive same number segment of the deviation exceeds the preset proportion of the total number of samples and the absolute value of the first-order autocorrelation coefficient exceeds the preset threshold, it is determined that the number of nodes in the current thermal network model is insufficient to characterize the actual heat distribution inside the box. In the thermal network model, auxiliary heat dissipation nodes are added to improve the model order. The thermal network model is the node network structure composed of thermal resistance and thermal capacity in the electrothermal coupling dynamic model. When the number of circuits in the power metering box changes due to increases or decreases, the dimension of the state space equation is automatically reconstructed, and the corresponding heat source nodes and row and column elements of the thermal conductivity matrix are added or deleted.

9. The intelligent load control method for an electricity metering box according to claim 3, characterized in that, The rolling time-domain optimization also includes adaptive adjustment methods for the prediction time domain and the control time domain: The average deviation between the actual temperature response and the predicted temperature trajectory within the most recent preset number of control cycles is statistically analyzed. When the average deviation increases, the prediction time domain length is shortened to reduce the uncertainty of long-term prediction, while the control time domain length is shortened to increase the control frequency. When the average deviation decreases, the prediction time domain length is extended to increase the time margin of the feedforward prediction. The adjustment range of the prediction time domain length is limited to ensure that the temperature prediction uncertainty of the farthest prediction node does not exceed a preset confidence level. The preset confidence level is the maximum value of the temperature prediction standard deviation of each node in the prediction time domain does not exceed a preset threshold. The preset threshold is determined based on the difference between a safety threshold and a control trigger margin. The control trigger margin is the difference between the safety threshold and the temperature threshold that triggers the control action.

10. A load intelligent control system for an electricity metering box, used to implement the load intelligent control method for an electricity metering box as described in any one of claims 1-9, characterized in that, include: The sensing module is used to collect electrical quantity data of each outgoing circuit and temperature data of each node in the power metering box. The modeling and prediction module is used to build an electrothermal coupling dynamic model of a multi-node thermal network with lumped parameters based on the collected data. The off-diagonal elements of the thermal conductivity matrix represent the thermal coupling propagation relationship between different loops. The temperature prediction trajectory of each node is obtained by forward integration in each control cycle, and the optimal control sequence is solved by rolling time-domain optimization. The decision module is used to generate load control instructions before the temperature prediction trajectory of any node reaches the safety threshold, and to perform a global temperature trajectory pre-simulation of the proposed power restoration sequence based on the electrothermal coupling dynamic model during power restoration to determine the power restoration sequence and waiting interval. The execution module is used to perform on / off switching or power regulation on the corresponding outgoing circuit according to the load control command; The parameter online update module is used to compare the deviation between the actual temperature response and the temperature prediction trajectory. When the deviation exceeds the preset allowable range, the parameters of the electrothermal coupling dynamic model are updated online and fed back to the modeling and prediction module to form a closed-loop control.

Images