Intelligent power grid power-saving control system

Abstract

The invention relates to the technical field of power electronics, and particularly discloses an intelligent power grid electricity-saving control system which comprises a tail end sensing layer, a data processing layer and a control execution layer. The tail end sensing layer comprises a distributed multi-parameter sensing unit, the distributed multi-parameter sensing unit is arranged in a workshop power supply loop far away from a power distribution room, and the distributed multi-parameter sensing unit comprises a voltage sensor, a current sensor and a harmonic detection sensor; the voltage sensor and the current sensor are used for synchronously collecting terminal real-time voltage data and current data, and the harmonic detection sensor is used for collecting harmonic component data of a terminal power grid; according to the technical scheme, the tail end voltage detection is high in reliability, high in real-time performance and high in adaptability, the single-point fault risk can be avoided, and the electric energy quality can be guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power electronics, and particularly relates to an intelligent power grid energy-saving control system. BACKGROUND

[0002] With the expansion of industrial production scale and the improvement of the complexity of electrical equipment, the power grid system needs to realize energy efficiency optimization under the premise of ensuring the normal operation of the equipment, and the intelligent power grid energy-saving control system emerges as the times require. The intelligent power grid energy-saving control system is based on modern power electronics technology, automatic control technology, Internet of Things (IoT) and artificial intelligence (AI) algorithm, covers the whole link of power generation, power transmission, power distribution and power consumption, realizes real-time monitoring, intelligent analysis and accurate regulation of the flow direction of electric energy, the load state and the energy consumption loss, balances the supply and demand relationship of the power grid, reduces the invalid loss, and the core goal is to realize the dual values of the overall energy efficiency improvement and the safe and stable operation of the power grid. The intelligent power grid energy-saving control system has been widely applied to large industrial parks, manufacturing enterprises and other scenes, and has become a key technical means for energy saving and loss reduction in the industrial field.

[0003] In the actual application of the current mainstream intelligent power grid energy-saving control system, there are still problems of insufficient reliability and real-time performance of end voltage detection. The intelligent power grid energy-saving control system relies on the end voltage sensor arranged far away from the power distribution room and the workshop to collect key data. The transmission distance between the sensor and the power distribution room / regional control center is usually tens of meters to hundreds of meters, and some large industrial parks are even farther. Influenced by the physical characteristics of the link, the signal is easily affected by industrial electromagnetic interference, wireless signal shielding and other factors in the transmission process, and problems such as data jumping and packet loss occur. At the same time, long-distance transmission leads to obvious delay of data, and the delay time in some scenes can be more than 500 ms. Distorted or lagged voltage data as the core input of control decision will cause the intelligent detection and control logic to deviate from the actual working condition, which not only cannot realize accurate energy saving, but also may exacerbate the end voltage fluctuation and affect the stability of equipment operation.

[0004] The existing control algorithm model also has the problem of poor adaptability. The reason is that the intelligent control algorithm of the existing smart grid power saving control system is designed around a specific power grid model. Common algorithms include static optimization algorithms such as genetic algorithms and particle swarm optimization algorithms. These algorithms need to be based on a pre-constructed power grid mathematical model (such as a node impedance matrix model) when applied, and the model parameters use the rated value or design value of the power grid. The default parameters remain unchanged during the running period. For example, a particle swarm optimization algorithm is used to calculate the voltage reduction amplitude in a certain industrial park power saving system. The line resistance, load power factor and other parameters initially input into the algorithm are all fixed values during the design stage of the power grid, without considering the dynamic changes in actual operation. However, in actual power grid operation, parameters will be dynamically adjusted according to the environment and working conditions. Changes in environmental temperature will cause fluctuations in line resistance (higher than the rated value in summer), and the start and stop of devices such as motors and machine tools will cause fluctuations in parameters such as load power factor and current. These dynamic changes cause a serious mismatch between the fixed model and the actual power grid. The mismatch between the model and the actual power grid results in low efficiency of the optimal power saving set value calculated by the algorithm when actually executed, large deviation in voltage reduction amplitude, and in some cases even causes device abnormalities, such as insufficient torque, overheating or locked rotor problems in motor loads due to excessive voltage reduction.

[0005] And there is a single point failure risk during operation. The control system needs to connect the end sensor, central controller and execution unit through a communication network. Most current systems use a single communication link (such as relying only on RS485 wired transmission or LoRa wireless transmission), which lacks redundancy design. When the communication link is interrupted or delayed due to line aging failure, industrial environment electromagnetic interference or network attack, a single point failure occurs, causing the entire system to fail or issue incorrect control instructions. For example, the interruption of a LoRa wireless link due to electromagnetic interference from workshop equipment causes the central controller to not receive end voltage data, triggering an incorrect instruction, which can cause the workshop production line motors to stop rotating, completely losing the power saving control function, and also causing production interruption, resulting in direct economic losses.

[0006] Existing systems focus on voltage amplitude adjustment to achieve power saving, only calculating the optimal voltage reduction value using the above-mentioned genetic algorithm and particle swarm optimization algorithm, ignoring possible harmonics, three-phase imbalance and other problems in the end power grid. The single voltage reduction strategy and these power quality problems are prone to coupling effects, i.e. the voltage reduction operation may amplify harmonic components, resulting in increased line loss compared to normal operating conditions, which is contrary to the goal of power saving; at the same time, harmonics can accelerate the insulation aging of devices such as motors and transformers, shorten the service life of devices, increase the operating and maintenance costs of enterprises, and further reduce power saving benefits.

[0007] Therefore, there is an urgent need for a smart grid power saving control system with high reliability, strong real-time performance, strong adaptability, the ability to avoid single point failure risks, and the ability to ensure power quality. SUMMARY

[0008] The application provides an intelligent power grid energy-saving control system, which has high reliability and real-time performance in end voltage detection, strong adaptability, avoids single-point failure risk and guarantees power quality.

[0009] To solve the above technical problems, the application provides the following technical scheme:

[0010] An intelligent power grid energy-saving control system comprises an end perception layer, a data processing layer and a control execution layer.

[0011] The end perception layer comprises a distributed multi-parameter sensing unit, which is arranged on a workshop power supply circuit far from a power distribution room, and comprises a voltage sensor, a current sensor and a harmonic detection sensor, wherein the voltage sensor and the current sensor are used for synchronously collecting end real-time voltage data and current data, and the harmonic detection sensor is used for collecting harmonic component data of an end power grid; meanwhile, the distributed multi-parameter sensing unit is internally provided with an edge preprocessing module, which preliminarily filters and denoises the collected voltage, current and harmonic data to obtain preliminary preprocessing data, and removes abnormal data caused by industrial electromagnetic interference.

[0012] The data processing layer comprises a data verification and retransmission unit and a dynamic model identification unit; the data verification and retransmission unit is used for performing real-time verification on the preliminary preprocessing data of the edge preprocessing module according to a CRC (Cyclic Redundancy Check) algorithm, and if data errors or loss are found, immediately sends a retransmission instruction to the distributed multi-parameter sensing unit to ensure data integrity; if no data errors or loss are found, outputs clean data; the dynamic model identification unit is used for calculating real-time parameters of a power grid by using a recursive least square method and an extended Kalman filter algorithm according to the clean data, correcting the real-time parameters by combining with acquired workshop environment temperature data to obtain corrected parameters, continuously iteratively updating a power grid mathematical model based on the corrected parameters, establishing a light-load, heavy-load and variable-temperature sub-scenario model library, calculating a real-time load rate according to the clean data, calling a matched sub-model from the sub-scenario model library according to the real-time load rate, and obtaining a dynamic power grid model adapted to actual working conditions.

[0013] The control execution layer comprises an adaptive regulation unit and a power quality treatment unit; the adaptive regulation unit is used for receiving the dynamic power grid model output by the dynamic model identification unit, comparing the terminal voltage theoretical value calculated by the dynamic power grid model with the clean data output by the data verification and supplementary transmission unit, calculating a deviation value, and automatically adjusting the control strategy parameters if the deviation value is greater than a preset threshold; meanwhile, the adaptive regulation unit is preset with a model prediction algorithm, generates a pre-regulation instruction according to the clean data to predict the future terminal load change trend in a preset time period, and fuses the adjusted control strategy parameters and the pre-regulation instruction to obtain a cooperative regulation instruction; the power quality treatment unit is used for receiving the harmonic component data collected by the distributed multi-parameter sensing unit and the cooperative regulation instruction output by the adaptive regulation unit, determining the harmonic suppression demand according to the harmonic component data, outputting a harmonic suppression signal to the filter device of the workshop power supply circuit in combination with the cooperative regulation instruction, and simultaneously monitoring the clean data in real time to judge the three-phase voltage balance degree, and triggering a three-phase voltage regulation mechanism to generate a three-phase voltage regulation signal when the three-phase voltage imbalance degree exceeds a limited value.

[0014] The basic scheme principle and beneficial effects are as follows: through the cooperative work of the three-layer architecture of the terminal sensing layer, the data processing layer and the control execution layer, the present application realizes the full-process closed-loop management of data from collection, processing to regulation.

[0015] The distributed multi-parameter sensing unit of the terminal sensing layer is directly arranged in the workshop power supply circuit away from the distribution room, synchronously collects terminal real-time voltage / current data through voltage sensors and current sensors, and additionally collects harmonic component data through a harmonic detection sensor, so as to realize multi-dimensional coverage of the power grid operation state; the built-in edge preprocessing module preliminarily denoises abnormal fluctuations (such as voltage spikes and current clutter) in the collected data caused by industrial electromagnetic interference through a filtering algorithm (such as mean filtering and Kalman filtering), converts the original data into preliminary preprocessing data, and completes the first purification at the data collection end, laying a foundation for subsequent data processing.

[0016] After receiving the preliminary preprocessing data output by the terminal sensing layer, the data verification and supplementary transmission unit verifies the data integrity and accuracy in real time based on the CRC cyclic redundancy check algorithm: if the verification finds data errors (such as missing data bits and unmatched check codes) or loss, immediately sends a retransmission instruction to the distributed multi-parameter sensing unit to avoid invalid data entering the subsequent process; if the verification is passed, clean data without errors or loss is output, ensuring the reliability of the core control data.

[0017] The dynamic model identification unit takes clean data as input, calculates real-time impedance, load power factor and other parameters of the power grid by recursive least squares method, and then corrects the real-time parameters in combination with the environmental temperature data of the workshop (such as the increase of line resistance caused by high temperature) to obtain corrected parameters that fit the actual working conditions; based on the corrected parameters, the power grid mathematical model is continuously iterated and updated, and the real-time load rate is calculated according to the current data in the clean data, and the matching sub-model is called from the preset light load, heavy load and variable temperature scene sub-model library, and finally a dynamic power grid model completely adapted to the current power grid state is constructed, which gets rid of the limitations of traditional fixed models.

[0018] After receiving the dynamic power grid model, the adaptive control unit compares the terminal voltage theoretical value calculated by the model with the clean data (voltage data) output by the data processing layer, calculates the deviation value: if the deviation value exceeds the preset threshold, the control strategy parameters (such as the voltage reduction amplitude and the control frequency) are automatically adjusted; at the same time, based on the load change law (such as current fluctuation caused by the start and stop of workshop equipment) in the clean data, the future load trend in a preset time period (such as 5-10 seconds) is predicted through the built-in model prediction algorithm, and a pre-control instruction is generated; the adjusted control strategy parameters and the pre-control instruction are fused to obtain a collaborative control instruction that takes into account real-time correction and advance adaptation.

[0019] The power quality management unit receives harmonic component data collected by the distributed multi-parameter sensing unit on the one hand, analyzes the harmonic type and amplitude, and determines the harmonic suppression requirement; on the other hand, it receives the collaborative control instruction, combines the harmonic suppression requirement with the voltage control target, and outputs the harmonic suppression signal to the filtering device (such as an active power filter) of the workshop power supply loop; at the same time, by monitoring the three-phase voltage data in the clean data, the three-phase voltage unbalance degree (i.e. the voltage difference between phase A, B and C) is calculated, and when the unbalance degree exceeds the limit value, the three-phase voltage regulation mechanism (such as adjusting the output of each phase voltage regulation module) is triggered to generate a three-phase voltage regulation signal, realizing the synchronous performance of voltage optimization and power quality management.

[0020] Through the double protection of edge preprocessing module preliminary noise reduction + CRC check and retransmission, abnormal data (such as voltage data jump) caused by industrial electromagnetic interference can be effectively filtered out, and real-time retransmission of error / missing data can be performed to improve data accuracy and avoid control inaccuracy caused by data distortion in traditional systems;

[0021] The distributed multi-parameter sensing unit is directly laid in the terminal power supply loop, shortening the data collection distance, and the edge preprocessing completes preliminary data processing at the collection end, reducing the subsequent data transmission amount; at the same time, the data verification and retransmission adopts a real-time response mechanism to ensure that the clean data can be quickly output to the control execution layer, the total delay of data transmission and processing is controlled within an acceptable delay range, which is much lower than that of traditional systems, meeting the real-time monitoring demand of terminal voltage.

[0022] The dynamic model identification unit iteratively updates the dynamic power grid model in real time as the power grid parameters (such as line resistance, load rate) and environmental conditions (such as temperature) change through real-time parameter calculation, environmental correction, and scene-specific sub-model invocation. For example, when the line resistance increases by 15% due to the increased temperature in the workshop during summer, the parameter correction and sub-model switching can be completed within 1 second. In the face of different working conditions such as light load (e.g., load rate < 30%) and heavy load (e.g., load rate > 80%), the model adaptation accuracy is improved, ensuring that the control strategy always matches the actual power grid state, solving the low efficiency problem of traditional fixed models when the working condition changes, and adapting to the complex and variable load demand in industrial scenarios.

[0023] Compared with the traditional system design relying on a single sensor or a single data link, the system avoids single-point failures through multi-dimensional redundancy. The end perception layer uses a distributed sensing layout (e.g., 2-3 sensing units are installed in each workshop), so that other units can normally collect data when a single sensor fails. The dynamic model identification unit in the data processing layer has a model self-healing capability (e.g., automatically switches to other adaptive sub-models when a sub-model invocation fails). The adaptive control unit in the control execution layer can automatically adjust the strategy according to the data deviation without human intervention, avoiding the failure of the entire system due to a single component failure, and improving the stability of the system.

[0024] By accurately collecting harmonic data through harmonic detection sensors and outputting targeted suppression signals through filtering devices, the total harmonic distortion (THD) can be reduced, and the line loss caused by harmonics can be reduced.

[0025] The three-phase voltage regulation mechanism can control the three-phase voltage imbalance within an acceptable range, avoiding problems such as motor overheating and shortened equipment life caused by three-phase imbalance. At the same time, the cooperative control instruction ensures that voltage regulation and power quality management are synchronized, avoiding the coupling effect (such as amplifying harmonics) of traditional single voltage reduction strategies and power quality problems, achieving double optimization of power saving and power quality.

[0026] The model prediction algorithm of the adaptive control unit can capture the load mutation trend (such as the sudden start of multiple motors in the workshop causing a sudden increase in current) in advance, and the pre-control instruction can complete parameter adaptation before the load changes, shortening the response time to load mutations and avoiding voltage fluctuations caused by response lag in traditional systems, ensuring the stable operation of precision equipment (such as CNC machine tools).

[0027] The dynamic model identification unit adopts a combination of recursive least square method and extended Kalman filter algorithm, and the identification error of real-time parameters (such as impedance and power factor) of the power grid is small; the dynamic power grid model constructed based on accurate parameters improves the precision of the cooperative control instruction output by the adaptive control unit, reduces the control error of the voltage reduction amplitude, ensures the power saving effect, and avoids problems such as insufficient motor torque and locked rotor caused by excessive voltage reduction.

[0028] The application has the capabilities of autonomous data verification, model updating and fault warning (such as automatically adjusting the data transmission path when the link is abnormal, and sending alarm information when the harmonic exceeds the standard), reduces the frequency of manual inspection, and reduces the operation and maintenance labor cost;

[0029] All data processing and control logic are automatically completed by each unit without human intervention, and clean data can be used to generate power grid operation reports (such as load change curve and power quality index) to provide accurate energy consumption analysis basis for enterprises and help subsequent energy efficiency optimization.

[0030] The distributed multi-parameter sensing unit can support different voltage levels (such as 380V and 10kV) of workshop power supply circuits according to actual needs, the scene sub-model library of the dynamic model identification unit can be extended according to user needs (such as adding "impulse load" and "intermittent operation" sub-models according to actual conditions), and the power quality treatment unit can be compatible with different types of filtering devices and voltage regulating equipment, which is suitable for various industrial scenes such as automobile manufacturing, mechanical processing and chemical industry, without the need for large-scale system modification for different scenes, and has a wide range of applications.

[0031] In summary, the application realizes high reliability and strong real-time performance of end voltage detection, has strong adaptability, can avoid single point failure risk, and can ensure the purpose of power quality.

[0032] Further, it further includes a dual-link communication layer, the dual-link communication layer includes a main communication link and a backup communication link, the main communication link adopts optical fiber communication, and the backup communication link adopts a 5G industrial private network; the dual-link communication layer is used for real-time receiving of data transmitted between units of the end sensing layer, the data processing layer and the control execution layer, and simultaneously monitoring signal strength and error code rate of the main communication link, when the error code rate of the main communication link is greater than a first error code preset value or the signal strength is lower than a first strength preset value, the data transmission link is automatically switched to the backup communication link, and link switching alarm information is sent to a preset address.

[0033] Further, the edge preprocessing module adopts the following contents for preliminary filtering and noise reduction:

[0034]

[0035] Among them:

[0036] is the preliminary preprocessed data after filtering;

[0037] is the current data collection time, taking a positive integer, corresponding to the first data point collected per second by the sensor;

[0038] is the filter window length, taking a value of 5-10, which is adaptively adjusted according to the intensity of industrial electromagnetic interference in the workshop, taking 10 when the interference intensity is >80dB, and taking 5 when the interference intensity is <50dB;

[0039] is the weighting coefficient of the original data, satisfying , and , i.e., the center data has the highest weight, and the edge data has a decreasing weight, ensuring that the filtered data retains the original change trend;

[0040] is the original data collected at the time;

[0041] At the same time, the edge preprocessing module is also used to judge the intensity of electromagnetic interference by collecting the fluctuation variance of the original data, and if the fluctuation variance is > the preset threshold value for 3 consecutive times, the filter window length is automatically increased by 2, with a maximum of 10; if the fluctuation variance is ≤ the preset threshold value for 5 consecutive times, the filter window length is automatically decreased by 1, with a minimum of 5.

[0042] On the one hand, the precise filtering of industrial electromagnetic interference (such as high-frequency noise generated by the frequency converter in the workshop) can be achieved by dynamically allocating the weighting coefficient, which improves the signal-to-noise ratio of the preliminary preprocessed data; on the other hand, without manual adjustment of the filter parameters, the window length can be adaptively changed according to the interference intensity, which not only adapts to the high-interference scenario in the heavy machinery workshop, but also avoids data lag caused by excessive filtering in the light-interference scenario, while realizing the dual effects of noise reduction and adaptive adaptation, solving the contradiction between incomplete noise reduction and data distortion in traditional fixed window filtering.

[0043] Further, the data verification and retransmission unit synchronously records the data error type and the corresponding collection time when sending the retransmission instruction, and establishes an error-environment association database, and calculates the data error probability in different environments through the following content:

[0044]

[0045] Among them:

[0046] is the data error probability under specific environmental conditions; ​​

[0047] The workshop environment temperature is valued in the range of -5℃-45℃, and is divided into intervals at 5℃ intervals;

[0048] The workshop electromagnetic interference intensity is valued in the range of 40dB-100dB, and is divided into intervals at 10dB intervals;

[0049] The total number of errors / missing data found per unit time in the temperature , interference intensity environment;

[0050] The total amount of data collected per unit time in the temperature , interference intensity environment;

[0051] The data verification retransmission unit is further configured to maintain only regular verification when , shorten the data retransmission response time when 1%< , and automatically send environment optimization suggestions to a preset address when > .

[0052] The above-mentioned mode not only realizes real-time retransmission of data errors, but also establishes an environment and data reliability correlation model through error probability calculation, early warns of high-risk environments, reduces subsequent data error frequency, provides precise guidance for workshop operation and maintenance, realizes dual functions of data retransmission and environment early warning, and avoids the limitations of traditional systems that can only passively handle errors and cannot trace the source.

[0053] Further, when calculating real-time parameters of the power grid, the dynamic model identification unit fuses clean data and workshop environment temperature data by using a multi-dimensional weighted correction formula, and the formula is:

[0054]

[0055] Among them:

[0056] is the corrected real-time resistance of the power grid line;

[0057] is the initial value of the line resistance calculated based on clean data by using the recursive least squares method;

[0058] is the temperature correction coefficient, which is determined based on the temperature resistance characteristics of copper transmission lines and conforms to the commonly used wire material parameters of industrial power grids;

[0059] The real-time temperature of the workshop environment is collected by the temperature sensor deployed in the workshop;

[0060] The reference temperature for the power grid design is 25°C;

[0061] The load correction coefficient is obtained based on the correlation characteristics experiment of the line load current and resistance;

[0062] The real-time load power is calculated based on the current and voltage data in the clean data, wherein, the real-time voltage is the real-time current is and the power factor is

[0063] The reference load power for the power grid design is 50% of the rated load power of the workshop;

[0064] The dynamic model identification unit is also used to calculate the deviation of the resistance value before and after correction. If the deviation is ≤5%, it is determined that the current power grid parameter is stable, and the correction frequency is maintained. If the deviation is >5%, it is determined that the power grid parameter fluctuates violently, and the correction frequency is increased to once every 10 seconds, and the fluctuation time and the corresponding load change are recorded synchronously for optimizing the subsequent correction coefficient.

[0065] The above method corrects the line resistance calculation error through temperature and load double dimensions, ensures the accuracy of the power grid mathematical model parameters, and determines the coefficients (a, b) in the formula based on the commonly used wire material and load characteristics in the industry, without the need to recalibrate for different workshops, which is suitable for multiple types of industrial scenes, realizes the dual effects of accurate parameter correction and scene general adaptation, and solves the parameter inaccuracy problem caused by traditional models relying only on temperature correction and ignoring load influence. 、

[0066] Further, the dynamic model identification unit calculates the real-time load rate as follows:

[0067]

[0068] wherein:

[0069] The real-time load rate at time t is used to determine the light load, heavy load or variable temperature sub-model in the scene sub-model library, wherein the light load is , the heavy load is , and the variable temperature is 30% < P < 70% and ;

[0070] ​​ is the number of data points in the sliding window;

[0071] is the current time, in s;

[0072] is the index of the data point in the window, taking values from 0 to m-1;

[0073] is the load data collection interval;

[0074] is the real-time load power at time t-p·Δt, calculated from the clean data;

[0075] is the rated load power of the workshop power supply loop, determined according to the workshop power distribution design parameters;

[0076] The dynamic model identification unit is further configured to switch the sub-model when the real-time load rate When switching from one scenario interval to another, the sub-model is not immediately switched, but the average trend of the load rate of the previous 5 times is calculated. If the trend consistently points to the new scenario, the sub-model is switched using a method of calling the new model weight 40% + old model weight 60% for the previous 3 times, and gradually transitioning to 100% new model weight thereafter. If the trend is repeated, the original model is maintained.

[0077] On the one hand, the sliding window smooths the load fluctuations (such as sudden load rise and fall caused by short-term start and stop of workshop equipment), reducing the fluctuation range of the load rate calculation result and ensuring the accuracy of the sub-model calling. On the other hand, the window length (m) and the collection interval (Δt) are adapted to the load change period of the industrial workshop (most equipment start-stop interval > 10s), which can accurately capture the long-term trend of the load, while realizing the dual effects of load rate smoothing calculation and sub-model accurate calling, avoiding the problems of frequent switching of sub-models and increased system energy consumption caused by traditional instantaneous load rate calculation.

[0078] Further, when the dynamic model identification unit iteratively updates the power grid mathematical model, the model parameters are adjusted using an error feedback iteration formula, which is:

[0079]

[0080] wherein:

[0081] is the power grid mathematical model parameter vector after the k+1th iteration;

[0082] is the model parameter vector after the kth iteration;

[0083] For the iteration step length coefficient, it is determined by experiment to ensure the balance between the iteration convergence speed and stability, and to avoid divergence caused by too large step length and slow convergence caused by too small step length;

[0084] The model error of the kth iteration, that is, the difference between the terminal voltage value calculated by the dynamic power grid model and the actual voltage value in the clean data;

[0085] The model loss function of the kth iteration, taking the value of , is used to measure the deviation between the model calculated value and the actual value;

[0086] The partial derivative of the loss function with respect to the model parameter vector, reflecting the influence degree of parameter change on the model error;

[0087] The dynamic model identification unit is also used to determine the model convergence when the model error of the last three iterations is , and extend the iteration period from 5 seconds to 15 seconds; if , it is determined that the model deviation is too large, and an emergency iteration is triggered immediately, the iteration interval is shortened to 1 second, and the historical correction parameter library storing the effective correction parameters in the last 24 hours is called to assist in adjusting ; if the iteration is still not reduced to below 0.5V after 10 times , send a model abnormal alarm information to the preset address.

[0088] On the one hand, the model parameter is adjusted in real time through error feedback, which shortens the model iteration convergence time and ensures that the model quickly adapts to the change of power grid parameters; on the other hand, the parameter adjustment direction is guided by the partial derivative of the loss function, avoiding blind iteration, so that the model error is finally stabilized within the expected range, realizing the dual effects of fast iteration convergence and minimum model error, and solving the problems of slow convergence and easy to fall into local optimum of traditional iteration algorithm.

[0089] Further, the adaptive control unit calculates the deviation value as follows:

[0090]

[0091] Wherein:

[0092] The weighted deviation value of the terminal voltage;

[0093] The static deviation weight coefficient, taking the value of 0.7;

[0094] The theoretical value of the terminal voltage calculated by the dynamic power grid model;

[0095] The actual voltage value in the clean data output by the data check and retransmission unit is obtained;

[0096] The dynamic deviation weight coefficient is 0.3;

[0097] The change rate of the voltage theoretical value is calculated by the difference value between adjacent time points and a 1s time interval;

[0098] The change rate of the voltage actual value is calculated by the difference value between adjacent time points and a 1s time interval; Meanwhile, the adaptive control unit is preset with a deviation threshold grading mechanism, when

[0099] , it is determined that the voltage deviation is within the allowable range, and the control strategy parameter is not adjusted; when , a mild adjustment is triggered, and only the voltage reduction amplitude is fine-tuned; when , a deep adjustment is triggered, and the voltage reduction amplitude and the control frequency are adjusted synchronously. On the one hand, the design fuses static and dynamic deviations with weighting, which not only ensures the voltage amplitude accuracy, but also avoids the equipment impact caused by too fast voltage change; on the other hand, the threshold grading adjustment avoids the control shock caused by the traditional “one-size-fits-all” adjustment, adapts to different load sensitivity requirements of the workshop (such as mild adjustment for precision machine tools and deep adjustment for ordinary motors), realizes the three effects of “static accuracy guarantee + dynamic fluctuation suppression + load adaptive adjustment”, and reduces the voltage control deviation rate from the traditional single static calculation of 5%-8% to below 2%.

[0100] Further, the adaptive control unit calculates the load change trend of the future preset time period as follows:

[0101]

[0102] Among them:

[0103]

[0104] The predicted load power at t+Δt is obtained;

[0105] The actual load power at t is obtained from the clean data;

[0106] The trend order is 1 or 2;

[0107] The first-order trend coefficient is calculated by the linear fitting slope of the load power in the previous 10s;

[0108] ​​ is the coefficient of 2nd order trend, calculated by the quadratic fitting curvature of load power in the first 10s;

[0109] is the prediction time interval, fixed at 5s;

[0110] is the prediction error correction term, taking the average of the previous 5 prediction errors;

[0111] Before generating the pre-control instruction, the adaptive control unit also sets the prediction reliability determination logic, calculates the predicted load power deviation fluctuation amplitude from the previous 3 predicted values, if the fluctuation amplitude ≤ 5%, it is determined that the prediction is reliable, and the pre-control instruction is generated directly based on if the predicted load increases, the step-down amplitude is reduced in advance; if the fluctuation amplitude > 5%, it is determined that there is random load interference, and the current fluctuation frequency in the clean data is used to assist in judgment and correction then generate the instruction.

[0112] This design, on the one hand, accurately captures the linear and nonlinear changes of the workshop load through the fusion of 1st order + 2nd order trend, improves the prediction accuracy; on the other hand, through reliability determination and auxiliary judgment and correction based on current fluctuation frequency in clean data, it avoids prediction failure caused by random load, and realizes the triple effects of complex trend capture, error dynamic compensation and prediction reliability guarantee, so that the pre-control instruction is adapted to the load change in advance, reduces the adjustment amplitude of the subsequent control strategy, and reduces the energy loss of frequent regulation of equipment.

[0113] Further, the power quality management unit calculates the three-phase voltage unbalance degree by the following content:

[0114]

[0115] Wherein:

[0116] is the weighted three-phase voltage unbalance degree;

[0117] are the actual voltage values of phase A, B and C in the clean data, respectively;

[0118] is the average value of three-phase voltage,

[0119] are the actual load powers of phase A, B and C, respectively, all obtained from the clean data;

[0120] is the total three-phase load power, ​​

[0121] When , the power quality treatment unit triggers a three-phase voltage regulation mechanism, and sets a split-phase regulation logic, first identifies the high-load phase, and fine-tunes the voltage regulation module of the high-load phase first, and after adjustment, re-collects clean data for calculation after 2s ; if drops to within 2%, stop adjusting; if it still exceeds the standard, adjust the low-load phase again, and the adjustment range is 1 / 2 of the high-load phase, to avoid excessive adjustment of the low-load phase leading to new imbalance.

[0122] This design, on the one hand, by integrating the proportion of each phase load power, prioritizes the high-load phase voltage deviation, solving the problem of excessive regulation of light-load phase caused by traditional calculation of voltage amplitude difference and ignoring load distribution; on the other hand, through the split-phase regulation logic, precise voltage control is achieved, and the three-phase voltage unbalance degree can be stabilized within the expected range after adjustment, ensuring the stability of precision equipment power supply, reducing motor copper loss caused by three-phase imbalance, and achieving the three effects of load adaptive regulation, equipment loss reduction, and regulation accuracy guarantee. BRIEF DESCRIPTION OF DRAWINGS

[0123] Figure 1 is a logic block diagram of an embodiment of an intelligent power grid energy-saving control system. DETAILED DESCRIPTION

[0124] The following will be further described in detail through specific embodiments:

[0125] An intelligent power grid energy-saving control system includes an end sensing layer, a data processing layer, a control execution layer, and a double-link communication layer (as shown in Figure 1 );

[0126] The end sensing layer includes a distributed multi-parameter sensing unit, which is arranged in a workshop power supply circuit away from the power distribution room. The distributed multi-parameter sensing unit includes a voltage sensor, a current sensor, and a harmonic detection sensor. The voltage sensor and the current sensor are used to synchronously collect real-time voltage data and current data at the end. The harmonic detection sensor is used to collect harmonic component data of the end power grid. At the same time, the distributed multi-parameter sensing unit is internally provided with an edge preprocessing module. The edge preprocessing module performs preliminary filtering and noise reduction on the collected voltage, current, and harmonic data to obtain preliminary preprocessing data, and removes abnormal data caused by industrial electromagnetic interference.

[0127] The data processing layer comprises a data verification and retransmission unit and a dynamic model identification unit; the data verification and retransmission unit is used for verifying the preliminary pretreatment data of the edge pretreatment module in real time according to the CRC cyclic redundancy check algorithm, and if data errors or loss are found in the verification, a retransmission instruction is immediately sent to the distributed multi-parameter sensing unit to ensure data integrity; if no data errors or loss are found, clean data is output; the dynamic model identification unit is used for calculating real-time parameters of the power grid by using the recursive least squares method and the extended Kalman filter algorithm according to the clean data, correcting the real-time parameters to obtain corrected parameters by combining the acquired workshop ambient temperature data, and continuously iteratively updating the mathematical model of the power grid based on the corrected parameters, while establishing a sub-model library of light load, heavy load and variable temperature scenes, calculating the real-time load rate according to the clean data, calling the matched sub-model from the sub-model library according to the real-time load rate, and obtaining the dynamic power grid model adapted to the actual working condition;

[0128] The control execution layer comprises an adaptive control unit and a power quality management unit; the adaptive control unit is used for receiving the dynamic power grid model output by the dynamic model identification unit, comparing the end voltage theoretical value calculated by the dynamic power grid model with the clean data output by the data verification and retransmission unit, calculating a deviation value, and automatically adjusting the control strategy parameters if the deviation value is greater than a preset threshold; at the same time, the adaptive control unit is preset with a model prediction algorithm, predicts the future end load change trend of a preset time period according to the clean data, generates a pre-control instruction, fuses the adjusted control strategy parameters and the pre-control instruction to obtain a cooperative control instruction; the power quality management unit is used for receiving the harmonic component data collected by the distributed multi-parameter sensing unit and the cooperative control instruction output by the adaptive control unit, determining the harmonic suppression demand according to the harmonic component data, outputting a harmonic suppression signal to the filtering device of the workshop power supply circuit in combination with the cooperative control instruction, and simultaneously monitoring the clean data in real time to judge the three-phase voltage balance degree, and when the three-phase voltage imbalance degree exceeds a limited value, triggering a three-phase voltage regulation mechanism to generate a three-phase voltage regulation signal;

[0129] The dual-link communication layer comprises a main communication link and a backup communication link, the main communication link adopts optical fiber communication, and the backup communication link adopts a 5G industrial private network; the dual-link communication layer is used for receiving the data transmitted between the units of the end perception layer, the data processing layer and the control execution layer in real time, while monitoring the signal strength and error code rate of the main communication link, and when the error code rate of the main communication link is greater than a first error code preset value or the signal strength is lower than a first strength preset value, automatically switching the data transmission link to the backup communication link and sending link switching alarm information to a preset address.

[0130] Specific use: this embodiment takes the intelligent power grid energy-saving control system of an automobile parts manufacturing enterprise as the application object, the enterprise has two core production workshops (machining workshop, injection molding workshop) away from the power distribution room, the workshop power supply loop voltage level is 380V, the main load is numerical control machine tool (power 11kW-37kW), injection molding machine (power 30kW-75kW), there are typical industrial characteristics such as industrial electromagnetic interference (interference intensity 55dB-90dB), periodic load fluctuation (light load period load rate 25%, heavy load period load rate 85%), environmental temperature fluctuation (winter 5℃-15℃, summer 28℃-40℃), the present application realizes whole-process energy-saving control through end sensing layer-data processing layer-control execution layer-double-link communication layer.

[0131] One set of distributed multi-parameter sensing unit is arranged on each of the three main shaft power supply loops in the machining workshop and the two forming machine power supply loops in the injection molding workshop, each set of unit contains:

[0132] Voltage sensor (selected type: VSM025A, measurement range 0-500V, accuracy ±0.3%), used for collecting real-time voltage data at the end ;

[0133] Current sensor (selected type: CSM050A, measurement range 0-100A, accuracy ±0.2%), used for collecting real-time current data at the end ;

[0134] Harmonic detection sensor (selected type: HSM010, detection range 3-21 harmonics, accuracy ±0.8%), used for collecting end power grid harmonic component data .

[0135] The edge preprocessing module is preprocessed by the following contents:

[0136]

[0137] Among them,

[0138] : The preliminary pretreatment data output after filtering (including voltage, current, harmonic component data), the data source is the original data collected by the sensor after filtering calculation result;

[0139] : Current data collection time (value 1, 2,..., 30, corresponding to 30 data points collected by the sensor every second), generated by the built-in clock module of the sensor;

[0140] : Filter window length (value 5-10), adaptively adjusted according to the electromagnetic interference intensity of the workshop, and the interference intensity data is collected by the interference detection module matched with the sensor;

[0141] : the weighting coefficient of the first original data (satisfying ), calculated by the formula , and the character is a natural constant (with a value of 2.718);

[0142] : the original data (voltage, current or harmonic component data) collected at the first moment, directly derived from the real-time collection results of each sensor.

[0143] If the electromagnetic interference intensity detection of the machining workshop is 82dB, set to 10; if the original data fluctuation variance is greater than the preset threshold value (voltage data threshold value 0.5V) for 3 consecutive times, increase by 2 (the maximum does not exceed 10). The preliminary pretreatment data obtained according to the above method has an improved signal-to-noise ratio compared with the original data, and abnormal data caused by industrial electromagnetic interference (such as voltage spikes and current clutter) is effectively filtered out.

[0144] The main communication link adopts a single-mode optical fiber (model: GJFJV-4B1, transmission rate 100Mbps, transmission distance 1.5km), connecting the end sensing layer and the data processing layer, and is used for transmitting preliminary pretreatment data and clean data; the standby communication link adopts a 5G industrial private network (frequency band 2.6GHz, bandwidth 50Mbps, time delay <20ms), which works in parallel with the main link, and is realized by an industrial-grade 5G module (model: MG300-5G) provided by the operator.

[0145] The dual-link communication layer monitors the signal strength and the bit error rate of the main link in real time, and automatically switches to the standby link when the bit error rate (first bit error preset value) or the signal strength (first strength preset value) of the main link is greater than the preset value. The bit error rate and the signal strength are collected in real time by a link monitoring module (integrated in the communication gateway, model: IGW-500); the preset values , are set according to the reliability requirements of industrial power grid data transmission and stored in the system configuration file.

[0146] If the electromagnetic interference of the injection molding machine in the injection molding workshop causes the bit error rate , automatically switch to the standby 5G link, and send alarm information (such as the main link error rate exceeding the standard, and the 5G standby link has been switched) to the preset address of the operation and maintenance terminal. If the link switching is completed according to the above logic, the data transmission interruption time can be controlled within 50ms, ensuring the continuity of data transmission at all levels.

[0147] The data verification and retransmission unit verifies the preliminary preprocessed data based on the CRC cyclic redundancy check algorithm, synchronously records the data error type (such as data bit loss, check code mismatch) and the collection time, establishes an error-environment association database, and calculates the data error probability through the following content:

[0148]

[0149] Among them,

[0150] : Data error probability in a specific environment, data source is the ratio of error data volume to total data volume;

[0151] : Workshop environment temperature (value range-5℃-45℃, interval is divided by 5℃), collected by temperature sensor (model: TSM010, accuracy ±0.5℃) deployed in the workshop;

[0152] : Workshop electromagnetic interference intensity (value range 40dB-100dB, interval is divided by 10dB), collected by interference detection module of distributed multi-parameter sensing unit;

[0153] : Total number of error / missing data in a unit of time (such as 1 hour), generated by data verification and retransmission unit statistics;

[0154] : Total data volume collected in a unit of time (such as 1 hour), calculated by sensor collection frequency (30 data points per second).

[0155] If the workshop temperature is ℃ , the interference intensity is , the total number of errors is , If the data error probability is according to the above formula, keep regular verification; if , send environment optimization suggestions (such as suggesting to start the workshop cooling equipment to reduce the environment temperature to below 30℃) to the operation and maintenance terminal.

[0156] The dynamic model identification unit calculates the real-time resistance of the power grid line as follows:

[0157]

[0158] Wherein:

[0159] : The corrected real-time resistance of the power grid line, the data source is the result of the initial resistance value after temperature and load correction;

[0160] : The initial value of the line resistance calculated based on clean data, calculated by recursive least squares method, the clean data is derived from the output result of the data verification and supplement unit;

[0161] : Temperature correction coefficient (the value of this embodiment is 0.004 / ℃), determined based on the temperature-resistance characteristic experiment of commonly used copper conductors (workshop power supply line material), of course, it can be set to other values according to actual situation;

[0162] : Real-time temperature of the workshop environment, collected by the temperature sensor; : Power grid design reference temperature (value 25℃), set according to the industrial power grid standard;

[0163] : Load correction coefficient (value 0.002 / kW), derived based on the correlation characteristic experiment of line load current and resistance;

[0164] : Real-time load power (P) ), , derived from clean data, (power factor) collected by the power factor detection module;

[0165] : Reference load power (value 50% of the workshop rated load), set according to the workshop power distribution design parameters (rated load 150kW).

[0166] If the clean data , , , ℃ If the corrected line resistance is obtained according to the above content, compared with the uncorrected , it is more consistent with the actual power grid parameters, ensuring the model precision.

[0167] The following content is used to calculate the real-time load rate:

[0168]

[0169] Wherein:

[0170] Real-time load rate at time t, data source is the ratio of real-time load power to rated load in the sliding window;

[0171] Number of data points in the sliding window (value 12), corresponding to load data in 1 minute (1 data point collected every 5 seconds), set by the sampling period;

[0172] Current time (unit s), generated by the system clock module; Data point index in the window (value 0 to m-1);

[0173] Load data collection interval (value 5s), set according to the load variation frequency;

[0174] Real-time load power at time t-p·Δt, derived from clean data;

[0175] Rated load power of workshop power supply loop (value 150kW), determined according to the power distribution design parameters.

[0176] If the average of the 12 data points in the sliding window is 105kW. If the real-time load rate is obtained according to the above content , it is determined that it is a heavy load condition, and the heavy load sub-model in the scene sub-model library is called.

[0177] The model parameters are adjusted as follows:

[0178]

[0179] Among them:

[0180] , Model parameter vector after the k+1th and kth iteration (including line resistance, reactance, and load admittance), data source is the parameter iteration calculation result;

[0181] Iteration step size coefficient (value 0.05), determined by experiment (balance convergence speed and stability);

[0182] Model error (e) of the kth iteration, is the model calculated voltage, is the actual voltage in the clean data;

[0183] Model loss function​ ), for measuring the model bias;

[0184] : the partial derivative of the loss function with respect to the parameter vector, reflecting the degree of influence of parameter changes on errors, calculated by calculus.

[0185] If the kth iteration of , If the k+1th iteration of , the model error can be gradually converged to within 0.2V, ensuring that the model is adapted to the actual power grid.

[0186] The adaptive regulation unit calculates the voltage deviation as follows:

[0187]

[0188] Among them:

[0189] : the terminal voltage weighted deviation value, the data source is the weighted sum result of static deviation and dynamic deviation;

[0190] : static deviation weight coefficient (value 0.7), : dynamic deviation weight coefficient (value 0.3), set according to the demand of industrial equipment for voltage static precision and dynamic stability;

[0191] : the voltage theoretical value calculated by the dynamic power grid model, : the actual voltage in the clean data;

[0192] , : the change rate of voltage theoretical value and actual value, calculated by the voltage difference between adjacent time and 1s time interval.

[0193] If , , the voltage change rates are 0.8V / s and 0.6V / s respectively. If according to the above content, it is obtained , at this time (preset threshold), deep adjustment (synchronous adjustment of step-down amplitude and regulation frequency) is triggered.

[0194] The load trend is predicted as follows:

[0195]

[0196] Among them:

[0197] : Predicted load power at t+Δt, data source is the result of actual load power plus trend item and error correction item;

[0198] : Actual load power at t, from clean data;

[0199] : Trend order (value 1, 2), 1 order reflects linear change, 2 order reflects nonlinear change;

[0200] : 1st order trend coefficient, calculated by linear fitting slope of previous 10s load power; : 2nd order trend coefficient, calculated by quadratic fitting curvature of previous 10s load power;

[0201] : Prediction time interval (value 5s), set according to load response speed;

[0202] : Prediction error correction item, value is the average of previous 5 prediction errors, calculated from historical prediction data.

[0203] If , , 2 , If the above content is obtained , at this time the prediction is reliable (fluctuation amplitude ≤5%), then generate pre-control instruction based on this value.

[0204] The power quality management unit calculates the three-phase voltage unbalance degree by the following content:

[0205]

[0206] Among them:

[0207] : Weighted three-phase voltage unbalance degree, data source is the calculation result of voltage deviation and load proportion;

[0208] : Actual voltage of A, B, C phase, from clean data; : Three-phase voltage average value ( );

[0209] : Actual load power of A, B, C phase, from clean data; : Three-phase total load power ( ).

[0210] If , , , , , If the above conditions are met , at this time (the limit value), the high load phase (A phase, 38.9% of the proportion) is fine-tuned first, and after adjustment it can be reduced to within 2%, avoiding equipment failure caused by three-phase imbalance.

[0211] The above is only an embodiment of the application, and the application is not limited to this embodiment. The specific structure and characteristics of the scheme and other common knowledge are not described in detail here. The ordinary skilled person in the art knows all the ordinary technical knowledge in the field of the application before the filing date or the priority date, can know all the prior art in the field, and has the ability to apply conventional experimental means before that date. The ordinary skilled person in the art can improve and implement the scheme under the guidance of this application, combining their own ability. Some typical known structures or known methods should not be an obstacle for the ordinary skilled person in the art to implement this application. It should be noted that for those skilled in the art, without departing from the structure of the application, a number of modifications and improvements can be made, which should also be considered within the scope of protection of the application. These will not affect the effect and practicality of the application. The scope of protection claimed in this application should be based on the content of the claims, and the specific implementation in the specification can be used to explain the content of the claims.

Claims

1. A smart grid energy-saving control system, characterized in that, It includes the end-sensing layer, the data processing layer, and the control execution layer; The terminal sensing layer includes a distributed multi-parameter sensing unit, which is deployed in the workshop power supply circuit far from the power distribution room. The distributed multi-parameter sensing unit includes a voltage sensor, a current sensor, and a harmonic detection sensor. The voltage sensor and current sensor are used to synchronously collect real-time voltage and current data at the terminal, and the harmonic detection sensor is used to collect harmonic component data of the terminal power grid. At the same time, the distributed multi-parameter sensing unit has a built-in edge preprocessing module. The edge preprocessing module performs preliminary filtering and noise reduction on the collected voltage, current, and harmonic data to obtain preliminary preprocessed data, removing abnormal data caused by industrial electromagnetic interference. The data processing layer includes a data verification and retransmission unit and a dynamic model identification unit. The data verification and retransmission unit is used to perform real-time verification on the preliminary preprocessed data of the edge preprocessing module according to the CRC cyclic redundancy check algorithm. If the verification finds data errors or loss, it immediately sends a retransmission command to the distributed multi-parameter sensing unit to ensure data integrity. If no data errors or loss are found, output clean data; The dynamic model identification unit is used to calculate the real-time parameters of the power grid based on the cleaning data using the recursive least squares method and the extended Kalman filter algorithm. It then corrects the real-time parameters by combining the acquired workshop ambient temperature data to obtain corrected parameters. Based on the corrected parameters, it continuously iterates and updates the power grid mathematical model. At the same time, it establishes a sub-model library for light load, heavy load, and variable temperature scenarios. It calculates the real-time load rate based on the cleaning data and calls the matching sub-model from the sub-model library based on the real-time load rate to obtain a dynamic power grid model adapted to the actual working conditions. The control execution layer includes an adaptive regulation unit and a power quality management unit; The adaptive control unit receives the dynamic power grid model output by the dynamic model identification unit, compares the theoretical value of the terminal voltage calculated by the dynamic power grid model with the clean data output by the data verification and supplementary transmission unit, calculates the deviation value, and automatically adjusts the control strategy parameters if the deviation value is greater than a preset threshold. At the same time, the adaptive control unit has a preset model prediction algorithm, which predicts the terminal load change trend in the future preset time period based on the clean data, generates a pre-control instruction, and merges the adjusted control strategy parameters with the pre-control instruction to obtain a coordinated control instruction. The power quality management unit is used to receive harmonic component data collected by the distributed multi-parameter sensing unit and the collaborative control command output by the adaptive control unit. Based on the harmonic component data, it determines the harmonic suppression requirement, and outputs the harmonic suppression signal to the filter device of the workshop power supply circuit in combination with the collaborative control command. At the same time, it monitors the cleanliness data in real time to judge the three-phase voltage balance. When the three-phase voltage imbalance exceeds the limit value, it triggers the three-phase voltage regulation mechanism and generates a three-phase voltage regulation signal.

2. The smart grid energy-saving control system according to claim 1, characterized in that, It also includes a dual-link communication layer, which comprises a primary communication link and a backup communication link. The primary communication link uses optical fiber communication, and the backup communication link uses a 5G industrial private network. The dual-link communication layer is used to receive data transmitted between the units of the terminal sensing layer, data processing layer, and control execution layer in real time, and to monitor the signal strength and bit error rate of the primary communication link. When the bit error rate of the primary communication link is greater than a first preset bit error rate or the signal strength is lower than a first preset strength, the data transmission link is automatically switched to the backup communication link, and a link switching alarm message is sent to a preset address.

3. The smart grid energy-saving control system according to claim 2, characterized in that, The edge preprocessing module performs preliminary filtering and noise reduction using the following methods: ； in: This is the initial preprocessed data output after filtering; This represents the current data acquisition time, with a positive integer value, corresponding to the first data acquisition time per second collected by the sensor. One data point; The value is the length of the filtering window, ranging from 5 to 10. It is adaptively adjusted according to the intensity of industrial electromagnetic interference in the workshop. When the interference intensity is >80dB, it is 10, and when the interference intensity is <50dB, it is 5. For the first The weighting coefficients of the original data satisfy the following conditions: ,and That is, the data in the center of the window has the highest weight, and the data at the edge has decreasing weight, so as to ensure that the filtered data retains the original trend of change. For the first Raw data collected in real time; Meanwhile, the edge preprocessing module is also used to determine the electromagnetic interference intensity by collecting the fluctuation variance of the raw data. If the fluctuation variance is detected to be greater than a preset threshold for three consecutive times, the filter window length is automatically adjusted. Increase by 2, with a maximum of 10; if the variance of fluctuation is ≤ preset threshold for 5 consecutive times, then automatically... Reduce by 1, with a minimum of 5.

4. The smart grid energy-saving control system according to claim 3, characterized in that, When sending a retransmission command, the data verification and retransmission unit synchronously records the data error type and corresponding acquisition time, and establishes an error-environment association database. It then calculates the data error probability under different environments using the following methods: ； in: The probability of data error under specific environmental conditions; The ambient temperature of the workshop is taken as -5℃ to 45℃, divided into intervals of 5℃. The electromagnetic interference intensity in the workshop is measured in the range of 40dB-100dB, divided into intervals of 10dB. For temperature Interference intensity Under the given environment, the total number of erroneous / lost data detected by verification per unit of time; For temperature Interference intensity The total amount of data collected per unit time under the given environment; The data verification and retransmission unit is also used when When only regular checks are performed, if 1% < In this way, the data retransmission response time can be shortened; When > At that time, it automatically sends environment optimization suggestions to the preset address.

5. The smart grid energy-saving control system according to claim 4, characterized in that, When calculating real-time power grid parameters, the dynamic model identification unit uses a multi-dimensional weighted correction formula to fuse cleaning data and workshop ambient temperature data. The formula is as follows: ； in: The corrected real-time resistance of the power grid line; The initial value of the line resistance is calculated using the recursive least squares method based on clean data; The temperature correction factor is determined based on the temperature resistance characteristics of copper transmission lines and conforms to the material parameters of commonly used conductors in industrial power grids. The real-time temperature of the workshop environment is collected by temperature sensors deployed in the workshop. The reference temperature for power grid design is set at 25℃. This is the load correction factor, derived from experiments based on the correlation characteristics of line load current and resistance; This is the real-time load power calculated based on current and voltage data from the clean data. ,in, For real-time voltage, For real-time current, Power factor; This is the reference load power for power grid design, and is taken as 50% of the workshop's rated load power. The dynamic model identification unit is also used to calculate the deviation of the resistance value before and after correction. If the deviation is ≤5%, it is determined that the current power grid parameters are stable and the correction frequency is maintained. If the deviation is >5%, it is determined that the power grid parameters fluctuate violently and the correction frequency is increased to once every 10 seconds. The fluctuation time and the corresponding load change are recorded simultaneously for optimization of subsequent correction coefficients.

6. The smart grid energy-saving control system according to claim 5, characterized in that, The dynamic model identification unit calculates the real-time load rate using the following methods: ； in: The real-time load rate at time t is used to determine whether to call the light-load, heavy-load, or variable-temperature sub-model in the scenario-specific sub-model library. Here, light-load is... Heavy load is Temperature variation is 30% < <70% and ℃ ; This represents the number of data points within the sliding window. The current time is expressed in seconds (s). This is the index of the data points within the window, with values ​​ranging from 0 to m-1; This refers to the load data acquisition interval; The real-time load power at time tp·Δt is calculated from clean data; The rated load power of the power supply circuit for the workshop is determined based on the workshop power distribution design parameters; The dynamic model identification unit is also used to identify the real-time load rate. When switching from one scene interval to another, the sub-model is not switched immediately. Instead, the average trend of the load rate of the previous 5 times is calculated. If the trend continues to point to the new scene, the sub-model is switched by calling the new model with a weight of 40% + the old model with a weight of 60% for the first 3 times, and then gradually transitioning to 100% new model weight. If the trend is repeated, the original model is maintained.

7. A smart grid energy-saving control system according to claim 6, characterized in that, When the dynamic model identification unit iteratively updates the power grid mathematical model, it uses an error feedback iterative formula to adjust the model parameters. The formula is as follows: ； in: This is the parameter vector of the power grid mathematical model after the (k+1)th iteration; This is the model parameter vector after the k-th iteration; The iteration step size coefficient is determined experimentally to ensure a balance between iteration convergence speed and stability, avoiding divergence caused by an excessively large step size and slow convergence caused by an excessively small step size. The model error for the kth iteration is the difference between the terminal voltage value calculated by the dynamic power grid model and the actual voltage value in the clean data. Let be the model loss function for the k-th iteration, and let its value be . This is used to measure the deviation between the model's calculated values ​​and the actual values; It is the partial derivative of the loss function with respect to the model parameter vector, reflecting the degree of influence of parameter changes on model error; The dynamic model identification unit is also used to detect model errors after three consecutive iterations. If the model converges, the iteration period is increased from 5 seconds to 15 seconds; if If the model is determined to have excessive deviation, an emergency iteration is immediately triggered, with the iteration interval shortened to 1 second. Simultaneously, a historical correction parameter library storing valid correction parameters from the past 24 hours is invoked to assist in adjustments. If after 10 iterations If the voltage still does not drop below 0.5V, a model anomaly alarm message is sent to the preset address.

8. The smart grid energy-saving control system according to claim 7, characterized in that, The adaptive control unit calculates the deviation value using the following methods: ； in: This is the weighted deviation value of the terminal voltage; This is the static deviation weighting coefficient, with a value of 0.7; The theoretical value of the terminal voltage calculated for a dynamic power grid model; To verify the actual voltage value in the clean data output by the data verification and supplementary transmission unit; This is the dynamic deviation weighting coefficient, with a value of 0.3; The rate of change of the theoretical voltage value, expressed as a function of adjacent time intervals. The difference is calculated using a 1-second time interval. The rate of change of the actual voltage value, expressed over adjacent time intervals. The difference is calculated using a 1-second time interval. Meanwhile, the adaptive control unit has a preset deviation threshold grading mechanism, when When the voltage deviation is within the allowable range, the control strategy parameters are not adjusted; when... When this is triggered, a slight adjustment is made, only slightly modifying the voltage drop; when At that time, a deep adjustment is triggered, and the voltage reduction amplitude and control frequency are adjusted simultaneously.

9. A smart grid energy-saving control system according to claim 8, characterized in that, The adaptive control unit calculates the load change trend over a future preset time period using the following methods: ； in: Let be the predicted load power at time t+Δt; The actual load power at time t is obtained from the clean data; This represents the trend order, with values ​​of 1 and 2. It is a first-order trend coefficient, calculated by linearly fitting the slope of the load power in the first 10 seconds; It is a second-order trend coefficient, calculated by quadratic fitting curvature of the load power in the first 10 seconds; The prediction time interval is fixed at 5 seconds. This is the prediction error correction term, and its value is the average of the previous 5 prediction errors; Before generating the pre-regulation command, the adaptive regulation unit also sets up prediction reliability judgment logic to calculate the predicted load power. If the deviation from the previous three predicted values ​​is ≤5%, the prediction is considered reliable and can be directly based on... Generate pre-regulation instructions: if the predicted load increase, reduce the voltage reduction amount in advance; if the fluctuation range is greater than 5%, it is determined that there is random load interference, and the current fluctuation frequency in the clean data is used as an auxiliary judgment for correction. Then generate the instructions.

10. A smart grid energy-saving control system according to claim 9, characterized in that, The power quality management unit calculates the three-phase voltage imbalance using the following methods: ； in: Weighted three-phase voltage imbalance; These are the actual voltage values ​​of phases A, B, and C in the cleaning data; This is the average value of the three-phase voltage. ; The actual load power for phases A, B, and C are respectively obtained from clean data; This is the total three-phase load power. ; when At that time, the power quality management unit triggers the three-phase voltage regulation mechanism and sets up phase-by-phase regulation logic. First, it identifies the high-load phase and prioritizes fine-tuning the voltage regulation module of the high-load phase. After adjustment, it re-collects clean data and calculates the results after an interval of 2 seconds. ;like Once the load decreases to below 2%, stop adjusting. If it still exceeds the limit, then perform auxiliary adjustments on the low-load phase, with the adjustment range being half that of the high-load phase, to avoid excessive adjustment of the low-load phase leading to new imbalances.

Images