Control method for low-voltage intelligent reactive compensation system of submerged arc furnace

Abstract

The invention relates to the technical field of power electronics, in particular to a control method for a low-voltage intelligent reactive compensation system of a submerged arc furnace. Comprising the following steps: acquiring three-phase voltage, current, power factor, harmonic content and temperature parameters of the low-voltage side of the submerged arc furnace through a sensor network; the load fluctuation of the submerged arc furnace in the future 1-5 minutes is predicted based on a deep learning algorithm, and a compensation demand pre-judgment value is generated; according to a load prediction result and a real-time working condition, performing dynamic switching from three modes of fixed compensation, grouped switching and continuous adjustment; harmonic component dynamic tracking compensation is carried out by combining an active power filter (APF) and a static var generator (SVG); the system overvoltage, overcurrent and temperature parameters are monitored in real time, and a grading protection mechanism is triggered and fed back to the compensation strategy adjusting module. The invention provides a control method for a low-voltage intelligent reactive compensation system of a submerged arc furnace so as to improve the response speed of the system, prolong the service life of equipment and guarantee the electric energy quality of a power grid.

Description

Technical Field

[0001] The present invention relates to the field of power electronics technology, and particularly to a control method for a low-voltage intelligent reactive power compensation system of a submerged arc furnace. Background Art

[0002] As a key device in the metallurgical industry, a submerged arc furnace generates a large amount of inductive reactive power during its operation, resulting in a significant reduction in the power factor of the power grid (usually below 0.7), and at the same time, accompanied by serious harmonic pollution (the total harmonic distortion rate THD can reach 15% - 30%). Traditional low-voltage reactive power compensation systems mostly adopt fixed capacitor banks or grouped switching strategies. However, during the smelting process of the submerged arc furnace, due to factors such as changes in raw material composition, frequent electrode lifting, and switching of smelting stages (such as the melting period and the refining period), the load fluctuates violently (the volatility can reach more than ±30%), and the existing compensation devices have problems such as a lag in dynamic response (in seconds), poor harmonic suppression effect, and serious equipment losses.

[0003] Currently, the solutions commonly adopted in the industry still have the following technical bottlenecks: (1) Traditional compensation strategies rely on static analysis of historical data and lack the fusion prediction of multi-dimensional parameters such as electrode position and raw material composition, resulting in a compensation amount matching error of more than 20% when the load suddenly changes; (2) The switching logic between the fixed compensation and grouped switching modes is single and cannot adapt to millisecond-level load fluctuations. Frequent switching causes the daily average action times of the capacitor bank to exceed the limit (>50 times / group), and the equipment life is shortened by more than 30%; (3) The traditional overvoltage and overcurrent protection thresholds are fixed and cannot be dynamically adjusted according to the real-time working conditions. The fault response time exceeds 100 ms, which is likely to cause accidents such as the explosion of the capacitor bank or the burning of the IGBT module. Summary of the Invention

[0004] The present invention provides a control method for low-voltage reactive power compensation of a submerged arc furnace that integrates dynamic prediction, multi-modal compensation, harmonic collaborative governance, and intelligent safety protection to improve the system response speed, extend the equipment life, and ensure the power quality of the power grid.

[0005] The technical solution adopted by the present invention is as follows: A control method for a low-voltage intelligent reactive power compensation system of a submerged arc furnace includes the following steps:

[0006] Step 1, real-time data acquisition: Collect three-phase voltage, current, power factor, harmonic content, and temperature parameters on the low-voltage side of the submerged arc furnace through a sensor network;

[0007] Step 2, dynamic load prediction: Based on a deep learning algorithm, predict the load fluctuation of the submerged arc furnace in the next 1 - 5 minutes to generate a predicted value of the compensation demand;

[0008] Step 3, multi-modal compensation strategy matching: Dynamically switch among the fixed compensation, grouped switching, and continuous adjustment modes according to the load prediction result and the real-time working conditions;

[0009] Step 4, harmonic collaborative suppression: Combine the active power filter (APF) with the static var generator (SVG) to perform dynamic tracking compensation for harmonic components;

[0010] Step 5, safety closed-loop control: Real-time monitor the system overvoltage, overcurrent and temperature parameters, trigger the hierarchical protection mechanism and feedback it to the compensation strategy adjustment module.

[0011] As a further improvement of the present invention, in Step 1, a multi-source data fusion technology is adopted. A Rogowski coil is added to the sensor network to collect current mutation signals, and the displacement sensor data of the electrode lifting mechanism is synchronized. The data sampling frequency is dynamically adjusted according to the smelting stage: 10 kHz sampling is used during the melting period, and the sampling is switched to 5 kHz during the refining period.

[0012] As a further improvement of the present invention, the dynamic load prediction adopts an improved LSTM neural network model. The input parameters include the historical load curve, electrode position signal, raw material composition data and smelting stage identifier. The model shares the training weights in multiple submerged arc furnace scenarios through transfer learning.

[0013] As a further improvement of the present invention, the improved LSTM neural network model includes an attention mechanism module. A working condition feature weighting unit is set in the decoder layer, and a dynamic weight coefficient of 0.3 - 0.5 is assigned to the FeO content index in the raw material composition data, and a priority coefficient of 0.6 - 0.8 is assigned to the submerged arc depth parameter in the electrode position signal.

[0014] As a further improvement of the present invention, in the multi-modal compensation strategy matching, when the load volatility exceeds ±15% and the duration > 30 seconds, the continuous adjustment mode is started, and the millisecond-level dynamic response of reactive power is realized through the IGBT converter.

[0015] As a further improvement of the present invention, when the grouped switching mode is implemented, a capacitor bank cyclic switching sequence is set. When it is detected that the daily switching times of a single capacitor bank > 50 times, this group is automatically marked as a standby unit, and the adjacent capacitor bank is activated to form a new switching combination.

[0016] As a further improvement of the present invention, the harmonic collaborative suppression adopts an adaptive harmonic separation algorithm, specifically including: analyzing the real-time harmonic spectrum through FFT to identify the characteristic harmonic components of 2 - 25 times; assigning the compensation tasks of the high-order harmonics of 3 times and above to the active power filter; assigning the compensation tasks of the dominant harmonics of 2 - 5 times to the static var generator, and performing coupled calculation with the reactive power compensation amount.

[0017] As a further improvement of the present invention, the adaptive harmonic separation algorithm integrates a harmonic liability quantification module. When it is detected that the proportion of the 3rd harmonic in the characteristic harmonic components is > 40%, the compensation capacity allocation ratio of the active power filter is automatically increased to 70% of the total compensation amount, and at the same time, the fundamental wave compensation intensity of the static var generator is reduced by 15%.

[0018] As a further improvement of the present invention, the safety closed-loop control includes a three-level protection mechanism: primary protection: when the voltage mutation > ±20% of the rated value, the compensation device is cut off within 10 ms; secondary protection: when the temperature of the capacitor bank > 85 °C, forced air cooling is started and derated operation is carried out; tertiary protection: when the harmonic distortion rate > 8%, an alarm is triggered and switched to the isolation compensation mode.

[0019] As a further improvement of the present invention, after the three-level protection mechanism is triggered, the system executes a self-diagnosis program for the state of the compensation device: by injecting characteristic harmonic signals to detect the equivalent series resistance value of the capacitor bank, if the detected value deviates from the initial value by ±30%, an early warning of equipment deterioration is generated, and a recommended maintenance plan is displayed on the human-machine interface.

[0020] Advantages of the present invention: (1) By using an improved LSTM neural network to fuse multi-dimensional operating parameters for load prediction, combining a continuous adjustment mode with a grouped switching intelligent rotation strategy, the present invention improves the compensation response speed to the millisecond level (<20 ms). At the same time, through dynamic load balancing, the daily switching times of the capacitor bank are reduced to less than 30 times, and the equipment life is extended by more than 40%.

[0021] (2) The present invention uses an adaptive harmonic separation algorithm to achieve collaborative compensation of APF and SVG, precisely suppresses high-order harmonics of the 3rd order and above, reduces the total harmonic distortion rate (THD) of the power grid from 30% to within 5%, stabilizes the power factor above 0.95, and controls the compensation capacity error within the range of ±3%.

[0022] (3) The present invention constructs a three-level protection mechanism and a self-diagnosis program for the state of the equipment, shortens the fault response time to the 10 ms level through dynamic threshold adjustment, combines the characteristic harmonic injection detection technology to achieve early warning of capacitor bank deterioration, improves the mean time between failures (MTBF) of the system to more than 8000 hours, and reduces the major accident rate by 90%. Detailed implementation manners

[0023] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present application clearer, the present application will be further described in detail below with reference to embodiments. It should be understood that the embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0024] The present invention provides a control method for a low-voltage intelligent reactive power compensation system of a submerged arc furnace, including the following steps:

[0025] Step 1, Real-time data acquisition: Collect the three-phase voltage, current, power factor, harmonic content, and temperature parameters on the low-voltage side of the submerged arc furnace through a sensor network.

[0026] Step 2, Dynamic load forecasting: Based on a deep learning algorithm, predict the load fluctuations of the submerged arc furnace in the next 1 - 5 minutes to generate a pre-judgment value of the compensation demand.

[0027] Step 3, Multi-modal compensation strategy matching: Dynamically switch among three modes: fixed compensation, grouped switching, and continuous regulation according to the load forecasting results and real-time operating conditions.

[0028] Step 4, Harmonic collaborative suppression: Combine an active power filter (APF) and a static var generator (SVG) to dynamically track and compensate harmonic components.

[0029] Step 5, Safe closed-loop control: Real-time monitor the overvoltage, overcurrent, and temperature parameters of the system, trigger a hierarchical protection mechanism, and feedback to the compensation strategy adjustment module.

[0030] In step 1 of the present invention, a multi-source data fusion technology is adopted. A Rogowski coil is added to the sensor network to collect current mutation signals, and the displacement sensor data of the electrode lifting mechanism is synchronized. The data sampling frequency is dynamically adjusted according to the smelting stage: 10 kHz sampling is used during the melting period, and the sampling is switched to 5 kHz during the refining period.

[0031] In the present invention, an improved LSTM neural network model is used for dynamic load forecasting. The input parameters include the historical load curve, electrode position signal, raw material composition data, and smelting stage identifier. The model shares training weights in multiple submerged arc furnace scenarios through transfer learning. The improved LSTM neural network model includes an attention mechanism module. A working condition feature weighting unit is set in the decoder layer, and a dynamic weight coefficient of 0.3 - 0.5 is assigned to the FeO content index in the raw material composition data, and a priority coefficient of 0.6 - 0.8 is assigned to the submerged arc depth parameter in the electrode position signal.

[0032] In the multi-modal compensation strategy matching of the present invention, when the load volatility exceeds ±15% and the duration > 30 seconds, the continuous regulation mode is started, and the millisecond-level dynamic response of reactive power is realized through an IGBT converter. When the grouped switching mode is implemented, a capacitor bank cycle switching sequence is set. When it is detected that the daily switching times of a single capacitor bank > 50 times, this group is automatically marked as a standby unit, and the adjacent capacitor bank is activated to form a new switching combination.

[0033] In the present invention, the harmonic collaborative suppression adopts an adaptive harmonic separation algorithm, which specifically includes: analyzing the real-time harmonic spectrum through FFT to identify the characteristic harmonic components of the 2nd to 25th order; assigning the compensation tasks of the 3rd and higher order harmonic components to the active power filter; assigning the compensation tasks of the 2nd to 5th order dominant harmonic components to the static var generator, and coupling and calculating with the reactive power compensation amount. The adaptive harmonic separation algorithm integrates a harmonic responsibility quantification module. When it is detected that the proportion of the 3rd harmonic in the characteristic harmonic components is > 40%, the compensation capacity allocation ratio of the active power filter is automatically increased to 70% of the total compensation amount, and at the same time, the fundamental wave compensation intensity of the static var generator is reduced by 15%.

[0034] The safety closed-loop control in the present invention includes a three-level protection mechanism: Primary protection: When the voltage mutation > ±20% of the rated value, the compensation device is cut off within 10 ms; Secondary protection: When the temperature of the capacitor bank > 85 °C, forced air cooling is started and the operation is derated; Tertiary protection: When the harmonic distortion rate > 8%, an alarm is triggered and the system is switched to the isolated compensation mode. After the three-level protection mechanism is triggered, the system executes a self-diagnosis program for the state of the compensation equipment: detecting the equivalent series resistance value of the capacitor bank by injecting a characteristic harmonic signal. If the detected value deviates from the initial value by ±30%, an equipment deterioration warning is generated, and a recommended maintenance plan is displayed on the human-machine interface.

[0035] Embodiment (Application example of a 35 MVA submerged arc furnace in a metallurgical plant):

[0036] (I) System configuration parameters

[0037] Equipment type Technical parameters Submerged arc furnace capacity 35 MVA, secondary side voltage 690 V, maximum reactive power demand 25 MVar Compensation device Fixed compensation: 4 groups × 2.5 MVar TSC; SVG: 8 MVar; APF: 5 MVar Sensor network Rogowski coil (bandwidth 0 - 50 kHz) + PT100 temperature sensor array (32 - point temperature measurement) Control processor Dual - core DSP + FPGA architecture, operation cycle 100 μs

[0038] (II) Implementation process

[0039] (1) Optimization of real-time data acquisition

[0040] A laser displacement sensor (accuracy ±1 mm) is installed on the electrode column to monitor the submerged arc depth in real time, and an XRF online component analyzer (FeO detection accuracy ±0.5%) is configured on the raw material conveyor belt. During the melting period, a sampling frequency of 10 kHz is used to capture the arc mutation characteristics, and during the refining period, the mode is switched to 5 kHz to reduce data redundancy.

[0041] (2) Training of the dynamic load prediction model

[0042] An improved LSTM network is constructed (structure: input layer with 64 nodes → LSTM layer with 128 nodes → attention mechanism layer → output layer with 3 nodes). The input parameters include: (1) Time series data: the load curve of the previous 60 minutes (1 s interval); (2) Operating parameters: FeO content (28.5%), submerged arc depth (1.2 m), and smelting stage code (melting period = 1). The model loads the pre-trained weights of similar submerged arc furnaces through transfer learning and fine-tunes them locally for 30 epochs, and the prediction error is reduced to 4.7%.

[0043] (3) Multi-modal Compensation Strategy Execution

[0044] When the system detects that the load volatility ΔQ = +18% (lasting for 35 seconds), it triggers the continuous regulation mode: (1) The SVG outputs 6.2 MVar of capacitive reactive power through the IGBT converter (switching frequency 2 kHz) within 15 ms; (2) At the same time, the TSC group switching is blocked to avoid capacitor surges. When the fluctuation subsides (ΔQ < ±5% lasting for 5 minutes), it automatically switches back to the group switching mode and evenly uses it according to the cyclic sequence of "Group 1 → Group 3 → Group 2 → Group 4".

[0045] (4) Harmonic Coordination Governance Implementation

[0046] FFT analysis shows that the proportion of the 3rd harmonic reaches 42.7% (THD = 27.5%), triggering the harmonic responsibility quantification module: (1) The APF allocates 70% of its capacity (3.5 MVar) to compensate for the 3rd, 5th, and 7th harmonics; (2) The SVG allocates 30% of its capacity (2.4 MVar) to handle the 2nd harmonic + fundamental wave compensation. After compensation, the measured THD drops to 4.3%, and the power factor is improved to 0.97.

[0047] (5) Safety Protection and Self-diagnosis

[0048] When simulating a 22% voltage swell fault: (1) The primary protection cuts off the SVG DC bus within 8 ms; (2) The self-diagnosis program injects 250 Hz characteristic harmonics, and the measured ESR of the capacitor increases from 0.12 Ω to 0.18 Ω (deviating by 50%); (3) The HMI displays "The capacitor of Group 3 is deteriorated, it is recommended to replace", and the system automatically removes it from the switching sequence.

[0049] (III) Performance Test Data

[0050] Indicators Traditional method This embodiment Improvement amplitude Dynamic response time 1.2s 18 ms 98.5％↓ Power factor 0.68-0.72 0.95-0.98 38.9％↑ THD (full load) 27.5％ 4.3％ 84.4％↓ Daily average switching times of capacitors 58 times / group 24 times / group 58.6％↓ Fault response time 120 ms 8 ms 93.3％↓ Power loss 8.7 kWh / t 5.1 kWh / t 41.4％↓

[0051] (IV) Conclusion

[0052] The test of this embodiment for 3 months of continuous operation of a 35 MVA submerged arc furnace shows that: (1) The improved LSTM model has an accuracy rate of 92.4% for load prediction 30 seconds later, enabling the compensation device to preset the best operating point 200 ms in advance; (2) Through the APF / SVG collaborative compensation strategy, the 3rd harmonic filtering efficiency reaches 89.7%, which is 53% higher than that of the traditional LC filter; (3) The intelligent rotation strategy reduces the usage deviation of each group of capacitors from ±35% to ±12%, and the expected equipment life is extended from 3 years to 5.2 years; (4) The characteristic harmonic injection method can give an early warning of the early failure of the capacitor 30 days in advance, avoiding the loss of unplanned shutdown.

[0053] In summary, for the control method of the low-voltage intelligent reactive power compensation system of the submerged arc furnace of the present invention, during the actual operation of the system, the safety protection and self-diagnosis mechanism play a crucial role. Once abnormal system parameters are detected, such as voltage, current or temperature exceeding the preset safety range, the system will immediately activate the corresponding protection mechanism. For example, in the case of a sudden voltage rise, the primary protection mechanism can quickly cut off the SVG DC bus within a very short time (such as within 8 ms), effectively preventing equipment damage. At the same time, the secondary and tertiary protection mechanisms will also take corresponding measures according to the specific situation, such as starting forced air cooling, derating operation or triggering an alarm and switching to the isolated compensation mode to ensure the stable operation of the system.

[0054] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A control method for a low-voltage intelligent reactive power compensation system of a submerged arc furnace, characterized in that, It includes the following steps: Step 1, Real-time data acquisition: Collect three-phase voltage, current, power factor, harmonic content, and temperature parameters on the low-voltage side of the submerged arc furnace through a sensor network; Step 2, Dynamic load prediction: Based on a deep learning algorithm, predict the load fluctuation of the submerged arc furnace in the next 1-5 minutes to generate a pre-judgment value of the compensation demand; Step 3, Multi-modal compensation strategy matching: Dynamically switch among three modes: fixed compensation, grouped switching, and continuous regulation according to the load prediction result and real-time working conditions; Step 4, Harmonic collaborative suppression: Combine an active power filter (APF) and a static var generator (SVG) to dynamically track and compensate harmonic components; Step 5, Safety closed-loop control: Real-time monitor the overvoltage, overcurrent, and temperature parameters of the system, trigger a hierarchical protection mechanism, and feedback it to the compensation strategy adjustment module.

2. The control method of a low-voltage intelligent reactive power compensation system for a submerged arc furnace according to claim 1, characterized in that, In Step 1, a multi-source data fusion technology is adopted. A Rogowski coil is added to the sensor network to collect current mutation signals, and the displacement sensor data of the electrode lifting mechanism is synchronized. The data sampling frequency is dynamically adjusted according to the smelting stage: 10 kHz sampling is used during the melting period, and it is switched to 5 kHz sampling during the refining period.

3. The control method of a low-voltage intelligent reactive power compensation system for a submerged arc furnace according to claim 1, characterized in that, The dynamic load prediction adopts an improved LSTM neural network model. The input parameters include the historical load curve, electrode position signal, raw material composition data, and smelting stage identifier. The model shares the training weights in multiple submerged arc furnace scenarios through transfer learning.

4. The control method of a low-voltage intelligent reactive power compensation system for a submerged arc furnace according to claim 3, characterized in that, The improved LSTM neural network model includes an attention mechanism module. A working condition feature weighting unit is set in the decoder layer, and a dynamic weight coefficient of 0.3-0.5 is assigned to the FeO content index in the raw material composition data, and a priority coefficient of 0.6-0.8 is assigned to the submerged arc depth parameter in the electrode position signal.

5. The control method of a low-voltage intelligent reactive power compensation system for a submerged arc furnace according to claim 1, characterized in that, In the multi-modal compensation strategy matching, when the load volatility exceeds ±15% and the duration > 30 seconds, the continuous regulation mode is started, and the millisecond-level dynamic response of reactive power is realized through an IGBT converter.

6. The control method of a low-voltage intelligent reactive power compensation system for a submerged arc furnace according to claim 5, characterized in that, When the grouped switching mode is implemented, a capacitor bank cyclic switching sequence is set. When it is detected that the daily switching times of a single capacitor bank > 50 times, this group is automatically marked as a standby unit, and the adjacent capacitor bank is activated to form a new switching combination.

7. The control method of a low-voltage intelligent reactive power compensation system for a submerged arc furnace according to claim 1, characterized in that, The harmonic collaborative suppression adopts an adaptive harmonic separation algorithm, which specifically includes: Analyze the real-time harmonic spectrum through FFT to identify the characteristic harmonic components of 2-25 times; Assign the compensation task of the 3rd and higher-order harmonics to the active power filter; Assign the compensation task of the 2-5th dominant harmonics to the static var generator, and perform a coupled calculation with the reactive power compensation amount.

8. The control method of a low-voltage intelligent reactive power compensation system for a submerged arc furnace according to claim 7, characterized in that, The adaptive harmonic separation algorithm integrates a harmonic responsibility quantification module. When it is detected that the proportion of the 3rd harmonic in the characteristic harmonic components > 40%, the compensation capacity allocation ratio of the active power filter is automatically increased to 70% of the total compensation amount, and at the same time, the fundamental wave compensation intensity of the static var generator is reduced by 15%.

9. The control method of a low-voltage intelligent reactive power compensation system for a submerged arc furnace according to claim 1, characterized in that, The described safety closed-loop control includes a three-level protection mechanism: primary protection: when the voltage mutation > ±20% of the rated value, the compensation device is cut off within 10 ms; secondary protection: when the temperature of the capacitor bank > 85 °C, forced air cooling is started and derated operation is carried out; tertiary protection: when the harmonic distortion rate > 8%, an alarm is triggered and switched to the isolated compensation mode.

10. The control method of a low-voltage intelligent reactive power compensation system for a submerged arc furnace according to claim 9, characterized in that, After the described three-level protection mechanism is triggered, the system executes a self-diagnosis program for the state of the compensation equipment: the equivalent series resistance value of the capacitor bank is detected by injecting a characteristic harmonic signal. If the detected value deviates from the initial value by ±30%, an equipment deterioration warning is generated and a recommended maintenance plan is displayed on the man-machine interface.