Control system of reactive voltage regulation and control device of mobile energy storage charging and discharging bin

Abstract

The invention discloses a control system for a reactive voltage regulation and control device of a mobile energy storage charging and discharging bin, and particularly relates to the technical field of automation and energy storage of a power system, and the control system comprises a multi-dimensional sensing unit, an edge intelligent decision-making unit, a self-adaptive execution unit, a safety protection unit and a man-machine interaction unit. According to the control system of the reactive voltage regulation and control device of the mobile energy storage charging and discharging bin, the multi-dimensional sensing unit realizes comprehensive and accurate acquisition of power grid, energy storage and environment parameters by integrating a power grid parameter detection module, an energy storage state monitoring module and an environment sensing module, a reliable data basis is provided for regulation and control, and meanwhile, the control system has the advantages of being simple in structure and convenient to operate. The edge intelligent decision-making unit can greatly improve the decision-making precision and the multi-bin cooperation efficiency by means of an LSTM-Kalman filtering model, a variable coefficient virtual inertia algorithm, federated learning and a TSN network, and effectively solves the problems of reactive power unbalance and voltage frequency fluctuation.

Description

Technical Field

[0001] This invention relates to the fields of power system automation and energy storage technology, and in particular to a control system for a mobile energy storage charging and discharging chamber reactive voltage regulation device. Background Technology

[0002] With the large-scale grid connection of new energy sources and the surge in demand for distributed power grids and emergency power supply scenarios, mobile energy storage charging and discharging warehouses have become an indispensable interactive unit in modern power systems due to their advantages such as flexible deployment, rapid response to load fluctuations, and ability to smooth out power grid oscillations.

[0003] However, reactive power imbalance and voltage frequency fluctuations remain core issues in grid interaction. Existing control systems have significant shortcomings. At the sensing level, parameter acquisition is limited in scope and accuracy, focusing only on basic grid parameters with sampling frequencies often below 1kHz, making it difficult to capture harmonic distortion rates. Energy storage device status monitoring (such as battery SOC and temperature) has errors exceeding ±1℃, and environmental temperature and humidity, as well as storage vibration monitoring, are ignored. Early warning systems for extreme conditions are lacking. At the decision-making and coordination level, intelligence and coordination capabilities are insufficient. Simple algorithms are frequently used, and high-precision prediction models are absent. Short-term parameter prediction accuracy is below 98%, and control... The system suffers from several drawbacks: Lag and frequency fluctuations are difficult to control within ±0.1Hz; multi-compartment linkage clock synchronization accuracy is only at the millisecond level; discrete control equipment operates frequently with response times exceeding 20ms; in terms of execution and safety protection, the execution unit topology is non-modular, current control mode switching is prone to problems, battery health management lacks a life prediction model based on the depth of discharge, fault warning accuracy is below 95%, and measures such as electrical insulation detection and communication encryption are lacking; in terms of human-machine interaction, only basic parameter display is provided, and there are no visualization functions such as voltage quality trend analysis; remote communication transmission rate is below 100Mbps, resulting in low operation and maintenance efficiency.

[0004] Therefore, a control system for the reactive voltage regulation device of the mobile energy storage charging and discharging chamber is needed. Summary of the Invention

[0005] The main objective of this invention is to provide a control system for a mobile energy storage charging and discharging chamber reactive voltage regulation device, which can effectively solve the problems mentioned above.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A control system for a mobile energy storage charging and discharging chamber reactive voltage regulation device includes a multi-dimensional sensing unit, an edge intelligent decision-making unit, an adaptive execution unit, a safety protection unit, and a human-machine interaction unit;

[0008] The output of the multi-dimensional perception unit is electrically connected to the input of the edge intelligent decision-making unit. The output of the edge intelligent decision-making unit is connected to the adaptive execution unit and the security protection unit respectively. The human-computer interaction unit is bidirectionally connected to the edge intelligent decision-making unit.

[0009] The multi-dimensional sensing unit is used to collect power grid operating parameters, energy storage device status, and environmental parameters.

[0010] The edge intelligent decision-making unit adopts a three-layer architecture to realize data processing, distributed collaboration and control strategy generation;

[0011] The adaptive execution unit uses a modular three-phase half-bridge topology to execute reactive power compensation and voltage regulation commands.

[0012] The security protection unit includes electrical safety, communication security, and health management modules;

[0013] The human-computer interaction unit is used for parameter configuration and status monitoring.

[0014] Preferably, the multi-dimensional sensing unit includes a power grid parameter detection module and an energy storage status monitoring module. The power grid parameter detection module collects voltage, current, frequency and harmonic distortion rate parameters, with a sampling frequency of not less than 1kHz. The energy storage status monitoring module includes a SOC sensor and a temperature monitoring circuit, with a temperature measurement accuracy of ±0.5℃.

[0015] Preferably, the multi-dimensional sensing unit further includes an environmental sensing module, which integrates a temperature and humidity sensor and a vibration sensor, wherein the humidity measurement range is 0-100%RH and the vibration detection frequency range is 10-1000Hz, for use in extreme working condition early warning.

[0016] Preferably, the decision layer of the edge intelligent decision-making unit integrates an LSTM-Kalman filter hybrid prediction model, and the short-term prediction accuracy is calculated using the following formula:

[0017]

[0018] The decision-making layer employs a variable coefficient virtual inertia control algorithm to control frequency fluctuations within ±0.1Hz.

[0019] Preferably, the collaborative layer of the edge intelligent decision-making unit adopts a federated learning framework and realizes clock synchronization between nodes through a time-sensitive network (TSN) with a synchronization accuracy of ±1μs.

[0020] The virtual inertia control algorithm satisfies:

[0021]

[0022] Among them, M EΔf is the dynamically adjustable virtual inertia coefficient, Δp is the frequency deviation, and ΔP is the output power adjustment.

[0023] Preferably, the health management module of the safety protection unit adopts a battery cycle life prediction model, and the life loss calculation formula based on the depth of discharge is as follows:

[0024]

[0025] Fault warning accuracy rate ≥95%.

[0026] Preferably, the edge intelligent decision-making unit adopts an improved artificial bee colony algorithm and introduces a dynamic adjustment mechanism for the virtual damping coefficient, which reduces the number of actions of the discrete control equipment by more than 60%, and supports seamless switching between direct current control and indirect current control modes, with a response time of ≤20ms.

[0027] Preferably, the human-machine interaction unit is equipped with a touch screen and a remote communication interface, providing visualization functions for voltage quality trend analysis and equipment status assessment. The communication interface adopts fiber optic Ethernet with a transmission rate of ≥100Mbps.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] 1. The multi-dimensional sensing unit of this invention integrates grid parameter detection, energy storage status monitoring and environmental sensing modules to achieve comprehensive and accurate acquisition of grid, energy storage and environmental parameters, providing a reliable data foundation for regulation. At the same time, the edge intelligent decision unit, with its LSTM-Kalman filter model, variable coefficient virtual inertia algorithm, federated learning and TSN network, can significantly improve decision accuracy and multi-warehouse collaboration efficiency, effectively solving the problems of reactive power imbalance and voltage frequency fluctuation.

[0030] 2. The adaptive execution unit of this invention adopts a modular three-phase half-bridge topology, supporting seamless switching between direct and indirect current control. The edge intelligent decision unit reduces the number of actions of discrete control equipment by more than 60% through an improved artificial bee colony algorithm and virtual damping adjustment, significantly reducing execution losses and failure probability. The safety protection unit comprehensively avoids safety risks by combining an electrical and communication safety module with a battery life prediction model and a fault warning accuracy of ≥95%. The human-machine interaction unit greatly improves operation and maintenance efficiency and adapts to complex scenario requirements by providing voltage quality trend analysis, equipment status assessment visualization, and ≥100Mbps fiber optic Ethernet transmission. Attached Figure Description

[0031] Figure 1 This is a flowchart of the overall modules of the present invention. Detailed Implementation

[0032] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0033] like Figure 1 As shown, a control system for a mobile energy storage charging and discharging chamber reactive voltage regulation device includes a multi-dimensional sensing unit, an edge intelligent decision-making unit, an adaptive execution unit, a safety protection unit, and a human-machine interaction unit.

[0034] The output of the multi-dimensional perception unit is electrically connected to the input of the edge intelligent decision-making unit. The output of the edge intelligent decision-making unit is connected to the adaptive execution unit and the security protection unit respectively. The human-computer interaction unit is bidirectionally connected to the edge intelligent decision-making unit.

[0035] The multi-dimensional sensing unit is used to collect power grid operating parameters, energy storage device status, and environmental parameters.

[0036] The edge intelligent decision-making unit adopts a three-layer architecture to realize data processing, distributed collaboration and control strategy generation;

[0037] The adaptive execution unit uses a modular three-phase half-bridge topology to execute reactive power compensation and voltage regulation commands.

[0038] The security protection unit includes electrical safety, communication security, and health management modules;

[0039] The human-computer interaction unit is used for parameter configuration and status monitoring.

[0040] In the above, the multi-dimensional sensing unit first deploys a grid parameter detection module at the grid access end of the mobile energy storage charging and discharging compartment. The module's voltage acquisition end is connected to the three-phase grid line through a voltage transformer, and the current acquisition end is connected to the grid line through a current transformer. An energy storage status monitoring module is installed next to the energy storage battery pack inside the energy storage compartment. The SOC sensor is connected to the communication port of the battery management system. The temperature detection circuit's temperature probe is attached to the outer shell of each battery cell. An environmental sensing module is fixed on the top and sides of the compartment. The temperature and humidity sensor faces the air circulation area, and the vibration sensor is fixed to the metal frame of the compartment. All modules are connected through a CAN bus to form a sensing network.

[0041] Its multi-dimensional sensing unit output uses RVVP type shielded cable (cross-sectional area 1.5mm²). 2 Connect to the input terminal of the edge intelligent decision unit. Before connecting, use a multimeter to check the continuity of the line and ensure that the resistance is ≤1Ω. After connecting, use a signal generator to send a 380V analog voltage signal and a 50A analog current signal to verify that the receiving deviation is ≤0.5%.

[0042] In the above, the output of the edge intelligent decision unit is connected to the control motherboard interface of the adaptive execution unit and the signal input interface of the safety protection unit through an industrial Ethernet cable. After connection, a 10kvar reactive power compensation command is sent to detect whether the IGBT conduction and shutdown of the adaptive execution unit is normal, and the equipment over-temperature signal (60℃) is simulated to check whether the safety protection unit triggers protection.

[0043] In the above, the human-machine interaction unit is bidirectionally connected to the Ethernet port of the edge intelligent decision-making unit via a Category 5e unshielded twisted-pair cable. After connection, a power grid frequency query command is sent to the touch screen, and the verification feedback delay is ≤50ms. When the multi-dimensional sensing unit collects power grid operating parameters, the power grid parameter detection module collects voltage, current, frequency, and harmonic rate every 0.833ms at a sampling frequency of 1.2kHz (meeting a requirement of not less than 1kHz). For example, if the secondary side voltage is 3.8V (transformer ratio 100:1, actual 380V) and the secondary side current is 0.5A (transformer ratio 100:1, actual 50A), the frequency is calculated to be 50Hz and the harmonic rate to be 1.2% using Fourier transform. When collecting the status of energy storage equipment, the SOC sensor reads the battery pack SOC every 5 seconds. For example, if the rated capacity is 200kWh and the discharged capacity is 40kWh, the SOC is calculated according to the formula. Calculation yields The temperature monitoring circuit checks the battery temperature every 2 seconds. For example, if a PT100 battery has a resistance of 109.8Ω, the temperature will be calculated according to the formula R. t =100(1+3.9083×10) -3 Calculations show that the temperature t = 25.08℃, with a deviation of ≤0.3℃.

[0044] When collecting environmental parameters, the temperature and humidity sensor collects the temperature at 28℃ and the humidity at 55%RH every 10 seconds, and the vibration sensor collects the vibration frequency at 80Hz every 1 second.

[0045] In the above three-layer architecture of the edge intelligent decision-making unit, the data processing layer processes 10 voltage data points (380.1V, 380.2V, etc.) using a moving average filtering method, according to the formula... (n=10, U i The average value (based on a single voltage acquisition) is calculated to be 380.05V. The distributed coordination layer synchronizes its clock with the decision-making units of the two surrounding warehouses through the TSN network, ensuring a deviation of ≤1μs. The control strategy layer, based on a grid load of 90kW and a power factor of 0.85, first calculates the apparent power S = Then follow the formula Calculations show that reactive power compensation is required. Generate control commands;

[0046] In the above, the adaptive execution unit controls the three-phase half-bridge module IGBT according to instructions. For example, when compensating for 15.77 kvar, it follows the formula... (U=380V,sinφ=0.5) Calculation yields the required output current. This can be achieved by adjusting the IGBT conduction angle;

[0047] In the above, the electrical safety module of the safety protection unit measures an insulation resistance ≥1MΩ, the communication safety module verifies data frames (e.g., "01020304" checksum 0A), and the health module is designed to cycle 2000 times, having already cycled 600 times, according to the formula. Calculations show that the remaining lifespan is 1400 cycles.

[0048] In the above, the human-machine interaction unit touch screen is set with voltage thresholds of 360V-400V and SOC thresholds of 20%-95%, and displays parameters in real time. Fiber optic Ethernet transmits data at a rate of 100Mbps. The multi-dimensional sensing unit is used to collect power grid operating parameters, energy storage device status, and environmental parameters. The edge intelligent decision-making unit adopts a three-layer architecture to realize data processing, distributed collaboration, and control strategy generation. The adaptive execution unit adopts a modular three-phase half-bridge topology to execute reactive power compensation and voltage regulation commands. The safety protection unit includes electrical safety, communication safety, and health management modules. The human-machine interaction unit is used for parameter configuration and status monitoring.

[0049] Furthermore, the multi-dimensional sensing unit includes a power grid parameter detection module and an energy storage status monitoring module. The power grid parameter detection module collects voltage, current, frequency and harmonic distortion rate parameters, with a sampling frequency of not less than 1kHz. The energy storage status monitoring module includes a SOC sensor and a temperature monitoring circuit, with a temperature measurement accuracy of ±0.5℃.

[0050] In the above, the power grid parameter detection module first installs a voltage transformer (the primary side is connected to the three-phase line of the power grid, and the secondary side outputs 0-5V) and a current transformer (the primary side is connected to the power grid line, and the secondary side outputs 0-5A) at the power grid access end of the mobile energy storage charging and discharging compartment. The signal conditioning circuit in the module filters and amplifies the analog signal.

[0051] Set the sampling frequency to 1.1kHz (ensuring it is not lower than 1kHz), and the sampling period... Each sampling triggers the A / D converter, such as a secondary voltage of 3.82V (turns ratio 100:1, actual U... 实际 =3.82×100=382V), secondary side current 0.45A (turns ratio 200:1, actual I 实际 =0.45×200=90A), performing a Fourier transform on the voltage signal yields the fundamental U1=382V, the second harmonic U2=4.58V, and the third harmonic U3=3.06V. According to the formula... Calculation yields

[0052] The SOC sensor in the energy storage status monitoring module sends a request to the BMS every 10 seconds via the CAN bus. The BMS uses the ampere-hour integration method; for example, if the rated capacity is 150kWh and the cumulative discharge is 37.5kWh, it calculates the value using the formula... Calculation yields It also provides feedback, and stores the sensor verification data after verification.

[0053] The temperature monitoring circuit uses 8 PT100 sensors attached to the battery casing to collect resistance data every 2 seconds. For example, if a cell's resistance is 109.8Ω, the resistance is calculated using the formula R. t =100(1+3.9083×10) -3 t) Calculate the temperature The deviation from the standard value is 0.2℃. In a construction site scenario, when two 50kW tower cranes are running, the module uses a secondary side voltage of 3.78V (actual 3.78×100=378V) and a current of 0.85A (actual 0.85×200=170A). The Fourier transform yields a 5th harmonic of 5.67V, which is calculated to be 1.596 according to the harmonic distortion rate formula. The average current of the 5th harmonic at a frequency of 1.1kHz is 170A.

[0054] Before construction, the SOC was 90%. After 8 hours of discharge, the current was 40A (voltage 500V, power P = UI = 500 × 40 = 20kW). The discharged capacity was Pt = 20 × 8 = 160kWh (rated 400kWh). Calculated using the SOC formula... The temperature of 16 batteries was tested. The temperature of the 8th battery was 26.3℃ and the temperature of the 14th battery were 25.7℃, with a deviation from the actual value of ≤0.1℃. The power grid parameter detection module collects voltage, current, frequency and harmonic distortion rate parameters and the sampling frequency is not less than 1kHz. The energy storage status monitoring module includes a SOC sensor and a temperature monitoring circuit and the temperature measurement accuracy is ±0.5℃.

[0055] Furthermore, the multi-dimensional sensing unit also includes an environmental sensing module, which integrates a temperature and humidity sensor and a vibration sensor. The humidity measurement range is 0-100%RH, and the vibration detection frequency range is 10-1000Hz, which is used for early warning of extreme working conditions.

[0056] In the above, the environmental sensing module has a mounting hole in the middle of the side of the mobile energy storage charging and discharging compartment. The integrated temperature and humidity sensor and vibration sensor are fixed by a waterproof bracket. The sensing surface of the temperature and humidity sensor faces outward and is 10cm away from the compartment. The vibration sensor is fixed to the compartment frame with M5 bolts (torque 8N·m).

[0057] The temperature and humidity sensor is first calibrated in a 50% RH standard environment so that 2.5V corresponds to 50% RH, with a sampling interval of 1 second. If the output is 3.2V, it is calculated according to the formula. Calculation yields The value is within 0-100%RH; the vibration sensor uses piezoelectric ceramic elements to convert vibration into voltage, with internal high-pass filtering (cutoff 10Hz) and low-pass filtering (cutoff 1000Hz), and frequency extraction by spectrum analysis. For example, the main peak of the voltage signal is found to be 120Hz, which is within 10-1000Hz.

[0058] Extreme operating condition warning settings include a high humidity threshold of 90% RH, a low humidity threshold of 1096 RH, a high-frequency vibration threshold of 800 Hz (lasting 10 seconds), and a low-frequency vibration threshold of 20 Hz (lasting 30 seconds); for outdoor camping scenarios, humidity is measured at 3 AM. High humidity warning triggered; 820Hz vibration (12 seconds) triggered vibration warning when the off-road vehicle passes through the sampling area;

[0059] Desert midday humidity measurement Low humidity warning is triggered; the cabin vibrates at 15Hz for 25 seconds, and no warning is issued if the vibration duration is less than 30 seconds; the temperature and humidity sensor uses 6096RH and the standard 60.296RH, with a deviation Δhumidity = 60 - 60.2 = -0.2%RH; the vibration sensor uses a 500Hz standard signal and measures 502Hz, with a deviation Δf = 502 - 500 = 2Hz; the environmental sensing module integrates temperature and humidity sensors and vibration sensors, with a humidity measurement range of 0-10096RH and a vibration detection frequency range of 10-1000Hz, used for warnings under extreme working conditions.

[0060] Furthermore, the decision layer of the edge intelligent decision-making unit integrates an LSTM-Kalman filter hybrid prediction model, and the short-term prediction accuracy is calculated using the following formula:

[0061]

[0062] The decision-making layer employs a variable coefficient virtual inertia control algorithm to control frequency fluctuations within ±0.1Hz.

[0063] In the above, the decision-making level retrieves power grid data from the multi-dimensional sensing unit over the past 48 hours (once every 30 seconds, for a total of...). The first 2400 groups were used as the training set, and the last 480 groups were used as the test set.

[0064] The training set frequency (49.8-50.2Hz) is normalized according to the formula. Processing, such as normalizing to 50.0Hz, yields... The normalized data is input into an LSTM network (time step 20, 2 hidden layers, 32 neurons, Adam optimizer), iterated 500 times until the loss value is ≤0.0005; the test set is input into the LSTM network to obtain a normalized prediction value of 0.52, and the inverse normalization formula x = xnorm ×(x max -x min )+x min The calculated inverse normalized prediction value is x = 0.52 × (50.2 - 49.8) + 49.8 = 50.008 Hz;

[0065] Start the Kalman filter, setting the initial state x0 = 50.0Hz, initial error covariance P0 = 0.001, process noise covariance Q = 0.0001, and measurement noise covariance R = 0.0003. Acquire the actual frequency z = 50.01Hz, and apply the Kalman gain formula. Calculation yields Calculated using the state update formula x1=x0+K(z-x0), we get x1=50.0+0.769×(50.01-50.0)=50.00769Hz. Calculated using the error covariance update formula P1=(1-K)P0, we get P1=(1-0.769)×0.001=0.000231. Using the weighted fusion formula… (With both LSTM and Kalman filter weights at 0.5), the final predicted value is obtained. The actual value is 50.01Hz. calculate,

[0066] The decision-making layer variable coefficient virtual inertia control algorithm sets the virtual inertia coefficient M. E Correspondence table:

[0067] When Δf∈[-0.03,0.03]Hz, M E =60kW·s / Hz;

[0068] When Δf∈[-0.07,-0.03)∪(0.03,0.07]Hz, M E = 90kW·s / Hz;

[0069] When Δf∈[-0.1,-0.07)∪(0.07,0.1]Hz, M E =130kW·s / Hz;

[0070] The current power grid frequency is 49.92Hz (rated frequency f). N =50Hz), according to the frequency deviation formula Δf = ff N Calculations show that Δf = 49.92 - 50 = -0.08 Hz, corresponding to M. E = 130 kW·s / Hz; within 0.5 seconds, the frequency drops from 49.93 Hz to 49.92 Hz, according to the frequency change formula Calculation yields According to the power regulation formula Calculations show that ΔP = 130 × (-0.02) = -2.6kW, meaning the output power is reduced by 2.6kW.

[0071] The frequency subsequently rose back to 49.96Hz. According to the frequency deviation formula, Δf = 49.96 - 50 = -0.04Hz, corresponding to M... E = 90 kW·s / Hz, the frequency increases from 49.94 Hz to 49.96 Hz in 0.5 seconds, according to the frequency change formula... According to the power regulation formula, ΔP = 90 × 0.02 = 1.8kW, meaning the output power increases by 1.8kW. The final grid frequency fluctuation range is 49.92-50.07Hz, within ±0.1Hz. The decision-making level uses a variable coefficient virtual inertia control algorithm to keep frequency fluctuations within ±0.1Hz.

[0072] Furthermore, the collaborative layer of the edge intelligent decision-making unit adopts a federated learning framework and realizes clock synchronization between nodes through a time-sensitive network (TSN), with a synchronization accuracy of ±1μs.

[0073] The virtual inertia control algorithm satisfies:

[0074]

[0075] Among them, M E Δf is the dynamically adjustable virtual inertia coefficient, Δp is the frequency deviation, and ΔP is the output power adjustment.

[0076] In the above process, the collaboration layer first deploys federated learning clients at the edge intelligent decision-making units of each mobile energy storage charging and discharging compartment. Each client connects to the local multi-dimensional sensing unit database to store locally collected grid parameters, energy storage status, and environmental data. For example, compartment A stores the data from the past 24 hours. The storage unit B contains 1440 sets of frequency and SOC data for the same period, including voltage and current data.

[0077] The federated learning iteration cycle is set to 50 rounds. In each round, each client randomly selects 200 sets of data from its local database and trains a local model using local computing resources. For example, warehouse A uses 200 sets of voltage and current data to train a reactive power compensation calculation model. The model parameters are adjusted using the least squares method to minimize the calculation error. To calculate the reactive power compensation for the model, Q 实际,i (The actual reactive power compensation) ≤ 0.5%;

[0078] Each client encrypts its trained local model parameters (excluding the original data) and transmits them to the federated learning server via a Time-Sensitive Network (TSN). Before transmission, the parameters are compressed, such as converting 32-bit floating-point numbers to 16-bit fixed-point numbers, to reduce the data volume. After receiving all client parameters, the server calculates weights based on the proportion of each client's data volume. The total data volume N = nA + nB + nC + ... = 1440 + 1440 + 1200 + 960 = 5040 sets (assuming 1200 sets for database C and 960 sets for database D). Update formula based on global model parameters (K is the number of clients, θ) k Calculate the updated global model parameters for the k-th client local model parameters;

[0079] The updated global model parameters are then distributed to each client. The clients update their local models with the new parameters and proceed to the next iteration, until 50 iterations are completed and the global model error stabilizes at ≤0.3%. During TSN time-sensitive network clock synchronization, each node first sends its local clock signal (e.g., warehouse A clock tA = 10:00:00.000000, warehouse B clock tB = 10:00:00.000002, and the server clock...). tS=10:00:00.000001), calculate the deviation of each node according to the clock deviation formula Δtk=tk-tS, the deviation of warehouse A ΔtA=10:00:00.000000-10:00:00.000001=-0.000001s, the deviation of warehouse B ΔtB=10:00:00.000002-10:00:00.000001=+0.000001s;

[0080] The server sends clock adjustment commands to each node. Container A adjusts its clock to t′A=tA-ΔtA=10:00:00.000000-(-0.000001)=10:00:00.000001s, and Container B adjusts its clock to t′B=tB-ΔtB=10:00:00.000002-0.000001=10:00:00.000001s. After adjustment, the deviation is checked again until the clock synchronization accuracy of all nodes reaches ±1μs.

[0081] In the variable coefficient virtual inertia control algorithm, the virtual inertia coefficient ME is dynamically adjusted based on the grid frequency deviation Δf and the frequency change rate.

[0082] When Δf ≤ -0.08Hz and At that time, M E = 150 kW·s / Hz; when -0.08 Hz < Δf ≤ -0.03 Hz and At that time, ME =100kW·s / Hz;

[0083] When -0.03Hz < Δf < 0.03Hz, M E =60kW·s / Hz;

[0084] When 0.03Hz≤Δf<0.08Hz and At that time, M E =100kW·s / Hz;

[0085] When Δf ≥ 0.08 Hz and At that time, M E =150kW·s / Hz;

[0086] For example, if the grid frequency is 49.89Hz (rated frequency 50Hz), according to the frequency deviation formula, Δf = 49.89 - 50 = -0.11Hz. The frequency drops from 49.90Hz to 49.89Hz within 1 second. According to the frequency change rate formula... Hz / s, corresponding to M E =150kW·s / Hz, according to the power regulation formula The calculation yields ΔP = 150 × (-0.01) = -1.5kW, meaning the output power is reduced by 1.5kW.

[0087] When the frequency rises to 49.95Hz, according to the frequency deviation formula, Δf = 49.95 - 50 = -0.05Hz. Within 0.8 seconds, the frequency rises from 49.93Hz to 49.95Hz. According to the frequency change rate formula... Corresponding to M E =100kW·s / Hz, calculated using the power adjustment formula, ΔP = 100 × 0.025 = 2.5kW, increasing the output power by 2.5kW. The collaborative layer uses a federated learning framework to achieve clock synchronization between nodes through a Time-Sensitive Network (TSN) with a synchronization accuracy of ±1μs. The virtual inertia control algorithm meets the requirements. (M E The virtual inertia coefficient Δf is dynamically adjustable. 为频率偏差 ΔP 为输出功率调节量 ).

[0088] Furthermore, the health management module of the safety protection unit adopts a battery cycle life prediction model, and the life loss calculation formula based on the depth of discharge is as follows:

[0089]

[0090] Fault warning accuracy rate ≥95%.

[0091] In the above, the health management module first obtains the design cycle number N of the energy storage battery. 设计 =2000 times, while setting the equivalent full cycle count standard: deep discharge (discharge depth DOD≥80%) 1 cycle counts as C 等效1 = 1 equivalent full cycle, moderate discharge (40% ≤ DOD < 80%) is counted as 1 cycle C 等效2 = 0.5 equivalent full cycles, mild discharge (DOD < 40%) is counted as 1 cycle. 等效3 = 0.2 equivalent full loop counts;

[0092] The module collects battery charge and discharge data once per hour and records the depth of discharge for each discharge. If the first discharge capacity is 160kWh (rated 200kWh), then according to the discharge depth formula, we get... (Counted as 1 time);

[0093] The second discharge capacity was 140 kWh. (Counted as 0.5 times);

[0094] The third discharge capacity is 60kWh. (Counted as 0.2 times);

[0095] The fourth discharge capacity was 130 kWh. (Counted as 0.5 times);

[0096] The fifth discharge capacity is 50 kWh. (Counted as 0.2 times);

[0097] The capacity of the 6th discharge is 180kWh. (Counted as 1 time);

[0098] According to the formula for cumulative equivalent total cycle number (m is the number of discharges) Calculate to get C 累计 =1 + 0.5 + 0.2 + 0.5 + 0.2 + 1 = 3.4 times;

[0099] According to the formula for remaining cycle life Calculation yields The result is 1997.

[0100] When a fault warning occurs, the module monitors the battery's voltage, temperature, and internal resistance parameters in real time and sets the voltage warning threshold: single cell voltage U < 2.5V or U > 3.7V, temperature warning threshold T > 45℃ or T < -10℃, and internal resistance warning threshold R > R0 × (1 + 20%) (R0 is the initial internal resistance).

[0101] If a battery initially has an internal resistance R0 = 50mΩ and its current internal resistance R = 62mΩ, then according to the formula for internal resistance increase... Calculation yields Trigger fault warning;

[0102] During battery use, at the 500th cycle, the cumulative equivalent full cycle count C 累计 =480 cycles, calculated using the remaining cycle life formula.

[0103] In the 1000th iteration, the cumulative equivalent total number of iterations is C. 累计 =950 cycles, calculated using the remaining cycle life formula.

[0104] During the monitoring, the temperature of a certain battery was found to be T = 46℃ > 45℃, exceeding the temperature warning threshold and triggering an alarm. When calculating the accuracy of the fault warning, 100 warning events were recorded, of which 96 were confirmed as actual faults. The accuracy was calculated using the formula... Calculation yields The health management module employs a battery cycle life prediction model, and the life loss calculation formula based on the depth of discharge is as follows: Fault warning accuracy rate ≥95%.

[0105] Furthermore, the edge intelligent decision-making unit adopts an improved artificial bee colony algorithm and introduces a dynamic adjustment mechanism for the virtual damping coefficient, which reduces the number of actions of discrete control equipment by more than 60%, and supports seamless switching between direct current control and indirect current control modes with a response time of ≤20ms.

[0106] In the above, the edge intelligent decision-making unit first sets the parameters of the improved artificial bee colony algorithm, including a colony size of 50 (25 hired bees and 25 observation bees), a maximum number of iterations of 100, and a search dimension corresponding to the action parameters of discrete control equipment (such as IGBT switching frequency and reactive power compensation).

[0107] A virtual damping coefficient ξ is introduced, with an initial value of ξ0 = 0.3. The rate of change of the number of actions of the reference discrete control device is dynamically adjusted. (Δn is the change in the number of movements, and Δt is the change in time). When r ≥ 5 times / minute, ξ = ξ0 + 0.1 = 0.4.

[0108] When 2 times / minute ≤ r < 5 times / minute, ξ = ξ0 = 0.3;

[0109] When r < 2 times / minute, ξ = ξ0 - 0.1 = 0.2, and the virtual damping coefficient ranges from 0.1 to 0.5.

[0110] During algorithm execution, the hired bees search according to the neighborhood search formula v ij =x ij +φ ij (x ij -x kj ), x ij Let x be the j-th dimension parameter of the i-th bee. kj Let φ be the j-th dimension parameter of the k-th random bee. ij Generate a new solution for a random number in the range [-1, 1]. For example, if the first dimension parameter of the 3rd hired bee is x_{31} = 1000Hz (IGBT switching frequency), and the 7th bee is randomly selected with x_{71} = 1200Hz, φ 31 =0.6, according to the formula, v 31 =1000+0.6×(1000-1200)=880Hz;

[0111] Calculate the fitness value of the new solution, and set the fitness function as follows: If the old solution corresponds to 15 actions / hour, the fitness The new solution corresponds to 9 actions / hour, with fitness... Because f 新 >f 旧 The hired bee is replaced with a new solution; the observation bee selects the hired bee using a roulette wheel selection method, with a selection probability of... If the fitness of the third hired bee is 0.1, the total fitness Σf i =1.8, therefore After selecting based on probability, a new solution is generated and updated;

[0112] The scout bee triggers when the mercenary bee fails to update its solution for 10 consecutive times, randomly generating a new solution x. ij =x min,j +rand(0,1)(x max,j -x min,j (where x_{min,j} and x_{max,j} are the upper and lower limits of the j-th dimension parameter), for example, the lower limit of the IGBT switching frequency is 800Hz and the upper limit is 1500Hz, and rand(0,1)=0.3, so x ij =800+0.3×(1500-800)=1010Hz;

[0113] After introducing a virtual damping coefficient, when the rate of change of the number of actions r = 6 times / minute (≥5) and ξ = 0.4, φ is adjusted during the neighborhood search. ij The range is [-0.4, 0.4], which reduces the variation of the new solution and decreases the number of actions from 15 times / hour to 6 times / hour, thus reducing...

[0114] When switching between direct current control and indirect current control modes, a switching threshold is set: indirect current control is used when the grid voltage fluctuation ΔU ≤ 2%, and direct current control is used when ΔU > 2%. Voltage fluctuation is determined according to the formula. Calculation (U) N =380V), if the mains voltage is 375V, then... Controlled by indirect current;

[0115] The voltage dropped to 370V, so... The switching is triggered, and the current control parameters (such as current command 100A) are saved during the switching process. The target control mode parameters (such as direct current control PI parameters) are loaded. The switching delay is ≤10ms and the total response time is ≤20ms. The edge intelligent decision unit adopts an improved artificial bee colony algorithm and introduces a virtual damping coefficient dynamic adjustment mechanism, which reduces the number of actions of discrete control equipment by more than 60%, and supports seamless switching between direct current control and indirect current control modes with a response time of ≤20ms.

[0116] Furthermore, the human-machine interaction unit is equipped with a touch screen and a remote communication interface, providing visualization functions for voltage quality trend analysis and equipment status assessment. The communication interface adopts fiber optic Ethernet with a transmission rate of ≥100Mbps.

[0117] In the above, the human-machine interaction unit first installs a 10.1-inch touch screen (1280×800 resolution) on the operation panel of the mobile energy storage charging and discharging compartment. The touch screen is connected to the edge intelligent decision unit through an RS485 bus. The communication baud rate is set to 115200bps, with 8 data bits, 1 stop bit, and no parity bit.

[0118] The remote communication interface uses a fiber optic Ethernet interface (SC type connector), connected to the remote monitoring center via single-mode fiber (transmission distance ≤ 2km), with a transmission rate set to 100Mbps (meeting 100Mbps requirements). In the voltage quality trend analysis function, the touchscreen stores grid voltage data every 5 minutes, generating a 24-hour trend curve based on the time series. Voltage quality assessment intervals are set as follows: Excellent (380±2%V, i.e., 372.4V-387.6V), Good (380±5%V, i.e., 361V-399V), Acceptable (380±10%V, i.e., 342V-418V), and Unacceptable (<342V or >418V). For example, if the voltage data for a certain period is: 10:00 382V (Excellent), 12:00 375V (Excellent), 14:00 365V (Good), 16:00 350V (Acceptable), the percentages of Excellent (50%), Good (25%), and Acceptable (25%) are tallied, and a trend analysis report is generated.

[0119] The equipment condition assessment function collects the SOC, temperature, and insulation resistance parameters of the energy storage device, and sets the assessment criteria: SOC > 80% and temperature > 80%. Furthermore, insulation resistance ≥2MΩ is preferred, 60% ≤ SOC < 80% and 30℃ < temperature. Furthermore, 1MΩ ≤ insulation resistance < 2MΩ is considered good; 40% ≤ 50°C < 60% and 35°C < temperature ≤ 40°C and 0.5MΩ ≤ insulation resistance < 1MΩ is considered medium; SOC < 400.5MΩ is considered poor. For example, if the current SOC is 75%, temperature is 32°C, and insulation resistance is 1.5MΩ, it is evaluated as good. The device status level and parameter details are displayed on the touch screen.

[0120] During remote communication, fiber optic Ethernet transmits data according to the TCP / IP protocol, sending one frame of status data per second (containing 20 bytes of parameters such as voltage, current, and SOO), according to the transmission rate formula. Calculations show that the amount of data transmitted per second is 20 × 8 = 160 bits, and the transmission rate is v = 160 bits / s << 100 × 10 6 bit / s, meets the requirements;

[0121] In the scenario of backup power supply in commercial complexes, the touch screen displays a voltage of 385V (excellent) and SOC of 92% (excellent) at 9:00 AM, and a voltage of 370V (good) and SOC of 85% (excellent) at 11:00 AM. The remote monitoring center receives data in real time and generates daily reports. The human-machine interaction unit is equipped with a touch screen and a remote communication interface, providing visualization functions for voltage quality trend analysis and equipment status assessment. The communication interface adopts fiber optic Ethernet with a transmission rate of ≥100Mbps.

[0122] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A control system for a mobile energy storage charging and discharging chamber reactive voltage regulation device, characterized in that, It includes a multi-dimensional perception unit, an edge intelligent decision-making unit, an adaptive execution unit, a security protection unit, and a human-computer interaction unit; The output of the multi-dimensional perception unit is electrically connected to the input of the edge intelligent decision-making unit. The output of the edge intelligent decision-making unit is connected to the adaptive execution unit and the security protection unit respectively. The human-computer interaction unit is bidirectionally connected to the edge intelligent decision-making unit. The multi-dimensional sensing unit is used to collect power grid operating parameters, energy storage device status, and environmental parameters. The edge intelligent decision-making unit adopts a three-layer architecture to realize data processing, distributed collaboration and control strategy generation; The adaptive execution unit uses a modular three-phase half-bridge topology to execute reactive power compensation and voltage regulation commands. The security protection unit includes electrical safety, communication security, and health management modules; The human-computer interaction unit is used for parameter configuration and status monitoring.

2. The control system of the mobile energy storage charging and discharging chamber reactive voltage regulation device according to claim 1, characterized in that: The multi-dimensional sensing unit includes a power grid parameter detection module and an energy storage status monitoring module. The power grid parameter detection module collects voltage, current, frequency and harmonic distortion rate parameters, with a sampling frequency of not less than 1kHz. The energy storage status monitoring module includes a SOC sensor and a temperature monitoring circuit, with a temperature measurement accuracy of ±0.5℃.

3. The control system of the mobile energy storage charging and discharging chamber reactive voltage regulation device according to claim 1, characterized in that: The multi-dimensional sensing unit also includes an environmental sensing module, which integrates a temperature and humidity sensor and a vibration sensor. The humidity measurement range is 0-100%RH, and the vibration detection frequency range is 10-1000Hz, which is used for early warning of extreme working conditions.

4. The control system of the mobile energy storage charging and discharging chamber reactive voltage regulation device according to claim 1, characterized in that: The decision layer of the edge intelligent decision-making unit integrates an LSTM-Kalman filter hybrid prediction model, and the short-term prediction accuracy is calculated using the following formula: The decision-making layer employs a variable coefficient virtual inertia control algorithm to control frequency fluctuations within ±0.1Hz.

5. The control system of the mobile energy storage charging and discharging chamber reactive voltage regulation device according to claim 1, characterized in that: The collaborative layer of the edge intelligent decision-making unit adopts a federated learning framework and realizes clock synchronization between nodes through a time-sensitive network (TSN), with a synchronization accuracy of ±1μs. The virtual inertia control algorithm satisfies: Among them, M E Δf is the dynamically adjustable virtual inertia coefficient, Δp is the frequency deviation, and ΔP is the output power adjustment.

6. The control system of the mobile energy storage charging and discharging chamber reactive voltage regulation device according to claim 1, characterized in that: The health management module of the safety protection unit adopts a battery cycle life prediction model, and the life loss calculation formula based on the depth of discharge is as follows: Fault warning accuracy rate ≥95%.

7. The control system of the mobile energy storage charging and discharging chamber reactive voltage regulation device according to claim 1, characterized in that: The edge intelligent decision-making unit adopts an improved artificial bee colony algorithm and introduces a dynamic adjustment mechanism for the virtual damping coefficient, which reduces the number of actions of discrete control equipment by more than 60%, and supports seamless switching between direct current control and indirect current control modes with a response time of ≤20ms.

8. The control system of the mobile energy storage charging and discharging chamber reactive voltage regulation device according to claim 1, characterized in that: The human-machine interaction unit is equipped with a touch screen and a remote communication interface, providing visualization functions for voltage quality trend analysis and equipment status assessment. The communication interface adopts fiber optic Ethernet with a transmission rate of ≥100Mbps.

Images