Multi-source intelligent anti-error scheduling method and system for transformer substation

Abstract

The invention provides a multi-source intelligent anti-error scheduling method and system for a transformer substation, and the method comprises the steps: collecting and preprocessing multi-source heterogeneous data, carrying out the dynamic recognition of an intermediate state, constructing an equipment state time sequence model based on a long-short-term memory network and a graph neural network, recognizing the intermediate state in switching operation in real time, and carrying out the logic judgment and updating of intelligent interlocking. A hybrid rule engine is adopted, a predefined interlocking logic rule is combined to carry out real-time interlocking logic judgment, anti-error decision and early warning, operation suggestions are generated based on a deep reinforcement learning model, real-time early warning is carried out on high-risk operation, and an operator is prompted to correct steps through an augmented reality interface. According to the invention, in an intermediate state caused by insufficient switching and closing, anti-error scheduling and interlocking judgment can be effectively carried out, and the operation safety and reliability of the transformer substation are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent anti-misoperation in power systems, and particularly to a multi-source intelligent anti-misoperation dispatching method and system for substations. Background Art

[0002] With the rapid development of intelligent substations, the number and complexity of substation equipment have been continuously increasing. Traditional substation anti-misoperation systems mainly rely on static rule libraries and manual experience, making it difficult to cope with complex dynamic scenarios. Switching operations are important links in substation operation and maintenance. Insufficient closing or opening of switches may cause equipment to be in an intermediate state, leading to the risk of misoperation and posing a hidden danger to the safe and stable operation of the power grid. Existing anti-misoperation systems mainly rely on fixed rules and logics for judgment, with poor adaptability to the complex and changeable operating states of substations. Moreover, they have deficiencies in dealing with intermediate states caused by insufficient closing or opening of switches and interlock judgment logics, making it difficult to adapt to the changing power grid operation modes.

[0003] The patent application with the application number 2024117969548 discloses an invention patent for an intelligent dispatching operation ticket anti-misoperation control system driven by power dispatching service data. It conducts anti-misoperation verification based on conditions such as the closed or open states of circuit breakers and disconnectors, and whether grounding wires are removed. However, in actual operation of the above-mentioned equipment in substations, the switches often enter an unexpected intermediate state due to insufficient closing or opening (such as the circuit breaker being disconnected but the disconnector not being fully closed), and the above anti-misoperation verification cannot identify it, leading to the risk event of misdispatching.

[0004] The patent application with the application number 2024113963576 discloses an invention patent for an anti-misoperation protection system for control elements in a power grid. The system includes multiple anti-misoperation interlock devices, which are correspondingly arranged one by one with multiple control elements that need interlock control in the power grid. The anti-misoperation interlock devices are used to collect the current state information of the corresponding control elements, transmit the current state information to the centralized control device, and perform anti-misoperation interlock protection on the control elements based on the target control instructions transmitted from the centralized control device. However, the above anti-misoperation interlock devices mainly consider the physical logical relationship of equipment cascades and do not consider dynamic interlock strategies related to power supply energy-saving strategies and load rates of electricity consumption, such as adjusting the grounding switch locking strategy according to real-time load.

[0005] Based on the above analysis, the present invention aims to provide a multi-source intelligent anti-misoperation dispatching method and system for substations, which can integrate multiple data sources, utilize artificial intelligence technology, and achieve accurate anti-misoperation judgment for substation dispatching operations. Especially in the intermediate state caused by insufficient closing or opening of switches, it can effectively perform anti-misoperation dispatching and interlock judgment, improving the safety and reliability of substation operation. Summary of the Invention

[0006] Aiming at the deficiencies of the prior art, the technical problem to be solved by the present invention is to provide a multi-source intelligent anti-error dispatching method and system for a substation to solve the problems existing in the prior art.

[0007] To solve the above technical problems, in the first aspect, a multi-source intelligent anti-error dispatching method for a substation is provided, including the following steps: S100: Multi-source heterogeneous data collection and preprocessing, collecting real-time status data, historical operation records, environmental parameters, video surveillance streams and power grid topology information of the substation, and constructing a dynamic knowledge graph; S0200: Intermediate state dynamic recognition, based on long short-term memory network and graph neural network, constructing a device state time series model to real-time recognize the intermediate state in the switching operation, where the intermediate state includes the semi-closed and ungrounded device working states in the switching operation; S300: Intelligent interlock logic judgment and update, using a hybrid rule engine to perform real-time interlock logic judgment in combination with predefined interlock logic rules, where the interlock logic judgment rules include real-time energy topology information and energy consumption dynamic data to adjust the locking conditions, and optimizing the locking conditions through real-time topology analysis to reduce redundant locking; S400: Anti-error decision-making and early warning, generating operation suggestions based on a deep reinforcement learning model, giving real-time early warning for high-risk operations, and prompting the operator to correct the steps through an augmented reality interface.

[0008] Optionally, the step S100: Multi-source heterogeneous data collection and preprocessing includes: S101: Internet of Things sensor data collection, real-time monitoring of the closing position change of the circuit breaker contacts through a high-precision position sensor installed on the circuit breaker, collecting pressure data of the circuit breaker operating mechanism through a pressure sensor, and monitoring vibration data of the circuit breaker during the operation through a vibration sensor; S102: SCADA system data collection, collecting current, voltage, active power, reactive power data of the circuit breaker, as well as the energy storage state signal and closing and opening indication signals of the circuit breaker; S103: Video surveillance device data collection, high-definition cameras and infrared thermal imagers real-time collect video surveillance data of the circuit breaker and disconnector, the closing action speed of the circuit breaker is shown in the video, and the temperature data at the contacts of the circuit breaker is shown by the infrared thermal imager; S104: Data preprocessing, the data cleaning module eliminates the out-of-range error data points caused by device vibration in the position sensor data, performs sliding window filtering on the pressure sensor data, the size of the filtering window is 5, and the normalization processing module normalizes the circuit breaker position, current, closing speed, vibration amplitude, energy and energy consumption data to the [0,1] interval.

[0009] Optionally, the intermediate state dynamic recognition in step S200 includes: S201: LSTM model analysis. Input the preprocessed circuit breaker position data sequence into the intermediate state recognition model based on LSTM, analyze the current circuit breaker position data sequence, and output the closing speed and position change curve data. S202: GNN topology modeling. Combine the topology relationship diagram of substation equipment. The GNN model analyzes the electrical connection relationship between the circuit breaker and the disconnector, as well as the association relationship with equipment such as the busbar and transmission line, and determines the closing state of the circuit breaker through impedance abnormality in the electrical connection. S203: Intermediate state judgment. If the closing position does not reach 95%, it is judged that the circuit breaker is in the intermediate state, and a misoperation risk assessment is carried out.

[0010] Optionally, the real-time energy topology information includes: when switching between the fan and the thermal power unit, if the output power of the fan is insufficient and the startup time of the thermal power unit does not meet the energy-saving requirements, the startup operation of the thermal power unit is locked; before the fan is connected to the grid, it is necessary to ensure that its speed is within the range of 90%-110% of the rated speed and the bus voltage is stable within ±5% of the rated voltage, otherwise the grid connection operation is locked; the energy consumption dynamic data includes: during the peak electricity price period, if the energy storage system has sufficient power and the photovoltaic power generation is greater than or equal to 50% of the real-time load, the connection operation of the commercial power is locked.

[0011] Optionally, the input of the deep reinforcement learning model is the current state of the substation, including the equipment closing position, real-time interlock topology information, load data, and energy supply status.

[0012] Optionally, the output of the deep reinforcement learning model is the Q value of each possible operation: Q(s,a)←Q(s,a)+ α×[R(s,a)+ γ×maxQ(s',a')-Q(s,a)] Where Q(s,a) is the current Q value of performing action a in state s; α is the learning rate, 0 < α < 1; R(s,a) is the immediate reward obtained after performing action a in state s; γ is the discount factor, used to balance future rewards and immediate rewards, 0 < γ < 1; maxQ(s',a') is the Q value corresponding to the action with the maximum Q value among all actions that can be performed in the new state s' after performing action a.

[0013] Optionally, the real-time warning form includes that the closing position of the circuit breaker is highlighted in red, the uncompleted closing steps are prompted to the operator in the form of animation, and remote alarms are sent through text messages and mobile APP push messages.

[0014] In a second aspect, the present invention further provides a multi-source intelligent anti-misoperation dispatching system for a substation, including: A multi-source data acquisition unit: used for acquiring and preprocessing multi-source heterogeneous data, acquiring real-time status data of the substation, historical operation records, environmental parameters, video surveillance streams, and power grid topology information, and constructing a dynamic knowledge graph; An intermediate state determination module: based on a long short-term memory network and a graph neural network, constructs a device state time series model to real-time identify intermediate states in switching operations, where the intermediate states include semi-closed and ungrounded device working states in switching operations; An interlock logic real-time update module: used for intelligent interlock logic judgment and update, adopts a hybrid rule engine, combines predefined interlock logic rules for real-time interlock logic judgment, where the interlock logic judgment rules include real-time energy topology information and energy consumption dynamic data to adjust the locking conditions, and optimizes the locking conditions through real-time topology analysis to reduce redundant locking; A strategy generation and early warning module, used for anti-misoperation decision-making and early warning, generates operation suggestions based on a deep reinforcement learning model, gives real-time early warnings for high-risk operations, and prompts operators to correct steps through an augmented reality interface.

[0015] In a third aspect, the present invention further provides an electronic device, including: One or more processors; A memory; and one or more programs stored in the memory, where the one or more programs include instructions for executing any of the above-mentioned multi-source intelligent anti-misoperation dispatching methods for the substation.

[0016] In a fourth aspect, the present invention further provides a computer-readable storage medium, including one or more programs for execution by one or more processors of an electronic device, where the one or more programs include instructions for executing any of the above-mentioned multi-source intelligent anti-misoperation dispatching methods for the substation.

[0017] Compared with the prior art, the technical solution of the present invention has the following beneficial effects: 1. Break through the limitations of traditional final state detection, real-time identify and early warn of unexpected intermediate states, and reduce the probability of misoperation.

[0018] 2. Integrate a rule engine and reinforcement learning, improve the flexibility and adaptability of the interlock strategy, adjust the locking conditions through real-time energy topology information and energy consumption dynamic data, and achieve energy conservation and improvement of power supply efficiency.

[0019] 3. Multi-modal data fusion: Integrate multi-source heterogeneous data through a knowledge graph to support accurate decision-making in complex scenarios. Description of the Drawings

[0020] Figure 1Flow chart of a multi-source intelligent anti-error dispatching method for a substation provided by an embodiment of the present invention; Figure 2 Flow chart of a multi-source intelligent anti-error dispatching system for a substation provided by an embodiment of the present invention. Specific implementation manners

[0021] Obviously, many modifications and variations made by those skilled in the art based on the purpose of the present invention fall within the protection scope of the present invention.

[0022] Those skilled in the art of the present technology can understand that unless specifically stated otherwise, the singular forms "a", "an", "the", and "said" used herein may also include the plural forms. It should be further understood that the term "including" used in the specification of the present invention means the presence of the described features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or their groups. It should be understood that when an element or component is referred to as being "connected" to another element or component, it can be directly connected to other elements or components, or there may also be intermediate elements or components. The phrase "and / or" used herein includes any and all combinations of one or more of the associated listed items.

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention. Embodiment 1

[0024] Refer to Figure 1 , which is an embodiment of the present invention. This embodiment provides a multi-source intelligent anti-error dispatching method for a substation, including: S100 Multi-source heterogeneous data collection and preprocessing: Collect real-time status data, historical operation records, environmental parameters, video surveillance streams, and power grid topology information of the substation, and construct a dynamic knowledge graph.

[0025] In an optional embodiment, in a certain 500 kV substation, an operator performs a switching operation of "transferring the main transformer high-voltage side circuit breaker from operation to maintenance": S100 Multi-source heterogeneous data collection and preprocessing specifically includes: S101: Internet of Things Sensor Data Acquisition: A high-precision position sensor (accuracy 0.01 mm) installed on the circuit breaker monitors the closing position change of the circuit breaker contacts in real time, and the collected position data is 0.02 mm. A pressure sensor (range 0 - 1000 Pa) collects the pressure data of the circuit breaker operating mechanism, and the current pressure is displayed as 500 Pa. A vibration sensor monitors the vibration condition of the circuit breaker during operation, and the collected vibration frequency is 13.5 Hz.

[0026] S102: SCADA System Data Acquisition: The SCADA system collects the current data of the circuit breaker as 200 A, the voltage data as 500 kV, the active power as 120 MW, and the reactive power as 80 Mvar. At the same time, the collected energy storage state signal of the circuit breaker is "energized", and the closing and opening indication signal is "closed".

[0027] S103: Video Surveillance Equipment Data Acquisition: A high-definition camera and an infrared thermal imager collect video surveillance data of the circuit breaker and disconnector in real time. The video shows that the closing action of the circuit breaker is slightly slow, and the infrared thermal imaging shows that the temperature at the circuit breaker contacts is slightly higher than other parts, and the temperature value is 65 °C.

[0028] S104: Data Preprocessing: The data cleaning module eliminates the out-of-range error data points in the position sensor data caused by equipment vibration, performs sliding window filtering on the pressure sensor data, the size of the filtering window is 5, and the smoothness of the processed pressure data is improved. The normalization processing module normalizes the circuit breaker position data to the [0, 1] interval, and the normalized position data is 0.02. The current data is normalized to the [0, 1] interval, and the normalized current data is 0.4. The feature extraction module calculates the closing speed of the circuit breaker as 0.02 mm / s, the closing synchronization deviation as 0.001 s, and the vibration amplitude as 0.05 mm.

[0029] S0200 Intermediate State Dynamic Identification: Based on the Long Short-Term Memory Network (LSTM) and the Graph Neural Network (GNN), construct a device state time series model to identify the intermediate state in the switching operation in real time.

[0030] In an optional embodiment, the construction of the intermediate state dynamic model specifically includes: S201: LSTM Model Analysis: Input the preprocessed circuit breaker position data sequence into the intermediate state recognition model based on LSTM. The LSTM model is trained with a large amount of historical operation data and learns the time series characteristics of normal closing operations. The model analyzes the current circuit breaker position data sequence and finds that its closing speed is slower than normal operations, and there are abnormal fluctuations in the position change curve.

[0031] S202: GNN Topological Modeling: Combining with the topological relation diagram of substation equipment, the GNN model analyzes the electrical connection relationship between circuit breakers and disconnectors, as well as the association relationship with equipment such as busbars and transmission lines. The model discovers that there is a certain impedance anomaly in the electrical connection between the circuit breaker and the disconnector, which may affect the closing state of the circuit breaker.

[0032] S203: Intermediate State Judgment: The model comprehensively judges that the circuit breaker is in an intermediate state (the closing position has not reached 95%), and there is a risk of misoperation (the probability of misclosing is 85%). The system records this judgment result and prepares for subsequent risk assessment and early warning operations.

[0033] S300 Intelligent Interlock Logic Judgment and Update: Using a hybrid rule engine, real-time interlock logic judgment is performed in combination with predefined interlock logic rules. The interlock logic judgment rules include real-time energy topology information and energy consumption dynamic data to adjust the blocking conditions, and the blocking conditions are optimized through real-time topology analysis to reduce redundant blocking.

[0034] Among them, the intelligent interlock logic judgment and update specifically include: In an optional embodiment, for preventing misoperation in dispatching operations under multiple energy supply modes, in a certain intelligent substation, a hybrid power supply mode of wind power generation and thermal power generation is adopted, and at the same time, energy-saving requirements need to be met. This substation is equipped with two wind turbine generators (wind turbine A and wind turbine B), one thermal power generator (thermal generator C), and related power transformation and distribution equipment.

[0035] High-precision rotational speed sensors and power sensors are respectively installed on wind turbine A and wind turbine B, and their rotational speeds are monitored in real time as 1200 rpm and 1150 rpm respectively, and the output powers are 1.2 MW and 1.0 MW respectively; a steam pressure sensor and a temperature sensor are installed on thermal generator C, and its steam pressure is monitored in real time as 1500 kPa, the temperature is 500 °C, and the output power is 3.5 MW.

[0036] The real-time energy consumption data of each generator set is collected through energy consumption monitoring equipment, including the unit power generation energy consumption of wind turbine A and wind turbine B being 0.12 kg standard coal / kWh and 0.13 kg standard coal / kWh respectively, and the unit power generation energy consumption of thermal generator C being 0.35 kg standard coal / kWh.

[0037] Substation operation data collection, the SCADA system collects the real-time load data of the substation as 5.0 MW, the bus voltage is 220 kV, and the status signals of each circuit breaker and disconnector are normal.

[0038] The data cleaning module eliminates the error data points beyond the range caused by equipment vibration in the data of the fan speed sensor, and performs sliding window filtering on the data of the lighter steam pressure sensor. The size of the filtering window is 5, and the smoothness of the processed pressure data is improved. The normalization processing module normalizes the output power data of each generator set to the interval [0, 1]. The normalized powers of Fan A, Fan B, and Lighter C are 0.4, 0.33, and 0.7 respectively. The feature extraction module calculates the power fluctuation characteristics, start-up time characteristics, etc. of each generator set. The power fluctuation amplitude of Fan A is 0.05 MW, the fluctuation frequency is 0.2 Hz, and the start-up time of Lighter C is 1.2 hours.

[0039] Pre-defined interlock logic rules: S301 has established the following interlock logic judgment rules according to the operation logic and energy-saving requirements of multiple energy power supply equipment: When switching between the fan and the lighter, if the output power of the fan is insufficient and the start-up time of the lighter does not meet the energy-saving requirements, the start-up operation of the lighter is blocked.

[0040] Before the fan is connected to the grid, it is necessary to ensure that its speed is within the range of 90% - 110% of the rated speed, and the bus voltage is stable within ±5% of the rated voltage. Otherwise, the grid connection operation is blocked.

[0041] Real-time interlock logic judgment: Judge according to the currently collected data and the pre-defined interlock logic rules: Calculate that the sum of the actual output powers of Fan A and Fan B is 2.2 MW. Compared with the real-time load of 5.0 MW, the remaining load of 2.8 MW needs to be borne by Lighter C. However, the current start-up time of Lighter C does not reach the energy-saving requirement of 1.5 hours. Therefore, the start-up operation of Lighter C is blocked, and a warning is issued to prompt the dispatcher.

[0042] Check that the speed of Fan A, 1200 rpm, is outside the range of 90% - 110% (900 rpm - 1100 rpm) of its rated speed of 1000 rpm, and the bus voltage of 220 kV is stable within ±5% (209 kV - 231 kV) of the rated voltage of 220 kV. Therefore, the grid connection operation of Fan A is blocked, and the operator is prompted to check the reason for the abnormal fan speed.

[0043] In another alternative embodiment, in the energy-saving operation mode of a substation, the substation adopts a hybrid power supply mode of photovoltaic power generation and commercial power, and is equipped with an energy storage system to improve the energy utilization efficiency.

[0044] Collect the real-time load data of the substation, the photovoltaic power generation power, the charge and discharge status of the energy storage system, and the peak-valley electricity price information of the power grid. At a certain moment, the real-time load is 3.0 MW, the photovoltaic power generation power is 1.8 MW, the energy storage system is in the charging state, the battery level is 60%, and it is currently in the peak electricity price period with a high electricity price.

[0045] The substation's energy conservation needs are assessed based on real-time peak and valley electricity prices and carbon emission limits. Due to peak electricity prices and strict carbon emission limits, the system prioritizes energy-saving operations, maximizing the use of photovoltaic power generation and energy storage systems to reduce utility power usage, thereby lowering energy costs and carbon emissions.

[0046] S302 establishes the following interlocking logic judgment rules based on energy-saving requirements: During peak electricity price periods, if the energy storage system has sufficient power and the photovoltaic power generation power is greater than or equal to 50% of the real-time load, the mains access operation will be blocked and the energy storage system and photovoltaic power generation will be used for power supply first.

[0047] For the operation of non-critical equipment, if its energy consumption is high and does not affect the safe and stable operation of the power grid, the relevant operations will be locked while meeting the energy-saving requirements.

[0048] According to the energy-saving demand assessment results, optimize the interlocking logic judgment rules: The current energy storage system is at 60% capacity, which is sufficient. The photovoltaic power generation power of 1.8MW is greater than or equal to 50% (1.5MW) of the real-time load of 3.0MW, so the mains access operation is blocked to ensure that the energy storage system and photovoltaic power generation are used first to achieve energy saving goals.

[0049] For a non-critical device with high energy consumption (such as an auxiliary heating device), the system will lock its startup operation under the premise of meeting operational safety to reduce unnecessary energy consumption.

[0050] S303 performs real-time interlocking judgment on the dispatching operation instructions based on the real-time status of the equipment and the interlocking rules in the knowledge graph. For example, when the disconnector is not fully closed, the closing operation of the related circuit breaker is automatically locked to prevent accidents such as short circuits caused by misoperation; when the bus is in the intermediate state, the operation of other equipment related to the bus is locked. The interlocking judgment logic uses a combination of logic gate circuits and state machines to monitor and judge the real-time status of the equipment in real time. Once an interlocking condition is found, the corresponding interlocking action is immediately triggered. At the same time, the interlocking judgment logic also has self-test and self-recovery functions, which can automatically release the interlocking state after the equipment status returns to normal.

[0051] S400 error-prevention decision-making and early warning: Generates operational recommendations based on deep reinforcement learning (DRL), provides real-time early warnings for high-risk operations, and prompts operators to correct steps through an augmented reality (AR) interface.

[0052] Optimize the scheduling strategy through deep reinforcement learning and build a neural network model: design a deep Q network (DQN) model, the input is the current state of the substation, and the output is the Q value of each possible operation.

[0053] Input layer: The input is the current state of the substation, including equipment status (such as circuit breaker position, disconnector position, etc.), real-time topology information, load data, energy supply status, etc. The design of the state space should accurately reflect the operation of the substation and provide sufficient information for the model to make decisions.

[0054] Hidden layer: Usually, a multi-layer neural network structure is adopted, such as 2 - 3 hidden layers, and each layer contains a certain number of neurons (such as 64, 128, 256 neurons). The activation function can use ReLU (Rectified Linear Unit), which can introduce non-linearity and enable the model to learn more complex feature representations.

[0055] Output layer: The output is the Q value of each possible operation. Each operation corresponds to a Q value, which represents the expected benefit of performing this operation in the current state. The number of neurons in the output layer is consistent with the size of the action space.

[0056] In an optional embodiment, assume that in a multi-source intelligent anti-misoperation dispatching system of a certain substation, the current state is that the disconnector is not fully closed (intermediate state), and the dispatching system needs to decide whether to allow the circuit breaker to close.

[0057] The current state s of the system is that the disconnector is not fully closed. At this time, perform the action a: allow the circuit breaker to close. According to the anti-misoperation rules, closing at this time may cause equipment damage or misoperation. Therefore, the immediate reward R(s, a) obtained is -10 (negative reward).

[0058] Optionally, the reward function considers the operation safety margin of the equipment: give rewards to operations that keep the equipment state within the safe range after operation; give lower rewards or penalties to operations that bring the equipment close to the safety limit. For example, for an operation where the closing position of the equipment is greater than 90%, the reward value can be set to +1; for an operation where the closing position of the equipment is less than 70%, the reward value can be set to -2. This helps the agent maintain sufficient safety margin during the dispatching process and reduce the risk of equipment failure. The specific form of the reward function can be a weighted sum that comprehensively considers the above factors: R = w1×R correct +w2×R efficiency +w3×R energy +w4×R stability +w5×R safety Where R correct represents the operation correctness reward, R efficiency represents the operation efficiency reward, R energy represents the energy utilization efficiency reward, R stability represents the system stability reward, Rsafety Indicates the safety margin reward; w1, w2, w3, w4, w5 are the corresponding weight coefficients, which are adjusted according to specific application scenarios and requirements to balance the relative importance of each reward factor.

[0059] After executing action a, enter the new state s′, and the new state may be a device failure state. The system calculates the Q values of each possible action (such as emergency tripping, equipment maintenance, etc.) in the new state s′. Suppose the maximum Q value among them is -20.

[0060] Suppose the current learning rate α is 0.1 and the discount factor γ is 0.9. According to the Q-value update formula, update the Q value of the action that allows the circuit breaker to close in the current state: Q(s,a)←Q(s,a)+0.1×[-10+0.9×(-20)-Q(s,a)] Through this process, the system continuously updates the Q values of each state-action pair, gradually learns the optimal scheduling strategy, and achieves the goal of multi-source intelligent anti-misoperation scheduling in the substation, effectively avoiding the misoperation risk caused by the intermediate state of the equipment.

[0061] In an optional embodiment, the system highlights the uncompleted steps through the AR interface, pops up a warning message: such as "The closing position of the circuit breaker does not meet the requirements, there is a risk of misoperation!", and locks the subsequent operation permissions. On the AR interface, the closing position of the circuit breaker is highlighted in red, and the uncompleted closing steps are prompted to the operator in the form of an animation. At the same time, the misclosing probability of 85% is displayed on the interface to remind the operator to pay attention.

[0062] The system synchronously updates the dynamic knowledge graph, marks the circuit breaker as "high-risk intermediate state", and associates the status information of related equipment (disconnecting switch, bus, etc.). In the knowledge graph, the status attribute of the circuit breaker node is updated to "intermediate state", and the status attributes of the connected disconnecting switch and bus nodes are also updated accordingly to reflect the current equipment status and potential risks.

[0063] The system automatically sends a notification to the operation and maintenance personnel, pushing messages through text messages and the mobile APP: "The circuit breaker on the high-voltage side of the main transformer is in an intermediate state, please review it in time!". The text message content includes the specific location, current status of the circuit breaker, and recommended measures. The mobile APP push message is attached with the detailed status data and historical operation records of the circuit breaker, which is convenient for the operation and maintenance personnel to quickly understand the on-site situation.

[0064] Based on the deep reinforcement learning model, combined with the current equipment status and the anti-misoperation rules in the knowledge graph, evaluate the risks and benefits of different decision-making schemes. The model considers the possible consequences caused by the circuit breaker being in an intermediate state, such as short-circuit faults, equipment damage, etc., and the impact of different decision-making schemes on the power grid operation.

[0065] The model generates an anti - error decision - making suggestion: "It is recommended to suspend the subsequent power - on operation, arrange on - site personnel to check the closing mechanism of the circuit breaker, and continue the operation after confirming the normal equipment status." The decision - making suggestion is displayed to the dispatcher through the man - machine interaction terminal, and detailed inspection steps and precautions are provided to guide the operation and maintenance personnel for on - site inspection and handling.

[0066] In other alternative embodiments of the present application, the load rate of each circuit breaker can also be calculated based on the active power and reactive power data collected by the SCADA system. According to the real - time load data, based on the updated interlock logic judgment rules, the closing sequence of the circuit breakers is automatically adjusted to avoid the overload risk caused by operations. The system recommends closing the circuit breaker with a lower load first, and then gradually closing other circuit breakers to ensure the stable operation of the power grid. On the man - machine interaction terminal, the system displays the adjusted closing sequence and provides detailed operation steps and precautions.

[0067] This embodiment also provides a multi - source intelligent anti - error dispatching system for substations, including: Multi - source data acquisition unit: used for collecting and pre - processing multi - source heterogeneous data, collecting real - time substation status data, historical operation records, environmental parameters, video surveillance streams, and power grid topology information, and constructing a dynamic knowledge graph; Intermediate state determination module: Based on the long short - term memory network and graph neural network, construct a device state time - series model to real - time identify the intermediate states in switching operations, where the intermediate states include the semi - closed and ungrounded device working states in switching operations; Interlock logic real - time update module: used for intelligent interlock logic judgment and update, adopting a hybrid rule engine, combining predefined interlock logic rules for real - time interlock logic judgment, where the interlock logic judgment rules include adjusting the blocking conditions based on real - time energy topology information and energy consumption dynamic data, and optimizing the blocking conditions through real - time topology analysis to reduce redundant blocking; Strategy generation and early - warning module, used for anti - error decision - making and early - warning, generating operation suggestions based on a deep reinforcement learning (DRL) model, giving real - time early - warning for high - risk operations, and prompting the operator to correct steps through an augmented reality (AR) interface.

[0068] Practical applications show that this system can reduce the misoperation rate by more than 60%, improve the operation efficiency by 30%, and successfully avoid 3 major accidents caused by intermediate states in the pilot project of a 500kV substation. By dynamically monitoring intermediate states, optimizing interlock logic, and integrating multi - source data, the present invention effectively improves the safety and reliability of substation dispatching operations, reduces power outages and equipment damage caused by misoperations, and has significant economic and social benefits.

[0069] In this specification, the various embodiments are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the embodiments, reference can be made to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple. For the relevant parts, reference can be made to the description in the method section.

[0070] Those skilled in the art can further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in the form of hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0071] Finally, it should also be noted that in this article, relationships such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "including", "comprising" or any other variant is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such process, method, article or device.

Claims

1. A multi-source intelligent anti-error dispatching method for a substation, characterized in that, It includes the following steps: S100. Multi-source heterogeneous data collection and preprocessing: Collect real-time substation status data, historical operation records, environmental parameters, video surveillance streams, and power grid topology information, and construct a dynamic knowledge graph; S200. Intermediate state dynamic recognition: Based on long short-term memory networks and graph neural networks, construct a device state time series model to real-time recognize the intermediate states during switching operations, where the intermediate states include the semi-closed and ungrounded device working states during switching operations; S300. Intelligent interlock logic judgment and update: Adopt a hybrid rule engine to perform real-time interlock logic judgment in combination with predefined interlock logic rules, where the interlock logic judgment rules include real-time energy topology information and energy consumption dynamic data to adjust the locking conditions, and optimize the locking conditions through real-time topology analysis to reduce redundant locking; S400. Anti-misoperation decision-making and early warning: Generate operation suggestions based on a deep reinforcement learning model, give real-time early warnings for high-risk operations, and prompt operators to correct steps through an augmented reality interface.

2. The multi-source intelligent anti-misoperation dispatching method for a substation according to claim 1, wherein The step S100, multi-source heterogeneous data collection and preprocessing includes: S101: Internet of Things sensor data collection: Real-time monitor the closing position change of the circuit breaker contacts through high-precision position sensors installed on the circuit breaker, collect pressure data of the circuit breaker operating mechanism through pressure sensors, and monitor vibration data of the circuit breaker during operation through vibration sensors; S102: SCADA system data collection: Collect current, voltage, active power, and reactive power data of the circuit breaker, as well as the energy storage state signal and closing and opening indication signals of the circuit breaker; S103: Video surveillance device data collection: High-definition cameras and infrared thermal imagers real-time collect video surveillance data of the circuit breaker and disconnector. The closing action speed of the circuit breaker is shown in the video, and the temperature data at the contacts of the circuit breaker is shown by the infrared thermal imager; S104: Data preprocessing: The data cleaning module eliminates the out-of-range error data points caused by device vibration in the position sensor data, performs sliding window filtering on the pressure sensor data with a filtering window size of 5, and the normalization processing module normalizes the circuit breaker position, current, closing speed, vibration amplitude, energy, and energy consumption data to the [0, 1] interval.

3. A multi-source intelligent anti-error dispatching method for a substation according to claim 1, characterized in that, The step S200, intermediate state dynamic recognition includes: S201: LSTM model analysis: Input the preprocessed circuit breaker position data sequence into the intermediate state recognition model based on LSTM, analyze the current circuit breaker position data sequence, and output the closing speed and position change curve data; S202: GNN topology modeling: Combine the topology relationship diagram of substation equipment. The GNN model analyzes the electrical connection relationship between the circuit breaker and the disconnector, as well as the association relationship with devices such as buses and transmission lines, and determines the closing state of the circuit breaker through impedance anomalies in the electrical connection; S203: Intermediate state judgment: If the closing position does not reach 95%, it is judged that the circuit breaker is in an intermediate state, and a misoperation risk assessment is carried out.

4. A multi-source intelligent anti-error dispatching method for a substation according to claim 1, characterized in that, The real-time energy topology information includes: when switching between the fan and the lighter, if the output power of the fan is insufficient and the starting time of the lighter does not meet the energy-saving requirements, the starting operation of the lighter is blocked; before the fan is connected to the grid, it is necessary to ensure that its speed is within the range of 90%-110% of the rated speed, and the bus voltage is stable within ±5% of the rated voltage, otherwise the grid connection operation is blocked; the dynamic energy consumption data includes: during the peak electricity price period, if the energy storage system has sufficient power and the photovoltaic power generation is greater than or equal to 50% of the real-time load, the access operation of the municipal power is blocked.

5. A multi-source intelligent anti-error dispatching method for a substation according to claim 1, characterized in that, The input of the deep reinforcement learning model is the current state of the substation, including the closing position of the equipment, the real-time interlocking topology information, the load data, and the energy supply status.

6. A multi-source intelligent anti-error dispatching method for a substation according to claim 5, characterized in that, The output of the deep reinforcement learning model is the Q value of each possible operation: Q(s,a)←Q(s,a)+ α×[R(s,a)+ γ×maxQ(s',a')-Q(s,a)] where Q(s,a) is the current Q value of executing action a in state s; α is the learning rate, 0<α<1; R(s,a) is the immediate reward obtained after executing action a in state s; γ is the discount factor, used to balance future rewards and immediate rewards, 0<γ<1; maxQ(s',a') is the Q value corresponding to the action with the maximum Q value among all actions that can be executed in the new state s' after executing action a.

7. A multi-source intelligent anti-error dispatching method for a substation according to claim 1, characterized in that The real-time warning form includes that the closing position of the circuit breaker is highlighted in red, the uncompleted closing steps are prompted to the operator in the form of animation, and remote warnings are sent through text messages and mobile APP push messages.

8. A multi-source intelligent anti-misoperation dispatching system for a substation, characterized in that, Including: Multi-source data acquisition unit: used for collecting and preprocessing multi-source heterogeneous data, collecting real-time state data of the substation, historical operation records, environmental parameters, video surveillance streams, and grid topology information, and constructing a dynamic knowledge graph; Intermediate state determination module: Based on the long short-term memory network and the graph neural network, construct a device state time series model to real-time identify the intermediate state in the switching operation, and the intermediate state includes the semi-closed and ungrounded device working states in the switching operation; Interlocking logic real-time update module: used for intelligent interlocking logic judgment and update, adopting a hybrid rule engine, combining predefined interlocking logic rules for real-time interlocking logic judgment, where the interlocking logic judgment rules include adjusting the blocking conditions based on real-time energy topology information and dynamic energy consumption data, and optimizing the blocking conditions through real-time topology analysis to reduce redundant blocking; Strategy generation and warning module, used for error prevention decision-making and warning, generating operation suggestions based on the deep reinforcement learning model, real-time warning of high-risk operations, and prompting the operator to correct the steps through the augmented reality interface.

9. An electronic device, characterized in that, Including: One or more processors; Memory; And one or more programs stored in the memory, the one or more programs include instructions for executing the substation multi-source intelligent error prevention scheduling method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, Including one or more programs executed by one or more processors of the power supply electronic device, the one or more programs include instructions for executing the substation multi-source intelligent error prevention scheduling method as described in any one of claims 1-7.

Images