Dynamic protection method and device for building electrical system

Abstract

The invention relates to a building electrical system dynamic protection method and device, and the method comprises the steps: collecting the real-time operation data of a power supply line, the real-time operation data comprising a three-phase current value, a cable surface temperature, environment humidity and a voltage fluctuation ratio; inputting the real-time operation data into a fault prediction model, calculating a health index of the current line according to the input real-time operation data by using the fault prediction model, and further calculating and outputting a fault probability according to the health index; according to the fault probability, an overload protection threshold value of the intelligent switch in the power supply line is dynamically adjusted according to a certain adjusting rule, and the adjusting rule meets the condition that the larger the fault probability is, the smaller the overload protection threshold value is. According to the invention, by predicting the fault probability and dynamically adjusting the overload protection threshold value of the intelligent switch based on the fault probability, a one-step passive protection strategy is avoided, and the safety and the power supply continuity are balanced.

Description

Technical Field

[0001] The present invention relates to the field of building electrical safety and intelligent technology, and in particular to a dynamic protection method and device for a building electrical system. Background Art

[0002] The existing protection method for building electrical systems is usually passive protection, that is, a fixed overload protection threshold is set for the circuit breaker in advance. When the current of the power supply line exceeds the overload protection threshold, the circuit breaker will automatically cut off the circuit (for details, please refer to the published patent CN110112728A).

[0003] The aforementioned passive protection method has the following drawbacks: The fixed threshold is rigid. According to GB 50054-2011, the circuit breaker's overload protection threshold reserves a 20% margin on the rated current, or 120% of the rated current. However, this approach fails to account for actual load fluctuations and equipment aging. It only disconnects the circuit after a line overload or short circuit occurs, failing to provide early warning of hidden faults caused by line aging or increased contact resistance. Statistics show that 28.7% of electrical fires in my country in 2022 were caused by deterioration of line insulation (source: China Fire Protection Yearbook). Summary of the Invention

[0004] The present invention provides a dynamic protection method and device for a building electrical system. A fault prediction model is used to calculate a health index based on the actual operating data of the power supply line. The health index is then used to quantify the degree of line aging to predict the probability of failure. The overload protection threshold of the intelligent switch is dynamically adjusted based on the probability of failure, avoiding a "one-size-fits-all" passive protection strategy and achieving a balance between safety and power supply continuity.

[0005] The present invention is implemented through the following scheme: a dynamic protection method for a building electrical system, wherein the building electrical system includes a power supply line and an intelligent switch provided on the power supply line, wherein the intelligent switch can automatically cut off the power supply line when the current of the power supply line exceeds its own overload protection threshold. The method comprises the following steps:

[0006] S1. Collecting real-time operating data of the power supply line, wherein the real-time operating data includes three-phase current value, cable surface temperature, ambient humidity and voltage fluctuation rate;

[0007] S2. Inputting the real-time operation data into a fault prediction model, using the fault prediction model to calculate a health index of the current line based on the input real-time operation data, and further calculating and outputting a fault probability based on the health index;

[0008] S3. Dynamically adjust the overload protection threshold of the intelligent switch according to a certain adjustment rule based on the fault probability, wherein the adjustment rule satisfies: the greater the fault probability, the smaller the overload protection threshold.

[0009] A further improvement of the dynamic protection method for building electrical systems of the present invention is that the adjustment rule is:

[0010]

[0011] Among them: I threshold It represents the overload protection threshold of the intelligent switch; I rated Expressed as the rated current value of the single-phase circuit; P f Expressed as the probability of failure.

[0012] The further improvement of the dynamic protection method of the building electrical system of the present invention is that: the building electrical system also includes a backup power supply line; before executing step S3, the fault probability P is first determined. f Is it greater than or equal to 20%?

[0013] If not, proceed to step S3;

[0014] If so, the current power supply line is identified as a fault circuit, and the following three-level response is triggered: cutting off the current fault circuit; turning on the backup power supply line to replace the fault circuit for power supply; and feeding back fault location information.

[0015] The further improvement of the dynamic protection method of the building electrical system of the present invention is:

[0016] The fault prediction model uses formula (1) to calculate the health index HI:

[0017]

[0018] Wherein: α, β, γ, δ represent the weight coefficients obtained by training the fault prediction model; I rms It is represented by the maximum phase current effective value among the three-phase current values; I rated It represents the rated current value of the single-phase line; T represents the cable surface temperature; T safe It represents the safety temperature threshold; ΔV represents the voltage fluctuation rate; H represents the ambient humidity; a represents the parameter affecting the ambient humidity (it is a known constant and its value varies depending on the application environment).

[0019] The fault prediction model uses formula (2) to calculate the fault probability P f :

[0020] P f =1-e -k(1 / HI) (2)

[0021] Wherein: HI represents the health index; k represents the attenuation coefficient (a known constant whose value varies depending on the application environment); e is a natural constant (also known as the Napier number).

[0022] A further improvement of the dynamic protection method for building electrical systems of the present invention is that the method for obtaining the weight coefficient by training the fault prediction model is:

[0023] Construct a data set containing historical fault data, and make each historical fault data set include a three-phase current value data set collected in time series, a cable surface temperature change data set, and an exposure value to ambient humidity;

[0024] The fault prediction model is iteratively trained based on the data set, and the weight coefficients are continuously adjusted during the iterative training process.

[0025] A further improvement of the dynamic protection method for building electrical systems of the present invention lies in the method of adjusting the weight coefficient: a bidirectional LSTM network is used to extract time series features from the data set, and based on the time series features, the attention mechanism is used to locate the critical fault symptom period, and the weight coefficient corresponding to the critical fault symptom period is increased.

[0026] A further improvement of the dynamic protection method for building electrical systems of the present invention is that the fault prediction model is adjusted by a trained fault prediction model suitable for industrial equipment. The adjustment method is: freezing the bottom network and fine-tuning the top fully connected layer.

[0027] The present invention also provides a building electrical system dynamic protection device for implementing the above-mentioned building electrical system dynamic protection method, the dynamic protection device comprising:

[0028] Multi-parameter sensing unit, used to collect real-time operating data of the power supply line;

[0029] An edge computing unit, configured to locally deploy a fault prediction model to calculate a health index and a fault probability, the edge computing unit being connected to the multi-parameter sensing unit;

[0030] A dynamic threshold control unit is used to dynamically adjust the overload protection threshold of the intelligent switch according to a certain adjustment rule based on the fault probability. The dynamic threshold control unit is connected to the edge computing unit and the intelligent switch.

[0031] The further improvement of the dynamic protection device for building electrical systems of the present invention is as follows:

[0032] The building electrical system further includes a backup power supply line, and the intelligent switch is a power supply switch, which can selectively switch on and off the power supply line and the backup power supply line based on the triggering of the dynamic threshold control unit;

[0033] When the fault probability is greater than or equal to 20%, the dynamic threshold control unit determines that the current power supply line is a fault circuit and triggers the power switching switch to perform the following three-level response: cutting off the current fault circuit; turning on the backup power supply line to replace the fault circuit for power supply; and feeding back fault location information.

[0034] A further improvement of the dynamic protection device for building electrical systems of the present invention is that: the number of the edge computing units is multiple; the dynamic protection device also includes a cloud-based collaborative module, which is used to receive desensitized data uploaded by multiple edge computing units and update global model parameters through federated learning, so as to dynamically optimize the fault prediction model deployed at each edge computing unit according to the building type, and all the edge computing units are connected to the cloud-based collaborative module.

[0035] The present invention includes but is not limited to the following beneficial effects:

[0036] A fault prediction model calculates a health index based on actual power line operating data. This health index then quantifies line aging to predict the probability of failure. This dynamically adjusts the overload protection threshold of the smart switch based on the probability of failure, avoiding a one-size-fits-all, passive protection strategy. By reducing the load factor rather than directly shutting off power when a potential fault is predicted, this approach avoids unplanned downtime in high-reliability locations and balances safety with power continuity.

[0037] By setting up the backup power supply line and the power switching device, when the power supply line is determined to be a faulty line, the backup power supply line can be used to replace the faulty line to supply power, so as to ensure the normal operation of the electrical equipment.

[0038] The cloud-based collaborative module is used to achieve iterative optimization of multi-building data aggregation and fault prediction models. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A flow chart of the dynamic protection method for a building electrical system of the present invention is shown.

[0040] Figure 2 The schematic diagram of the dynamic protection device for building electrical systems of the present invention is shown.

[0041] Figure 3 The data architecture diagram of the dynamic protection device for building electrical systems of the present invention is shown. DETAILED DESCRIPTION

[0042] In order to solve the problems of traditional passive protection methods that only cut off the circuit after line overload or short circuit occurs and cannot warn of hidden faults, the present invention provides a dynamic protection method and device for building electrical systems. The health index is calculated according to the actual operating data of the power supply line through a fault prediction model, and the degree of line aging is further quantified based on the health index to predict the probability of failure. The overload protection threshold of the intelligent switch is dynamically adjusted according to the fault probability, avoiding a "one-size-fits-all" passive protection strategy and balancing safety and power supply continuity.

[0043] The following is a further description of the dynamic protection method and device for the building electrical system with reference to specific embodiments and accompanying drawings.

[0044] See Figure 1 As shown, a dynamic protection method for a building electrical system includes a power supply line and an intelligent switch provided on the power supply line. The intelligent switch can automatically cut off the power supply line when the current of the power supply line exceeds its own overload protection threshold. The method includes:

[0045] Step S1: collecting real-time operation data of the power supply line, where the real-time operation data includes three-phase current values, cable surface temperature, ambient humidity, and voltage fluctuation rate.

[0046] Step S2: input the real-time operation data into a fault prediction model, use the fault prediction model to calculate the health index of the current line based on the input real-time operation data, and further calculate and output the fault probability based on the health index.

[0047] In some embodiments, the fault prediction model uses formula (1) to calculate the health index HI:

[0048]

[0049] Where: α, β, γ, δ represent the weight coefficients obtained by training the fault prediction model; I rms Expressed as the three-phase current value (I a , I b , I c ) in the maximum phase current effective value; I rated It represents the rated current value of the single-phase line; T represents the cable surface temperature; T safe It represents the safety temperature threshold; ΔV represents the voltage fluctuation rate; H represents the ambient humidity; a represents the parameter affecting the ambient humidity (it is a known constant and its value varies depending on the application environment. For example, for public buildings, a = 0.15; for residential buildings, a = 0.08).

[0050] The fault prediction model uses formula (2) to calculate the fault probability P f:

[0051] P f =1-e -k(1 / HI) (2)

[0052] Where HI is the health index; k is the attenuation coefficient (a known constant whose value varies depending on the application environment. For example, for residential distribution boxes, k = 0.12; for data center UPS, k = 0.25; for industrial motors, k = 0.30); and e is a natural constant (also known as the Napier number, e ≈ 2.71828).

[0053] It can be seen from formula (2) that the failure probability P f The prediction is based on the health index HI to quantify the degree of line aging. In other words, the failure probability P f The size of reflects the aging degree of the line to a certain extent.

[0054] Step S3: According to the failure probability P f The overload protection threshold of the intelligent switch is dynamically adjusted according to a certain adjustment rule, and the adjustment rule satisfies: the greater the fault probability, the smaller the overload protection threshold.

[0055] In some embodiments, the adjustment rule is:

[0056]

[0057] Among them: I threshold It represents the overload protection threshold of the intelligent switch; I rated Expressed as the rated current value of the single-phase circuit; P f Expressed as the probability of failure.

[0058] The adjustment rule sets the failure probability P f The size range is divided into three range intervals, and different range intervals correspond to different overload protection thresholds I threshold , and satisfy the failure probability P f The larger the value, the higher the overload protection threshold I threshold The smaller the overload protection threshold I threshold Not greater than the rated current value I of the single-phase circuit rated To meet the needs of potential faults such as line aging, the load rate can be reduced by dynamic threshold adjustment instead of directly cutting off the power supply line, which can effectively avoid unexpected downtime in high-reliability places such as data centers / hospitals.

[0059] In some preferred embodiments, the method for obtaining the weight coefficients (α, β, γ, δ) by training the fault prediction model is as follows: constructing a data set containing historical fault data, and making each historical fault data include a three-phase current value data set collected in time series (sampling frequency ≥ 1kHz), a cable surface temperature change data set (i.e., temperature change gradient ΔT / Δt), and an ambient humidity exposure (HE). The fault prediction model is iteratively trained based on the data set, and the weight coefficients (α, β, γ, δ) are continuously adjusted during the iterative training process. It should be noted that since the voltage fluctuation rate ΔV has a Pearson correlation coefficient < 0.05, the voltage fluctuation rate ΔV parameter is excluded from the above-mentioned data set used for training, and is instead analyzed and verified based on its importance.

[0060] Preferably, the above-mentioned data set also includes a fault type label. The method for adjusting the weight coefficient is: using a bidirectional LSTM network to extract the time series features of the data set, and locating the key fault symptom period through the attention mechanism based on the time series features, and increasing the weight coefficient corresponding to the key fault symptom period. Among them, the key symptom period can be located by the gradient weighted class activation map (Grad-CAM), for example: the window of 200ms before and after the current surge, the temperature acceleration stage (gradient ≥ 2℃ / min) is given a higher weight. In this way, the weight coefficient can be adjusted in a targeted manner, and the output fault probability can also reflect the fault type. When training the fault prediction model, it is helpful to better optimize the fault prediction model and obtain a more accurate weight coefficient.

[0061] Because fault prediction models for industrial equipment are relatively mature, a transfer learning strategy can be used to simplify the training of fault prediction models for building electrical systems. This strategy uses a previously trained fault prediction model for industrial equipment and adapts its parameters to the building electrical scenario. Specifically, the trained model's underlying network is frozen and the top fully connected layer is fine-tuned. This approach only requires fine-tuning the top fully connected layer, reducing the required training data by 70%.

[0062] The present invention also provides a building electrical system dynamic protection device for implementing the above-mentioned building electrical system dynamic protection method, see Figure 2 As shown, the dynamic protection device includes a multi-parameter sensing unit 1, an edge computing unit 2, and a dynamic threshold control unit 3. The building electrical system includes a power supply line 5 and an intelligent switch 6 arranged on the power supply line 5, and the power supply line 5 and the intelligent switch 6 are integrated and assembled in a distribution box 7.

[0063] The multi-parameter sensing unit 1 is used to collect real-time operating data of the power supply line 5 and is installed at the distribution box 7. Specifically, the multi-parameter sensing unit 1 includes a current sensor for collecting three-phase current values, an infrared thermal imager for measuring the surface temperature of the cable, and a temperature and humidity sensor for measuring ambient humidity. In this embodiment, the current sensor is a high-precision closed-loop Hall current sensor, model CHB-50S, installed at each phase line, with a range of 0 to 100A and an accuracy of ±0.5%. The infrared thermal imager is a non-contact infrared thermal imager, model FLIR T540, with a resolution of 320×240 and a temperature measurement range of -20°C to +150°C. The temperature and humidity sensor is a digital temperature and humidity sensor, model SHT35, with a humidity measurement error of ±1.5%RH.

[0064] The edge computing unit 2 is used to locally deploy a fault prediction model to calculate the health index and the probability of failure. The edge computing unit 2 is connected to the multi-parameter sensing unit 1. Specifically, the edge computing unit 2 includes a core processor and a model framework, and is equipped with at least 64GB of local storage space of eMMC to store historical data cache. In this embodiment, the core processor is NVIDIA Jetson Nano (an NVIDIA embedded system), which is equipped with a GPU acceleration module with a CUDA (i.e., a general-purpose parallel computing platform and programming model) core. The model framework is an LSTM network optimized by TensorRT (i.e., a high-performance deep learning inference engine), and the model size is compressed to less than 35MB.

[0065] The dynamic threshold control unit 3 is used to dynamically adjust the overload protection threshold of the intelligent switch according to a certain adjustment rule based on the failure probability. The dynamic threshold control unit 3 is connected to the edge computing unit 2 and the intelligent switch. Specifically, the dynamic threshold control unit 3 includes a programmable logic controller PLC. In this embodiment, the programmable logic controller PLC is Siemens' S7-1200, which supports the Modbus RTU protocol. Correspondingly, the intelligent switch 6 in the building electrical system is an adaptive circuit breaker with remotely adjustable rated current, such as Schneider's Acti9 series, whose rated current can be remotely adjusted in the range of 16A to 63A. By connecting the dynamic threshold control unit 3 to the intelligent switch 6, remote adjustment of the overload protection threshold of the intelligent switch 6 can be achieved. The adjustment rule for the remote adjustment is:

[0066]

[0067] Among them: I threshold It is represented as the overload protection threshold of the intelligent switch 6; rated Expressed as the rated current value of the single-phase circuit; P f Expressed as the probability of failure.

[0068] Specifically: When the failure probability P output by the fault prediction model f When it is less than 5%, the dynamic threshold control unit 3 automatically sets the overload protection threshold I of the intelligent switch 6 to threshold Adjust to the rated current value I of the unidirectional line rated Equal; when the failure probability P output by the fault prediction model f When the overload protection threshold value I of the intelligent switch 6 is less than 15% and not less than 5%, the dynamic threshold control unit 3 automatically increases the overload protection threshold value I of the intelligent switch 6. threshold Reduce to the rated current value I for unidirectional lines rated 0.9 times; and when the fault probability P output by the fault prediction model f When the overload protection threshold value is not less than 15%, the dynamic threshold control unit 3 automatically increases the overload protection threshold value I of the intelligent switch 6. threshold Reduce to the rated current value I for unidirectional lines rated 0.7 times of the load. As can be seen from the above adjustment rules, the greater the failure probability, the lower the overload protection threshold. Using these dynamic adjustment rules, when a power line has potential faults such as aging, the load factor is automatically reduced rather than directly shutting off the power line. This can effectively avoid unexpected downtime in high-reliability locations such as data centers and hospitals.

[0069] Preferably, to ensure continuous power supply in high-reliability locations, building electrical systems are typically equipped with a backup power supply so that, in the event of a power line failure, the system can switch to the backup power supply line to provide power. In this case, the intelligent switch 5 typically employs a power transfer switch, which is essentially composed of two adaptive circuit breakers and a switching module. The two adaptive circuit breakers can be remotely adjusted for overload protection thresholds by the dynamic threshold control unit 3, and the switching module can be triggered by the dynamic threshold control unit 3 to selectively turn on or off the two adaptive circuit breakers, thereby switching the backup power supply. In this embodiment, the switching module is based on a silicon carbide (SiC) MOSFET fast switching circuit, and the switching time is <10ms.

[0070] Specific as Figure 2 As shown, the building electrical system has two power supply lines 5, one of which serves as a backup power supply line and is connected to a backup power source. When the fault prediction model predicts a power supply line failure probability of 20% or greater, the dynamic threshold control unit 3 identifies the current power supply line as a faulty circuit and automatically triggers the power supply switch to execute the following three-level response: disconnecting the current faulty circuit; connecting the backup power supply line to replace the faulty circuit; and providing feedback on fault location information.

[0071] Through the dynamic threshold adjustment and active triggering of the three-level response described above, when the power supply line has a relatively low potential fault (e.g., a failure probability of less than 20%), the overload protection threshold can be actively lowered to protect the power supply line, rather than directly cutting off the power line, thus avoiding unexpected downtime in high-reliability locations. When the power line has a relatively high potential fault (e.g., a failure probability of not less than 20%), the fault circuit is actively cut off, rather than cutting off after the fault occurs, changing passive protection to active protection. At the same time, the backup power supply line can be automatically switched to ensure continuous power supply in high-reliability locations.

[0072] In some preferred embodiments, see Figure 2 and Figure 3 As shown: the number of edge computing units 2 is multiple, and they are deployed in different power supply locations. The dynamic protection device also includes a cloud collaboration module 4, which is used to receive the desensitized data uploaded by multiple edge computing units 2 (i.e., delete the user identification and Pan-China geographical location), and update the global model parameters through federated learning to dynamically issue differentiated protection strategy templates according to the building type (such as residential, commercial, and industrial) to optimize the fault prediction model deployed at each edge computing unit. All edge computing units 2 are connected to the cloud collaboration module 4, and the corresponding system architecture diagram is shown as follows: Figure 3 As shown in the figure, the multi-parameter sensing unit 1 belongs to the sensing layer, the edge computing unit 2 belongs to the edge layer, the dynamic threshold control unit 3 and related components in the building electrical system (such as smart switches 6) belong to the execution layer, and the cloud collaboration module 4 belongs to the cloud. The sensing layer uploads real-time operating data such as collected current to the edge layer, which issues threshold instructions, which are executed by the execution layer. Simultaneously, the edge layer also uploads data to the cloud, which uniformly issues differentiated protection strategy templates to optimize various fault prediction models.

[0073] The present invention has been described in detail above with reference to the embodiments of the accompanying drawings. A person skilled in the art can make various modifications to the present invention based on the above description. Therefore, certain details in the embodiments should not be construed as limiting the present invention. The scope of protection of the present invention shall be determined by the scope defined in the appended claims.

Claims

1. A dynamic protection method for a building electrical system, wherein the building electrical system includes a power supply line and an intelligent switch provided on the power supply line, wherein the intelligent switch can automatically cut off the power supply line when the current of the power supply line exceeds its own overload protection threshold, characterized in that: The method comprises the steps of: S1. Collecting real-time operating data of the power supply line, wherein the real-time operating data includes three-phase current value, cable surface temperature, ambient humidity and voltage fluctuation rate; S2. Inputting the real-time operation data into a fault prediction model, using the fault prediction model to calculate a health index of the current line based on the input real-time operation data, and further calculating and outputting a fault probability based on the health index; S3. Dynamically adjust the overload protection threshold of the intelligent switch according to a certain adjustment rule based on the fault probability, wherein the adjustment rule satisfies: the greater the fault probability, the smaller the overload protection threshold.

2. The dynamic protection method for building electrical systems according to claim 1, wherein: The adjustment rules are: Among them: I threshold It represents the overload protection threshold of the intelligent switch; I rated Expressed as the rated current value of the single-phase circuit; P f Expressed as the probability of failure.

3. The dynamic protection method for a building electrical system according to claim 1, wherein: The building electrical system also includes a backup power supply line; before executing step S3, first determine the fault probability P f Is it greater than or equal to 20%? If not, proceed to step S3; If so, the current power supply line is identified as a fault circuit, and the following three-level response is triggered: cutting off the current fault circuit; turning on the backup power supply line to replace the fault circuit for power supply; and feeding back fault location information.

4. The dynamic protection method for a building electrical system according to claim 1, wherein: The fault prediction model uses formula (1) to calculate the health index HI: Wherein: α, β, γ, δ represent the weight coefficients obtained by training the fault prediction model; I rms It is represented by the maximum phase current effective value among the three-phase current values; I rated It represents the rated current value of the single-phase line; T represents the cable surface temperature; T safe It represents the safety temperature threshold; ΔV represents the voltage fluctuation rate; H represents the ambient humidity; a represents the parameter affecting the ambient humidity (it is a known constant and its value varies depending on the application environment). The fault prediction model uses formula (2) to calculate the fault probability P f : P f =1-e -k(1 / HI) (2) Wherein: HI represents the health index; k represents the attenuation coefficient (a known constant whose value varies depending on the application environment); e is a natural constant (also known as the Napier number).

5. The dynamic protection method for building electrical systems according to claim 4, characterized in that: The method for obtaining the weight coefficient by training the fault prediction model is: Construct a data set containing historical fault data, and make each historical fault data set include a three-phase current value data set collected in time series, a cable surface temperature change data set, and an exposure value to ambient humidity; The fault prediction model is iteratively trained based on the data set, and the weight coefficients are continuously adjusted during the iterative training process.

6. The dynamic protection method for building electrical systems according to claim 5, characterized in that: The method for adjusting the weight coefficient is: using a bidirectional LSTM network to extract time series features from the data set, and locating the key fault symptom period through the attention mechanism based on the time series features, and increasing the weight coefficient corresponding to the key fault symptom period.

7. The dynamic protection method for building electrical systems according to claim 1, wherein: The fault prediction model is adjusted by a trained fault prediction model suitable for industrial equipment. The adjustment method is: freezing the bottom layer network and fine-tuning the top fully connected layer.

8. A dynamic protection device for a building electrical system, characterized in that: For implementing the building electrical system dynamic protection method according to any one of claims 1 to 7, the dynamic protection device comprises: Multi-parameter sensing unit, used to collect real-time operating data of the power supply line; An edge computing unit, configured to locally deploy a fault prediction model to calculate a health index and a fault probability, the edge computing unit being connected to the multi-parameter sensing unit; A dynamic threshold control unit is used to dynamically adjust the overload protection threshold of the intelligent switch according to a certain adjustment rule based on the fault probability. The dynamic threshold control unit is connected to the edge computing unit and the intelligent switch.

9. The dynamic protection device for a building electrical system according to claim 8, characterized in that: The building electrical system further includes a backup power supply line, and the intelligent switch is a power supply switch, which can selectively switch on and off the power supply line and the backup power supply line based on the triggering of the dynamic threshold control unit; When the fault probability is greater than or equal to 20%, the dynamic threshold control unit determines that the current power supply line is a fault circuit and triggers the power switching switch to perform the following three-level response: cutting off the current fault circuit; turning on the backup power supply line to replace the fault circuit for power supply; and feeding back fault location information.

10. The dynamic protection device for building electrical systems according to claim 8, characterized in that: There are multiple edge computing units; the dynamic protection device also includes a cloud-based collaborative module, which is used to receive desensitized data uploaded by multiple edge computing units and update global model parameters through federated learning, so as to dynamically optimize the fault prediction model deployed at each edge computing unit according to the building type, and all the edge computing units are connected to the cloud-based collaborative module.

Images