Power distribution room comprehensive safety monitoring alarm system and fault prediction method

Abstract

The invention discloses a power distribution room comprehensive safety monitoring alarm system and a fault prediction method.The alarm system comprises a data acquisition module, a programmable logic controller and a communication module, and the data acquisition module comprises a temperature and humidity sensor, a smoke sensor, a water sensor and a gas monitoring sensor; the programmable logic controller comprises a power supply module, an analog quantity input module, a calculation and storage module, an analog quantity output module, a contactor, an indicating lamp and a local area network, the communication module comprises a power distribution room monitoring center and a wireless communication device, and the data acquisition module, the programmable logic controller and the communication module are connected in sequence. The fault prediction method comprises the following steps of data acquisition and preprocessing, feature extraction, model training and fault prediction and preventive maintenance decision making. The power distribution room safety monitoring mode is comprehensive, the automation degree of alarm is improved, the inspection pressure of operation and maintenance personnel is shared, and emergency safety events can be prevented or handled.

Description

Technical Field

[0001] The present invention belongs to the field of electrical engineering and automation technology, and in particular relates to a comprehensive safety monitoring and alarm system for a power distribution room and a fault prediction method. Background Art

[0002] With the continuous development of the power industry, the safety of the distribution room, as an important part of the power system, has received more and more attention. Traditional distribution room monitoring mainly relies on manual inspections, which has problems such as low efficiency and poor accuracy. In order to improve the safety and reliability of the distribution room, various automatic monitoring and alarm systems have emerged in recent years.

[0003] These systems usually include sensors such as temperature, humidity, and smoke, which are used to monitor the environmental parameters in the distribution room in real time and trigger alarms in abnormal situations. However, existing monitoring and alarm systems often have single functions and lack comprehensive safety assessment capabilities. In existing systems, the layout and selection of sensors are often not reasonable, resulting in insufficient accuracy and real-time monitoring data. For example, some key areas may not be fully covered, or the sensors used may not be sensitive enough to detect potential safety hazards in a timely manner.

[0004] In terms of data processing and analysis, the existing system lacks the ability to conduct in-depth mining and intelligent analysis of complex data, which may lead to false alarms or missed alarms.

[0005] In terms of alarm mechanism, existing systems usually adopt local sound and light alarm, wired local monitoring alarm, wireless network background system alarm or other forms of alarm, either alone or in combination. They have high requirements for human inspection and monitoring, and lack automatic monitoring and instant alarm systems, especially the alarm methods of instant communication mobile terminals such as mobile phones. This makes it impossible for users to achieve full-time, instant and efficient distribution room safety monitoring and emergency response. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a comprehensive safety monitoring and alarm system and a fault prediction method for a distribution room, which can greatly improve the system warning accuracy, shorten the warning response time, and improve the intelligence and automation level of distribution room operation and maintenance.

[0007] The technical solution of the present invention is: a comprehensive safety monitoring and alarm system for a power distribution room, comprising: a data acquisition module, a programmable logic controller, a communication module and an alarm module, the data acquisition module comprising: a temperature and humidity sensor, a smoke sensor, a water immersion sensor, a gas monitoring sensor and a partial discharge detector, the programmable logic controller comprising: a power supply module, an analog input module, a calculation and storage module, an analog output module, a contactor, an indicator light and a local area network, the communication module comprising: a power distribution room monitoring center and a wireless communication device, the alarm module comprising: a text message notification, a telephone notification and an audible and visual alarm; the data acquisition module is connected to the analog input module, the power supply module is respectively connected to the analog input module, the calculation and storage module and the analog output module, the analog input module is connected to the calculation and storage module, the calculation and storage module is connected to the analog output module, the analog output module is respectively connected to the contactor and the indicator light, the analog input module and the analog output module are connected to the wireless communication device through the local area network, the wireless communication device is wirelessly connected to the power distribution room monitoring center and the alarm module, and the power distribution room monitoring center comprises: a power distribution room host, a platform server and a data server.

[0008] Furthermore, the temperature and humidity sensor is a digital sensor and is arranged in a power distribution room.

[0009] Furthermore, the smoke sensor is a photoelectric sensor and is arranged on the top of the power distribution room.

[0010] Furthermore, the water immersion sensor is arranged on the ground or a low-lying area of ​​the power distribution room.

[0011] Furthermore, the wireless communication device uses GPRS, 3G, 4G or 5G communication for data transmission.

[0012] Furthermore, the power distribution room monitoring center is wirelessly connected to the mobile device via a wireless communication device.

[0013] Furthermore, the analog output module is connected to the fire protection system of the power distribution room.

[0014] A fault prediction method for a comprehensive safety monitoring and alarm system of a power distribution room comprises the following steps:

[0015] (1) Data collection and preprocessing, including the following steps:

[0016] S1. Data collection: The following data are collected through the temperature and humidity sensors, smoke sensors, water immersion sensors, and gas monitoring sensors in the system: temperature data, humidity data, smoke concentration data, water immersion data, and gas detection data;

[0017] S2, data cleaning and processing: The data of each sensor is generated into time series data according to the sampling time, the noise signal is eliminated by sliding average, and the outliers are removed by triple standard deviation method;

[0018] (2) Feature extraction: extract key features from time series data, including mean, standard deviation, autocorrelation coefficient and frequency domain features;

[0019] (3) Model training and fault prediction, including the following steps:

[0020] S3. Select LSTM deep learning model to process time series data;

[0021] S4, divide the sensor data into training set and test set, and optimize the model parameters by mean square error;

[0022] S5. Use the trained model to predict future sensor data trends. When the error between the predicted value and the actual value exceeds a certain statistical threshold, the system will issue an early warning to indicate the fault.

[0023] (4) Preventive maintenance decision-making: Based on the model's prediction results, the system automatically generates the following maintenance recommendations:

[0024] Abnormal temperature: Check the heat dissipation of electrical equipment and clean the ventilation equipment;

[0025] Abnormal humidity: Check the moisture-proof measures in the power distribution room to prevent the equipment from getting wet;

[0026] Smoke concentration increases: Check whether the power distribution equipment is overheating and check for fire hazards;

[0027] Flooding: Check the working status of the drainage system to prevent water accumulation from further damaging the equipment;

[0028] (5) Continuous monitoring and optimization: The results of each fault prediction will be stored in the database. The system uses these historical data to continuously optimize the model and retrain the model regularly.

[0029] Furthermore, the expression of the sliding average in step S2 is:

[0030]

[0031] in, represents the data after smoothing, N represents the size of the sliding window, and X(ti) represents the original value of the time series data at time ti;

[0032] The expression of the triple standard deviation method is:

[0033] |X(t)-μ|>3σ,

[0034] Among them, X(t) represents the value of the time series data at time t, μ represents the mean of the time series data, and σ represents the standard deviation of the time series data;

[0035] The expression of the mean is:

[0036]

[0037] Among them, X(t) represents the value of the time series data at time t, and T represents the number of samples;

[0038] The expression of the standard deviation is:

[0039]

[0040] Among them, X(t) represents the value of the time series data at time t, T represents the number of samples, and μ represents the mean of the time series data;

[0041] The expression of the autocorrelation coefficient is:

[0042]

[0043] Among them, ACF(k) represents the autocorrelation coefficient at time lag k, and its range is [-1,1], X(t) represents the value of the time series data at time t, T represents the number of samples, μ represents the mean of the time series data, and k represents the time lag step;

[0044] The expression of the frequency domain feature is:

[0045]

[0046] Among them, X(t) represents the value of the time series data at time t, and T represents the number of samples;

[0047] The expression of the LSTM deep learning model in step S3 is:

[0048] Input Gate:

[0049] i t =σ(W i ·[h t-1 , x t ]+b i ),

[0050] Among them, i t represents the output of the input gate, σ represents the Sigmoid activation function, W i represents the weight matrix of the input gate, h t-1 represents the hidden state at the previous time t-1, x trepresents the input data at the current time t, b i represents the bias term of the input gate;

[0051] Forget Gate:

[0052] f t =σ(W f ·[h t-1 , x t ]+b f ),

[0053] Among them, f t represents the output of the forget gate, σ represents the Sigmoid activation function, W f represents the weight matrix of the forget gate, h t-1 represents the hidden state at the previous time t-1, x t represents the input data at the current time t, b f Represents the bias term of the forget gate;

[0054] Memory unit update:

[0055]

[0056] Among them, C t represents the cell state of the LSTM model at the current time t, f t represents the output of the forget gate, i t represents the output of the input gate, C t-1 Represents the cell state of the LSTM model at the previous time t-1, represents the candidate cell state of the LSTM model at the current time t, h t-1 represents the hidden state at the previous time t-1, x t represents the input data at the current time t, W C The weight matrix representing the candidate cell state, b C A bias term representing the candidate cell state;

[0057] Output Gate:

[0058] o t =σ(W o ·[h t-1 , x t ]+b o ),

[0059] Among them, t represents the output of the output gate, σ represents the Sigmoid activation function, W o represents the weight matrix of the output gate, h t-1 represents the hidden state at the previous time t-1, x t represents the input data at the current time t, bo represents the bias term of the output gate;

[0060] The expression of the mean square error in step S4 is:

[0061]

[0062] Among them, y i represents the true value, represents the predicted value, and n represents the number of samples.

[0063] Beneficial effects of the present invention:

[0064] 1. Improve monitoring accuracy and real-time performance: By optimizing the sensor layout and selection, the present invention significantly improves the accuracy and real-time performance of monitoring data. Specifically, the improvement of sensor sensitivity enables the system to detect potential safety hazards more quickly, thereby reducing the risk of false alarms and missed alarms. Compared with the prior art, the present invention effectively reduces the false alarm rate;

[0065] 2. Enhanced data processing and analysis capabilities: The introduction of advanced data processing and analysis algorithms enables the system to have stronger intelligent early warning functions. By building complex mathematical models and algorithms, the system can conduct in-depth mining and intelligent analysis of the collected data, effectively improving the accuracy and timeliness of early warnings, further improving the accuracy of early warnings, and shortening the early warning response time;

[0066] 3. Implement diversified alarm mechanisms: The designed diversified alarm mechanisms ensure that users can receive alarm information in a timely manner and take effective measures. Whether it is sound and light alarm, SMS notification or telephone voice, it can be customized and adjusted according to user needs. This flexibility not only improves user satisfaction but also enhances the practicality and reliability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention.

[0068] Figure 1 It is a structural schematic diagram of a comprehensive safety monitoring and alarm system for a power distribution room of the present invention.

[0069] Figure 2 It is a schematic diagram of the use status of a comprehensive safety monitoring and alarm system for a power distribution room of the present invention.

[0070] Figure 3 It is a three-dimensional structural schematic diagram of a comprehensive safety monitoring and alarm system for a power distribution room of the present invention.

[0071] In the figure, 1-temperature and humidity sensor, 2-smoke sensor, 3-water sensor, 4-distribution room host. DETAILED DESCRIPTION

[0072] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0073] like Figure 1-2 As shown, a comprehensive safety monitoring and alarm system for a power distribution room includes: a data acquisition module, a programmable logic controller, a communication module and an alarm module. The data acquisition module includes: a temperature and humidity sensor 1, a smoke sensor 2, a water immersion sensor 3, a gas monitoring sensor 4, and a partial discharge detector 5. The programmable logic controller includes: a power supply module, an analog input module, a calculation and storage module (providing a real-time data visualization interface, displaying the current status of various monitoring indicators, storing historical data, facilitating later query and analysis, and identifying the trend of equipment status changes through data analysis tools to provide a basis for subsequent maintenance decisions), an analog output module, a contactor, an indicator light and a local area network. The communication module includes: a power distribution room monitoring center and a wireless communication device. The alarm module includes: SMS notification (when the system detects an abnormal situation, it automatically sends a SMS notification to a preset mobile phone number. The SMS content contains a brief description of the abnormal situation, the time of occurrence and the recommended response measures), telephone notification (when the system detects an abnormal situation, it automatically sends a SMS notification to a preset mobile phone number. The SMS content contains a brief description of the abnormal situation, the time of occurrence and the recommended response measures), telephone notification (when the system detects an abnormal situation, it automatically sends a SMS notification to a preset mobile phone number, and the ... When a serious abnormality or emergency is detected, a preset telephone number is automatically dialed for voice notification, and the voice notification includes a detailed description of the abnormal situation, the time of occurrence, and suggestions for emergency response measures) and an audible and visual alarm (when the system detects an abnormal situation, the audible and visual alarm device on site is immediately triggered to alert on-site personnel through sound and light signals so that the problem can be quickly located and preliminary response measures can be taken); the data acquisition module is connected to the analog input module, the power supply module is respectively connected to the analog input module, the calculation and storage module, and the analog output module, the analog input module is connected to the calculation and storage module, the calculation and storage module is connected to the analog output module, the analog output module is respectively connected to the contactor and the indicator light, the analog input module and the analog output module are connected to the wireless communication device via a local area network, the wireless communication device is wirelessly connected to the distribution room monitoring center and the alarm module, and the distribution room monitoring center includes: a distribution room host 4, a platform server, and a data server.

[0074] Preferably, the temperature and humidity sensor 1 is a digital sensor and is arranged in a power distribution room.

[0075] Preferably, the smoke sensor 2 is a photoelectric sensor and is arranged on the top of the power distribution room.

[0076] Preferably, the water immersion sensor 3 is arranged on the ground or a low-lying area of ​​the power distribution room.

[0077] Preferably, the wireless communication device uses GPRS, 3G, 4G or 5G communication for data transmission.

[0078] Preferably, the power distribution room monitoring center is wirelessly connected to a mobile device (providing a mobile phone APP to support users to monitor the status of the power distribution room anytime and anywhere) via a wireless communication device.

[0079] Preferably, the analog output module is connected to the fire protection system of the power distribution room.

[0080] A fault prediction method for a comprehensive safety monitoring and alarm system of a power distribution room comprises the following steps:

[0081] (1) Data collection and preprocessing, including the following steps:

[0082] S1. Data collection: The following data are collected through the temperature and humidity sensors, smoke sensors, water immersion sensors, and gas monitoring sensors in the system: temperature data, humidity data, smoke concentration data, water immersion data, and gas detection data;

[0083] S2, data cleaning and processing: The data of each sensor is generated into time series data according to the sampling time, the noise signal is eliminated by sliding average, and the outliers are removed by triple standard deviation method;

[0084] (2) Feature extraction: extract key features from time series data, including mean, standard deviation, autocorrelation coefficient and frequency domain features;

[0085] (3) Model training and fault prediction, including the following steps:

[0086] S3. Select LSTM deep learning model to process time series data;

[0087] S4, divide the sensor data into training set and test set, and optimize the model parameters by mean square error;

[0088] S5. Use the trained model to predict the future sensor data trend. When the error between the predicted value and the actual value exceeds a certain statistical threshold (such as the prediction error exceeds a certain standard deviation), the system will issue an early warning to indicate the fault.

[0089] (4) Preventive maintenance decision-making: Based on the model's prediction results, the system automatically generates the following maintenance recommendations:

[0090] Abnormal temperature: Check the heat dissipation of electrical equipment and clean the ventilation equipment;

[0091] Abnormal humidity: Check the moisture-proof measures in the power distribution room to prevent the equipment from getting wet;

[0092] Smoke concentration increases: Check whether the power distribution equipment is overheating and check for fire hazards;

[0093] Flooding: Check the working status of the drainage system to prevent water accumulation from further damaging the equipment;

[0094] (5) Continuous monitoring and optimization: The results of each fault prediction will be stored in the database. The system uses these historical data to continuously optimize the model and ensures that its prediction performance continues to improve by regularly retraining the model.

[0095] Example scenario: If the temperature sensor continuously detects that the temperature in the power distribution room is gradually rising, and combined with the changing trends of humidity and smoke concentration, the system can intelligently predict that there may be a risk of equipment overheating or fire at a certain point in time, issue a warning in advance, and recommend equipment inspection.

[0096] Preferably, the expression of the sliding average in step S2 is:

[0097]

[0098] in, represents the data after smoothing, N represents the size of the sliding window, and X(ti) represents the original value of the time series data at time ti;

[0099] The expression of the triple standard deviation method is:

[0100] |X(t)-μ|>3σ,

[0101] Among them, X(t) represents the value of the time series data at time t, μ represents the mean of the time series data, and σ represents the standard deviation of the time series data;

[0102] The expression of the mean is:

[0103]

[0104] Among them, X(t) represents the value of the time series data at time t, and T represents the number of samples;

[0105] The expression of the standard deviation is:

[0106]

[0107] Among them, X(t) represents the value of the time series data at time t, T represents the number of samples, and μ represents the mean of the time series data;

[0108] The expression of the autocorrelation coefficient is:

[0109]

[0110] Among them, ACF(k) represents the autocorrelation coefficient at time lag k, and its range is [-1,1], X(t) represents the value of the time series data at time t, T represents the number of samples, μ represents the mean of the time series data, and k represents the time lag step;

[0111] The expression of the frequency domain feature is:

[0112]

[0113] Among them, X(t) represents the value of the time series data at time t, and T represents the number of samples;

[0114] The expression of the LSTM deep learning model in step S3 is:

[0115] Input Gate:

[0116] i t =σ(W i ·[h t-1 , x t ]+b i ),

[0117] Among them, i t represents the output of the input gate, σ represents the Sigmoid activation function, W i represents the weight matrix of the input gate, h t-1 represents the hidden state at the previous time t-1, x t represents the input data at the current time t, b i represents the bias term of the input gate;

[0118] Forget Gate:

[0119] f t =σ(W f ·[h t-1 , x t ]+b f ),

[0120] Among them, f t represents the output of the forget gate, σ represents the Sigmoid activation function, W f represents the weight matrix of the forget gate, h t-1 represents the hidden state at the previous time t-1, x t represents the input data at the current time t, b f Represents the bias term of the forget gate;

[0121] Memory unit update:

[0122]

[0123] Among them, C t represents the cell state of the LSTM model at the current time t, f t represents the output of the forget gate, i t represents the output of the input gate, C t-1 Represents the cell state of the LSTM model at the previous time t-1, represents the candidate cell state of the LSTM model at the current time t, h t-1 represents the hidden state at the previous time t-1, x t represents the input data at the current time t, W C The weight matrix representing the candidate cell state, b C A bias term representing the candidate cell state;

[0124] Output Gate:

[0125] o t =σ(W o ·[h t-1 , x t ]+b o ),

[0126] Among them, t represents the output of the output gate, σ represents the Sigmoid activation function, W o represents the weight matrix of the output gate, h t-1 represents the hidden state at the previous time t-1, x t represents the input data at the current time t, b o represents the bias term of the output gate;

[0127] The expression of the mean square error in step S4 is:

[0128]

[0129] Among them, y i represents the true value, represents the predicted value, and n represents the number of samples.

[0130] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A comprehensive safety monitoring and alarm system for a power distribution room, characterized in that: include: A data acquisition module, a programmable logic controller, a communication module and an alarm module, wherein the data acquisition module comprises a temperature and humidity sensor (1), a smoke sensor (2), a water immersion sensor (3), a gas monitoring sensor (4) and a partial discharge detector (5), the programmable logic controller comprises a power supply module, an analog input module, a calculation and storage module, an analog output module, a contactor, an indicator light and a local area network, the communication module comprises a distribution room monitoring center and a wireless communication device, and the alarm module comprises a text message notification, a telephone notification and an audible and visual alarm; the data acquisition module is connected to the analog input module, the power supply module is respectively connected to the analog input module, the calculation and storage module and the analog output module, the analog input module is connected to the calculation and storage module, the calculation and storage module is connected to the analog output module, the analog output module is respectively connected to the contactor and the indicator light, the analog input module and the analog output module are connected to the wireless communication device via the local area network, the wireless communication device is respectively wirelessly connected to the distribution room monitoring center and the alarm module, and the distribution room monitoring center comprises a distribution room host (4), a platform server and a data server.

2. A comprehensive safety monitoring and alarm system for a power distribution room according to claim 1, characterized in that: The temperature and humidity sensor (1) is a digital sensor and is arranged in a power distribution room.

3. A comprehensive safety monitoring and alarm system for a power distribution room according to claim 1, characterized in that: The smoke sensor (2) is a photoelectric sensor and is arranged on the top of the power distribution room.

4. A comprehensive safety monitoring and alarm system for a power distribution room according to claim 1, characterized in that: The water immersion sensor (3) is arranged on the ground or in a low-lying area of ​​the power distribution room.

5. The integrated safety monitoring and alarm system for a power distribution room according to claim 1 is characterized in that: The wireless communication device uses GPRS, 3G, 4G or 5G communication for data transmission.

6. A comprehensive safety monitoring and alarm system for a power distribution room according to claim 1, characterized in that: The power distribution room monitoring center is wirelessly connected to the mobile device via a wireless communication device.

7. The integrated safety monitoring and alarm system for power distribution room according to claim 1 is characterized in that The analog output module is connected to the fire protection system of the power distribution room.

8. A fault prediction method for a comprehensive safety monitoring and alarm system in a power distribution room, characterized in that: The following steps are involved: (1) Data collection and preprocessing, including the following steps: S1. Data collection: The following data are collected through the temperature and humidity sensors, smoke sensors, water immersion sensors, and gas monitoring sensors in the system: temperature data, humidity data, smoke concentration data, water immersion data, and gas detection data; S2, data cleaning and processing: The data of each sensor is generated into time series data according to the sampling time, the noise signal is eliminated by sliding average, and the outliers are removed by triple standard deviation method; (2) Feature extraction: extract key features from time series data, including mean, standard deviation, autocorrelation coefficient and frequency domain features; (3) Model training and fault prediction, including the following steps: S3. Select LSTM deep learning model to process time series data; S4, divide the sensor data into training set and test set, and optimize the model parameters by mean square error; S5. Use the trained model to predict future sensor data trends. When the error between the predicted value and the actual value exceeds a certain statistical threshold, the system will issue an early warning to indicate the fault. (4) Preventive maintenance decision: Based on the prediction results of the model, the system automatically generates the following maintenance recommendations: Abnormal temperature: Check the heat dissipation of the power equipment and clean the ventilation equipment; Abnormal humidity: Check the moisture-proof measures in the power distribution room to prevent the equipment from getting wet; Smoke concentration increases: Check whether the power distribution equipment is overheating and check for fire hazards; Flooding: Check the working status of the drainage system to prevent water accumulation from further damaging the equipment; (5) Continuous monitoring and optimization: The results of each fault prediction will be stored in the database. The system uses these historical data to continuously optimize the model and retrain the model regularly.

9. A fault prediction method for a comprehensive safety monitoring and alarm system of a power distribution room according to claim 8, characterized in that: The expression of the sliding average in step S2 is: in, represents the data after smoothing, N represents the size of the sliding window, and X(ti) represents the original value of the time series data at time ti; The expression of the triple standard deviation method is: |X(t)-μ|>3σ, Among them, X(t) represents the value of the time series data at time t, μ represents the mean of the time series data, and σ represents the standard deviation of the time series data; The expression of the mean is: Among them, X(t) represents the value of the time series data at time t, and T represents the number of samples; The expression of the standard deviation is: Among them, X(t) represents the value of the time series data at time t, T represents the number of samples, and μ represents the mean of the time series data; The expression of the white correlation coefficient is: Among them, ACF(k) represents the correlation coefficient at time lag k, and its range is [-1, 1], X(t) represents the value of the time series data at time t, T represents the number of samples, μ represents the mean of the time series data, and k represents the time lag step; The expression of the frequency domain feature is: Among them, X(t) represents the value of the time series data at time t, and T represents the number of samples; The expression of the LSTM deep learning model in step S3 is: Input Gate: i t =σ(W i ·[h t-1 ,x t ]+b i ), Among them, i t represents the output of the input gate, σ represents the Sigmoid activation function, W i represents the weight matrix of the input gate, h t-1 represents the hidden state at the previous time t-1, x t represents the input data at the current time t, b i represents the bias term of the input gate; Forget Gate: f t =σ(W f ·[h t-1 ,x t ]+b f ), Among them, f t represents the output of the forget gate, σ represents the Sigmoid activation function, W f represents the weight matrix of the forget gate, h t-1 represents the hidden state at the previous time t-1, x t represents the input data at the current time t, b f Represents the bias term of the forget gate; Memory unit update: Among them, C t represents the cell state of the LSTM model at the current time t, f t represents the output of the forget gate, i t represents the output of the input gate, C t-1 Represents the cell state of the LSTM model at the previous time t-1, represents the candidate cell state of the LSTM model at the current time t, h t-1 represents the hidden state at the previous time t-1, x t represents the input data at the current time t, W C The weight matrix representing the candidate cell state, b C A bias term representing the candidate cell state; Output Gate: The t =σ(W o ·[h t-1 ,x t ]+b o ), Among them, t represents the output of the output gate, σ represents the Sigmoid activation function, W o represents the weight matrix of the output gate, h t-1 represents the hidden state at the previous time t-1, x t represents the input data at the current time t, b o Represents the bias term of the output gate; The expression of the mean square error in step S4 is: Among them, y i represents the true value, represents the predicted value, and n represents the number of samples.

Images