Method and system for processing equipment performance statistical data

Abstract

The invention relates to the field of data processing, and particularly discloses an equipment performance statistical data processing method and system, and the method comprises the steps: obtaining historical performance statistical data, and extracting the performance characteristics of the historical performance statistical data; constructing a performance prediction model according to the performance characteristics and the LSTM neural network; acquiring real-time performance statistical data of the equipment, and substituting the real-time performance statistical data into the performance prediction model to obtain a working state of the equipment; and searching a corresponding coping strategy according to the working state of the equipment and executing the coping strategy. According to the method, feature extraction is performed on the historical performance statistical data of the ventilation equipment, then the prediction model is constructed by using the specific LSTM neural network, and then the real-time performance statistical data is substituted into the prediction model, so that whether the working state of the ventilation equipment is normal or not can be obtained, and therefore, the working state of the ventilation equipment can be monitored in real time; and the normal work of the ventilation equipment can be ensured without staring at the monitoring screen all the time.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a method and system for processing equipment performance statistical data. Background Art

[0002] Buildings are artificial environments created by people to meet the needs of social life, using the material and technical means they have mastered, and applying certain scientific laws and aesthetic principles. Many equipment are needed in the process of building a building, and ventilation equipment is one of the most common ones.

[0003] In order to ensure that the air quality in the building meets the requirements, ventilation equipment must be used for air exchange. The ventilation equipment needs to be in normal condition. The current monitoring method for ventilation equipment is to monitor it using image monitoring on the server. Although monitoring can be achieved, the manager needs to keep an eye on the monitoring screen, which is labor-intensive and easy to miss ventilation equipment with problems. Summary of the invention

[0004] The object of the present invention is to provide a method and system for processing equipment performance statistical data to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: A method for processing device performance statistical data, the method comprising: Obtain historical performance statistics, extract performance characteristics of historical performance statistics, and the curve formula for extracting performance characteristics is: , where W is the energy consumption of the ventilation equipment, m is the mass of the ventilation equipment, p is the power of the ventilation equipment, and t is the time. Then the curve is filtered with the SG filter. The construction of the SG filter is as follows: , p(i+j) is the original input signal, y(i) is the resulting output signal, F j is the coefficient given by the SG filter, and N is the number of convolution integers; A performance prediction model is constructed based on performance characteristics and LSTM neural network (Long Short-Term Memory Neural Network). Specifically, the forget gate is used to determine which performance characteristics in the unit are discarded and retained, determine the information stored in the unit, and integrate the data from the forget gate and input gate to obtain the performance prediction model. Obtain real-time performance statistics of the equipment, substitute the real-time performance statistics into the performance prediction model, and obtain the working status of the equipment; Find the corresponding response strategy and execute it according to the working status of the equipment.

[0006] As a further solution of the present invention: after extracting the performance characteristics of the historical performance statistical data, the method further includes: The performance characteristics are normalized, and the calculation formula for normalization is as follows: , is the normalized value, is the minimum value in the sequence, is the maximum value in the sequence, and S is the original data.

[0007] As a further solution of the present invention: the calculation formula of the forget gate is as follows: ,in is the output of the forget gate, is the weight matrix, is the bias vector, is the hidden state of the neuron at the previous time step, is the input of the LSTM neural network at the current time step, is the ReLU function; the information stored in the unit is determined by updating the information in the unit through the input gate and obtaining new information using the cosh layer. The calculation formula for updating the information in the unit through the input gate is as follows: ,The calculation formula for obtaining new information using the cosh layer is as follows: ,in and is the weight matrix, and is the bias vector, is the output of the input gate at the current time step, is the newly generated memory state of the neuron, is the hyperbolic cosine function.

[0008] As a further solution of the present invention: the step of integrating the data from the forget gate and the input gate includes: converting the old state c j-1 Multiply by the output of the forget gate f j , represents the data partial discarding process, and then adds the output i of the input gate j Multiply by the new memory state g of the neuron j , change the memory state of the neuron from c j-1 Updated to c j , c j It will be passed to the next neuron, and the specific calculation process is: , the output gate is based on the unit state c j 、The output of the previous cycle is h j-1 and power input x j Generate a new hidden state h j-1 , the calculation formula of the output gate is: , ,in and are the weight matrix and bias vector respectively, is the output of the output gate at the current time step, is the hidden state of the neuron at the current time step and is also the output of the current time step. Will make peace Pass it on to the next neuron.

[0009] As a further solution of the present invention: the step of substituting the real-time performance statistical data into the performance prediction model to obtain the working status of the device includes: Substitute the real-time performance statistics into the performance prediction model to obtain the output result of the output gate; The output result is compared with the preset rules to determine whether the working status of the equipment is normal, abnormal or dangerous.

[0010] The technical solution of the present invention also provides a system for processing equipment performance statistical data, the system comprising: The extraction module is used to obtain historical performance statistics and extract the performance characteristics of historical performance statistics. The curve formula for extracting performance characteristics is: , where W is the energy consumption of the ventilation equipment, m is the mass of the ventilation equipment, p is the power of the ventilation equipment, and t is the time. Then the curve is filtered with the SG filter. The construction of the SG filter is as follows: , p(i+j) is the original input signal, y(i) is the resulting output signal, F j is the coefficient given by the SG filter, and N is the number of convolution integers; The model construction module is used to build a performance prediction model based on performance characteristics and LSTM neural network (Long Short-Term Memory Neural Network). Specifically, the forget gate is used to determine which performance characteristics in the unit are discarded and retained, determine the information stored in the unit, and integrate the data from the forget gate and input gate to obtain the performance prediction model. A processing module is used to obtain real-time performance statistics of the equipment, substitute the real-time performance statistics into the performance prediction model, and obtain the working status of the equipment; The execution module is used to find and execute corresponding response strategies according to the working status of the equipment.

[0011] As a further solution of the present invention: the processing system for device performance statistical data further includes: The normalization module is used to normalize the performance characteristics. The calculation formula for normalization is as follows: , is the normalized value, is the minimum value in the sequence, is the maximum value in the sequence, and S is the original data.

[0012] As a further solution of the present invention: the model construction module includes: The screening unit is used to determine which performance features in the unit are discarded and retained using the forget gate. The calculation formula of the forget gate is as follows: ,in is the output of the forget gate, is the weight matrix, is the bias vector, is the hidden state of the neuron at the previous time step, is the input of the LSTM neural network at the current time step, is the ReLU function; The storage unit is used to determine the information stored in the unit. The information stored in the unit includes updating the information in the unit through the input gate and obtaining new information using the cosh layer. The calculation formula for updating the information in the unit through the input gate is as follows: ,The calculation formula for obtaining new information using the cosh layer is as follows: ,in and is the weight matrix, and is the bias vector, is the output of the input gate at the current time step, is the newly generated memory state of the neuron, is the hyperbolic cosine function; The output unit is used to integrate the data from the forget gate and the input gate, including: j-1 Multiply by the output of the forget gate f j , represents the data partial discarding process, and then adds the output i of the input gate j Multiply by the new memory state g of the neuron j , change the memory state of the neuron from c j-1 Updated to c j , c j It will be passed to the next neuron, and the specific calculation process is: , the output gate is based on the unit state c j 、The output of the previous cycle is h j-1 and power input x j Generate a new hidden state h j-1 , the calculation formula of the output gate is: , ,in and are the weight matrix and bias vector respectively, is the output of the output gate at the current time step, is the hidden state of the neuron at the current time step and is also the output of the current time step. Will make peace Pass it on to the next neuron.

[0013] As a further solution of the present invention: the execution module includes: A substitution unit, used to substitute the real-time performance statistical data into the performance prediction model to obtain the output result of the output gate; The execution unit is used to compare the output result with the preset rules to determine whether the working state of the equipment is normal, abnormal or dangerous.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention first performs feature extraction on the historical performance statistical data of the ventilation equipment, then uses a specific LSTM neural network to build a prediction model, and then substitutes the real-time performance statistical data into the prediction model, so as to obtain whether the working status of the ventilation equipment is normal, thereby being able to monitor the working status of the ventilation equipment in real time, and ensuring that the ventilation equipment can work normally without having to stare at the monitoring screen all the time. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention.

[0016] Figure 1 The flowchart is a method for processing device performance statistical data.

[0017] Figure 2 This is a second sub-flow chart of the method for processing device performance statistical data.

[0018] Figure 3 This is a third sub-flow chart of the method for processing device performance statistical data.

[0019] Figure 4 This is a structural block diagram of the system for processing equipment performance statistics.

[0020] Figure 5 This is a structural block diagram of the components of the model construction module in the system for processing equipment performance statistical data.

[0021] Figure 6 This is a structural block diagram of the components of the execution module in the processing system for device performance statistics. DETAILED DESCRIPTION

[0022] At present, it is still necessary for humans to judge the situation captured by the camera before taking measures. Monitoring personnel monitor multiple cameras at the same time and rely on their experience to make judgments. Unsafe behaviors may be judged as safe behaviors due to their experience, or monitoring personnel may not be able to detect unsafe behaviors in time.

[0023] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0024] Embodiment 1: Figure 1 The flowchart of the method for processing device performance statistical data is as follows. In an embodiment of the present invention, a method for processing device performance statistical data includes: Step S100: Obtain historical performance statistics data, extract performance characteristics of historical performance statistics data, and the curve formula for extracting performance characteristics is: , where W is the energy consumption of the ventilation equipment, m is the mass of the ventilation equipment, p is the power of the ventilation equipment, and t is the time. Then the curve is filtered with the SG filter. The construction of the SG filter is as follows: , p(i+j) is the original input signal, y(i) is the resulting output signal, F j is the coefficient given by the SG filter, and N is the number of convolution integers; In order to ensure that the indoor air quality meets the requirements, the ventilation equipment must work normally. However, as time goes by, dust and other substances will accumulate in the ventilation equipment, and the ventilation equipment will also be affected by the environment and corroded. At this time, the quality of the ventilation equipment will change. When the ventilation equipment is in an abnormal working state, the power consumed by the ventilation equipment will change. Therefore, the power and quality of each ventilation equipment are selected as performance statistics to realize the monitoring of the ventilation equipment. This step is to extract performance characteristics. In order to build a model later, the energy consumption of the ventilation equipment when it is working normally is almost the same. When the ventilation equipment is abnormal, the energy consumption of the ventilation equipment will change suddenly. According to this principle, a corresponding curve can be established to help identify the trend of energy consumption over time. Historical performance statistics can help identify the causal relationship between variables. The SG filter can remove noise interference and can effectively keep the basic form and key features of the signal unchanged, which can ensure that the filtered curve can still retain the key features of the performance statistics, thereby ensuring the accuracy of the subsequent model.

[0025] Step S200: constructing a performance prediction model based on the performance characteristics and the LSTM neural network, specifically using a forget gate to determine which performance characteristics in the unit are discarded and retained, determining the information stored in the unit, integrating data from the forget gate and the input gate, and obtaining a performance prediction model; The LSTM neural network contains three different types of gate structures, namely the input gate, forget gate, and output gate. This structure can solve the gradient problem and then use the screened performance characteristics to establish a performance prediction model. The resulting performance prediction model is very accurate.

[0026] Step S300: Obtain real-time performance statistics of the device, substitute the real-time performance statistics into the performance prediction model, and obtain the working status of the device; Collect real-time performance statistics of ventilation equipment, substitute the real-time performance statistics of ventilation equipment into the previously constructed performance prediction model for training. The performance prediction model will output the results, compare the results with the preset thresholds, and then determine the working state of the ventilation equipment.

[0027] Step S400: Find and execute a corresponding response strategy based on the working status of the device.

[0028] Pre-set response strategies for various working conditions and execute them. Here, ventilation equipment can be divided into normal state, abnormal state and dangerous state. Normal state means that there is no problem with the ventilation equipment and the response strategy is no need to change; abnormal state means that the ventilation equipment has a minor problem but will not endanger the rest of the equipment. The response strategy is to find a solution from the strategy library. The system itself can repair the abnormal state to normal state without notifying the management personnel; dangerous state means that the ventilation equipment has a major problem and will endanger the rest of the equipment. The system itself cannot repair the dangerous state to normal state. The response strategy is to find a technician who can solve the state from the strategy library. The technician will solve the problem of the equipment and repair the dangerous state to normal state. At the same time, notify the management personnel for supervision to ensure that all ventilation equipment can be restored to normal state as soon as possible and the expected indoor air quality can be obtained within the expected construction period.

[0029] The obtaining of historical performance statistical data and extracting performance characteristics of the historical performance statistical data further includes: The performance characteristics are normalized, and the calculation formula for normalization is as follows: , is the normalized value, is the minimum value in the sequence, is the maximum value in the sequence, S is the original data, that is, find the maximum and minimum values ​​in the performance characteristics, and adjust each data point according to these extreme values ​​to maintain their relative position in the overall data set. This processing helps to maintain the scale consistency of the data in subsequent processing, which can improve the model training efficiency and prediction accuracy.

[0030] Figure 2 The second sub-flow chart of the method for processing device performance statistical data, wherein the step of constructing a performance prediction model based on performance characteristics and LSTM neural network includes: Step S201: Use the forget gate to determine which performance features in the unit are discarded and retained. The calculation formula of the forget gate is as follows: ,in is the output of the forget gate, is the weight matrix, is the bias vector, is the hidden state of the neuron at the previous time step, is the input of the LSTM neural network at the current time step, is the ReLU function; This is to screen various performance characteristics in the performance statistics, retain only the main performance characteristics, and minimize the workload on the basis of accurate model construction, thereby shortening the working time. The use of structured weight matrices can reduce the number of parameters, model complexity and storage requirements, and may also lead to faster training convergence and better generalization performance; bias vectors allow the model to better adapt to the data, making the neural network more flexible and able to fit more complex functions; is the input data, here is the power of the device; the mathematical definition of the ReLU function is: , that is, when the input x is less than 0, the output is 0; when x is greater than or equal to 0, the output is x itself.

[0031] Step S202: Update the information in the input gate unit and obtain new information using the cosh layer. The calculation formula for the information in the input gate unit is as follows: ,The calculation formula for obtaining new information using the cosh layer is as follows: ,in and is the weight matrix, and is the bias vector, is the output of the input gate at the current time step, is the newly generated memory state of the neuron, is the hyperbolic cosine function; and Using a weight matrix that includes a bias, that is, adding the bias vector of each layer directly into the weight matrix as a row, can obtain more accurate results. and They are and The corresponding bias vector uses the above calculation formula to update the state, and then stores the updated information in the unit.

[0032] Step S203: The old state c j-1 Multiply by the output of the forget gate f j , represents the data partial discarding process, and then adds the output i of the input gate j Multiply by the new memory state g of the neuron j , change the memory state of the neuron from cj-1 Updated to c j , c j It will be passed to the next neuron, and the specific calculation process is: , the output gate is based on the unit state c j 、The output of the previous cycle is h j-1 and power input x j Generate a new hidden state h j-1 , the calculation formula of the output gate is: , ,in and are the weight matrix and bias vector respectively, is the output of the output gate at the current time step, is the hidden state of the neuron at the current time step and is also the output of the current time step. Will make peace Pass it on to the next neuron.

[0033] The essence of this process of using a structured weight matrix is ​​to randomly and temporarily remove a part of the neurons during the training process, so that each neuron must rely on the group behavior of the entire network rather than the output of a single neuron. The specific reason is that when the neural network is forward propagating, each neuron has a certain probability of being temporarily disabled, so that the disabled neurons will not have any effect on the output of the network. This mechanism of randomly disabling some neurons is essentially to create multiple different "scaled" networks based on the original network. The model complexity of these networks is relatively low because they contain fewer neurons. In this way, it helps to reduce the complexity of the entire model, speed up the convergence of the model, and thus improve the generalization ability of the model.

[0034] Figure 3 The third sub-flow chart of the method for processing device performance statistical data is as follows: the steps of obtaining real-time performance statistical data of the device, substituting the real-time performance statistical data into the performance prediction model, and obtaining the working status of the device include: Step S301: Substitute the real-time performance statistical data into the performance prediction model to obtain the output result of the output gate; When constructing the performance prediction model, performance statistical data of the ventilation equipment in different working states are used. Each performance statistical data corresponds to an output result. According to the rated power of the ventilation equipment and the corresponding fault tolerance range, it can be known what power is in the normal state. The output results of these normal state powers constitute the normal state result set. Similarly, the output results of the abnormal state constitute the abnormal state result set, and the output results of the dangerous state constitute the dangerous state result set. Substituting the real-time performance statistical data into the performance prediction model will obtain a real-time output result.

[0035] Step S302: Compare the output result with the preset rules to determine whether the working state of the equipment is normal, abnormal or dangerous.

[0036] Observe which of the normal state result set, abnormal state result set and dangerous state result set the real-time output result is in, and obtain the corresponding working state of the ventilation equipment according to the belonging structure set. When the real-time output result is included in the normal state result set, the working state of the ventilation equipment is normal; when the real-time output result is included in the abnormal state result set, the working state of the ventilation equipment is abnormal; when the real-time output result is included in the dangerous state result set, the working state of the ventilation equipment is dangerous.

[0037] Embodiment 2: Figure 4 The structure block diagram of the system for processing device performance statistical data is as follows. In an embodiment of the present invention, a system for processing device performance statistical data includes: The extraction module is used to obtain historical performance statistics and extract the performance characteristics of historical performance statistics. The curve formula for extracting performance characteristics is: , where W is the energy consumption of the ventilation equipment, m is the mass of the ventilation equipment, p is the power of the ventilation equipment, and t is the time. Then the curve is filtered with the SG filter. The construction of the SG filter is as follows: , p(i+j) is the original input signal, y(i) is the resulting output signal, F j is the coefficient given by the SG filter, and N is the number of convolution integers; The model construction module is used to build a performance prediction model based on performance characteristics and LSTM neural network (Long Short-Term Memory Neural Network). Specifically, the forget gate is used to determine which performance characteristics in the unit are discarded and retained, determine the information stored in the unit, and integrate the data from the forget gate and input gate to obtain the performance prediction model. A processing module is used to obtain real-time performance statistics of the equipment, substitute the real-time performance statistics into the performance prediction model, and obtain the working status of the equipment; The execution module is used to find and execute corresponding response strategies according to the working status of the equipment.

[0038] As a further solution of the present invention: the processing system for device performance statistical data further includes: The normalization module is used to normalize the performance characteristics. The calculation formula for normalization is as follows: , is the normalized value, is the minimum value in the sequence, is the maximum value in the sequence, and S is the original data.

[0039] The model construction module includes: The screening unit is used to determine which performance features in the unit are discarded and retained using the forget gate. The calculation formula of the forget gate is as follows: ,in is the output of the forget gate, is the weight matrix, is the bias vector, is the hidden state of the neuron at the previous time step, is the input of the LSTM neural network at the current time step, is the ReLU function; The storage unit is used to determine the information stored in the unit. The information stored in the unit includes updating the information in the unit through the input gate and obtaining new information using the cosh layer. The calculation formula for updating the information in the unit through the input gate is as follows: ,The calculation formula for obtaining new information using the cosh layer is as follows: ,in and is the weight matrix, and is the bias vector, is the output of the input gate at the current time step, is the newly generated memory state of the neuron, is the hyperbolic cosine function; The output unit is used to integrate the data from the forget gate and the input gate, including: j-1 Multiply by the output of the forget gate f j , represents the data partial discarding process, and then adds the output i of the input gate j Multiply by the new memory state g of the neuron j , change the memory state of the neuron from c j-1 Updated to c j , c j It will be passed to the next neuron, and the specific calculation process is: , the output gate is based on the unit state c j 、The output of the previous cycle is h j-1 and power input x j Generate a new hidden state hj-1 , the calculation formula of the output gate is: , ,in and are the weight matrix and bias vector respectively, is the output of the output gate at the current time step, is the hidden state of the neuron at the current time step and is also the output of the current time step. Will make peace Pass it on to the next neuron.

[0040] The execution module includes: A substitution unit, used to substitute the real-time performance statistical data into the performance prediction model to obtain the output result of the output gate; The execution unit is used to compare the output result with the preset rules to determine whether the working state of the equipment is normal, abnormal or dangerous.

[0041] It should be noted that, in this article, the term "comprises" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of more restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0042] The above are only preferred embodiments of the present invention, and are not intended to limit the scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the protection scope of the present invention.

Claims

1. A method for processing equipment performance statistical data, characterized in that: The method comprises: Obtain historical performance statistics, extract performance characteristics of historical performance statistics, and the curve formula for extracting performance characteristics is: , where W is the energy consumption of the ventilation equipment, m is the mass of the ventilation equipment, p is the power of the ventilation equipment, and t is the time. Then the curve is filtered with the SG filter. The construction of the SG filter is as follows: , p(i+j) is the original input signal, y(i) is the resulting output signal, F j is the coefficient given by the SG filter, and N is the number of convolution integers; A performance prediction model is constructed based on the performance characteristics and the LSTM neural network. Specifically, the forget gate is used to determine which performance characteristics in the unit are discarded and retained, determine the information stored in the unit, and integrate the data from the forget gate and the input gate to obtain the performance prediction model. Obtain real-time performance statistics of the equipment, substitute the real-time performance statistics into the performance prediction model, and obtain the working status of the equipment; Find the corresponding response strategy and execute it according to the working status of the equipment.

2. The method for processing equipment performance statistical data according to claim 1, characterized in that: The step of extracting the performance characteristics of historical performance statistics further includes: The performance characteristics are normalized, and the calculation formula for normalization is as follows: , is the normalized value, is the minimum value in the sequence, is the maximum value in the sequence, and S is the original data.

3. The method for processing equipment performance statistical data according to claim 1 or 2, characterized in that: The calculation formula of the forget gate is as follows: ,in is the output of the forget gate, is the weight matrix, is the bias vector, is the hidden state of the neuron at the previous time step, is the input of the LSTM neural network at the current time step, is the ReLU function; the information stored in the unit is determined by updating the information in the unit through the input gate and obtaining new information using the cosh layer. The calculation formula for updating the information in the unit through the input gate is as follows: ,The calculation formula for obtaining new information using the cosh layer is as follows: ,in and is the weight matrix, and is the bias vector, is the output of the input gate at the current time step, is the newly generated memory state of the neuron, is the hyperbolic cosine function.

4. The method for processing equipment performance statistical data according to claim 3, characterized in that: The step of integrating data from the forget gate and the input gate includes: converting the old state c j-1 Multiply by the output of the forget gate f j , represents the data partial discarding process, and then adds the output i of the input gate j Multiply by the new memory state g of the neuron j , change the memory state of the neuron from c j-1 Updated to c j , c j It will be passed to the next neuron, and the specific calculation process is: , the output gate is based on the unit state c j 、The output of the previous cycle is h j-1 and power input x j Generate a new hidden state h j-1 , the calculation formula of the output gate is: , ,in and are the weight matrix and bias vector respectively, is the output of the output gate at the current time step, is the hidden state of the neuron at the current time step and is also the output of the current time step. Will make peace Pass it on to the next neuron.

5. A system for processing equipment performance statistics, characterized in that: The system comprises: The extraction module is used to obtain historical performance statistics and extract the performance characteristics of historical performance statistics. The curve formula for extracting performance characteristics is: , where W is the energy consumption of the ventilation equipment, m is the mass of the ventilation equipment, p is the power of the ventilation equipment, and t is the time. Then the curve is filtered with the SG filter. The construction of the SG filter is as follows: , p(i+j) is the original input signal, y(i) is the resulting output signal, F j is the coefficient given by the SG filter, and N is the number of convolution integers; The model construction module is used to build a performance prediction model based on performance characteristics and LSTM neural network. Specifically, the forget gate is used to determine which performance characteristics in the unit are discarded and retained, determine the information stored in the unit, and integrate the data from the forget gate and input gate to obtain the performance prediction model. A processing module is used to obtain real-time performance statistics of the equipment, substitute the real-time performance statistics into the performance prediction model, and obtain the working status of the equipment; The execution module is used to find and execute corresponding response strategies according to the working status of the equipment.

6. The processing system for device performance statistical data according to claim 5, characterized in that: Also includes: The normalization module is used to normalize the performance characteristics. The calculation formula for normalization is as follows: , is the normalized value, is the minimum value in the sequence, is the maximum value in the sequence, and S is the original data.

7. The system for processing equipment performance statistical data according to claim 5 or 6, characterized in that: The model construction module includes: The screening unit is used to determine which performance features in the unit are discarded and retained using the forget gate. The calculation formula of the forget gate is as follows: ,in is the output of the forget gate, is the weight matrix, is the bias vector, is the hidden state of the neuron at the previous time step, is the input of the LSTM neural network at the current time step, is the ReLU function; The storage unit is used to determine the information stored in the unit. The information stored in the unit includes updating the information in the unit through the input gate and obtaining new information using the cosh layer. The calculation formula for updating the information in the unit through the input gate is as follows: ,The calculation formula for obtaining new information using the cosh layer is as follows: ,in and is the weight matrix, and is the bias vector, is the output of the input gate at the current time step, is the newly generated memory state of the neuron, is the hyperbolic cosine function; The output unit is used to integrate the data from the forget gate and the input gate, including: j-1 Multiply by the output of the forget gate f j , represents the data partial discarding process, and then adds the output i of the input gate j Multiply by the new memory state g of the neuron j , change the memory state of the neuron from c j-1 Updated to c j , c j It will be passed to the next neuron, and the specific calculation process is: , the output gate is based on the unit state c j 、The output of the previous cycle is h j-1 and power input x j Generate a new hidden state h j-1 , the calculation formula of the output gate is: , ,in and are the weight matrix and bias vector respectively, is the output of the output gate at the current time step, is the hidden state of the neuron at the current time step and is also the output of the current time step. Will make peace Pass it on to the next neuron.

Images