High-temperature fan power consumption prediction method, device and equipment and storage medium

Abstract

The invention provides a high-temperature fan power consumption prediction method, device and equipment and a storage medium, and the method comprises the steps: defining a prediction period, obtaining the historical operation parameters of a high-temperature fan, and predicting an initial coding sequence of the change of a working condition in the prediction period through a machine learning model; receiving adjustment of the user on the initial coding sequence according to actual conditions to obtain a final coding sequence; inputting the fan volume flow in the historical operation parameters into a memory network model optimized by using a sparrow search method, and correcting the output of the memory network model by using an extreme learning machine model to obtain a value sequence of the fan volume flow changing in a prediction period; traversing the final coding sequence, and predicting the power consumption according to the time step in each working condition stage: judging the working condition category of the current to-be-detected stage, and predicting the power consumption based on a value sequence by introducing a temperature correction formula if the working condition category is a non-frequency-conversion category; otherwise, using online learning to predict the power consumption in a rolling manner and updating the value sequence after the current to-be-detected stage.

Description

Technical Field

[0001] The present invention relates to the technical field of energy engineering, and in particular, to a method, device, equipment and storage medium for predicting the power consumption of high-temperature fans Background Art

[0002] A high-temperature fan is a ventilation device specifically designed to work in a relatively high-temperature environment in industrial production, and is widely used in industries such as metallurgy, chemical industry, electric power and cement. The prediction of its power consumption is of great significance for optimizing energy use, reducing operating costs, and improving equipment reliability

[0003] In related technologies, traditional methods for predicting high-temperature fans include the formula method and the time series analysis method. The existing formula method only calculates based on the theoretical power demand of the fan, ignoring the losses generated during the energy conversion process of the motor, and does not consider the influence of temperature on the prediction results in specific high-temperature scenarios. The original time series method directly predicts the power consumption value, resulting in poor interpretability of the model. It is relatively sensitive to noise data and has weak robustness. With the development of artificial intelligence technology, deep machine learning models can capture more complex non-linear relationships to a certain extent. However, whether it is the traditional method or the machine model method for predicting power consumption, it cannot well adapt to different working conditions. For example, in the three cases of start-up and stop working conditions, emergency working conditions, and commissioning working conditions, the load and efficiency change violently. The performance of the offline-trained model will decline when facing frequent changes, and in this case, the traditional method performs worse

[0004] Based on the analysis of the development status of this technical field above, the existing technologies lack a solution that first uses machine learning algorithms to predict the changes in working conditions, and then uses corresponding methods to predict the power consumption of high-temperature fans in this stage according to the characteristics of working condition stability Summary of the Invention

[0005] The purpose of the present invention is to provide a method, device, equipment and storage medium for predicting the power consumption of high-temperature fans, aiming to solve the above problems in the prior art

[0006] According to the first aspect of the embodiments of the present invention, a method for predicting the power consumption of a high-temperature fan is provided, including:

[0007] Define a prediction period, obtain the historical operation parameters of the high-temperature fan, and use a machine learning model to predict the initial coding sequence of the changes in working conditions within the prediction period

[0008] Receive the adjustment of the initial coding sequence by the user according to the actual situation to obtain the final coding sequence

[0009] Input the fan volume flow rate in the historical operation parameters into a memory network model optimized by the sparrow search algorithm, and use an extreme learning machine model to correct the output of the memory network model to obtain the value sequence of the change in the fan volume flow rate within the prediction period

[0010] Traverse the final encoded sequence and predict the power consumption at each working condition stage according to the time step: Determine the working condition category of the current stage to be measured. If it is a non-variable frequency type, predict the power consumption based on the value sequence using the introduced temperature correction formula; otherwise, use online learning to roll predict the power consumption and update the value sequence after the current stage to be measured.

[0011] According to the second aspect of the embodiments of the present invention, there is provided a power consumption prediction device for a high-temperature fan, including:

[0012] A working condition prediction module, configured to define a prediction period, obtain the historical operation parameters of the high-temperature fan, and use a machine learning model to predict the initial encoded sequence of the change of the working condition within the prediction period;

[0013] A user adjustment module, configured to receive the adjustment of the initial encoded sequence by the user according to the actual situation to obtain the final encoded sequence;

[0014] A flow rate correction module, configured to input the fan volume flow rate in the historical operation parameters into a memory network model optimized by the sparrow search method, and use an extreme learning machine model to correct the output of the memory network model to obtain the value sequence of the change of the fan volume flow rate within the prediction period;

[0015] A power consumption prediction module, configured to traverse the final encoded sequence and predict the power consumption at each working condition stage according to the time step: Determine the working condition category of the current stage to be measured. If it is a non-variable frequency type, predict the power consumption based on the value sequence using the introduced temperature correction formula; otherwise, use online learning to roll predict the power consumption and update the value sequence after the current stage to be measured.

[0016] According to the third aspect of the embodiments of the present invention, there is provided an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the computer program is executed by the processor, the steps of the high-temperature fan power consumption prediction method provided in the first aspect of the present disclosure are implemented.

[0017] According to the fourth aspect of the embodiments of the present invention, there is provided a computer-readable storage medium, on which an implementation program for information transmission is stored. When the program is executed by the processor, the steps of the high-temperature fan power consumption prediction method provided in the first aspect of the present disclosure are implemented.

[0018] The technical solutions provided by the embodiments of the present invention have the following beneficial effects: Considering the influence of different working conditions on the prediction of the power consumption of high-temperature fans: For relatively stable non-inverter working conditions, the memory network model optimized by the sparrow search algorithm is used to predict the volume flow rate of the fan, and then the volume flow rate of the fan is substituted into the formula for calculation, fully considering that the essence of the power consumption change lies in the change of the volume flow rate of the fan, which enhances the interpretability of the calculation; For the inverter working conditions with relatively drastic changes, online learning is used to fully adapt to the dynamic changes; Overall, the solution has achieved good generalization ability and robustness.

[0019] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in one or more embodiments of the present specification or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the present specification. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0021] Figure 1 is a flowchart of the method for predicting the power consumption of a high-temperature fan according to an embodiment of the present invention;

[0022] Figure 2 is a schematic diagram of online rolling prediction according to an embodiment of the present invention;

[0023] Figure 3 is a schematic diagram of the complete prediction framework according to an embodiment of the present invention;

[0024] Figure 4 is a schematic diagram of the device for predicting the power consumption of a high-temperature fan according to an embodiment of the present invention;

[0025] Figure 5 is a schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of the present specification, the following will clearly and completely describe the technical solutions in one or more embodiments of the present specification with reference to the drawings in one or more embodiments of the present specification. Obviously, the described embodiments are only some embodiments of the present specification, rather than all embodiments. Based on one or more embodiments of the present specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this document.

[0027] Method Embodiment

[0028] According to an embodiment of the present invention, a method for predicting the power consumption of a high-temperature fan is provided. Figure 1 It is a flowchart of the method for predicting the power consumption of a high-temperature fan according to an embodiment of the present invention, as Figure 1 shown. The method for predicting the power consumption of a high-temperature fan according to an embodiment of the present invention specifically includes:

[0029] In step S110, a prediction period is defined, historical operating parameters of the high-temperature fan are obtained, and an initial coding sequence of the changes in the operating conditions within the prediction period is predicted using a machine learning model. Specifically, it includes:

[0030] In the embodiment of the present invention, the prediction period is 15 minutes, that is, it is necessary to predict the change of the power consumption of the high-temperature fan with time within 15 minutes after the current node;

[0031] Historical operating parameters including the volume flow rate, power, and load rate of the fan are obtained, and the specific types of the historical operating parameters are not limited;

[0032] Based on the historical operating parameters, a trained XGBoost machine learning model is used to predict the initial coding sequence. Among them, the initial coding sequence divides the prediction period into time window slices that are integer multiples. In each time window slice, the label corresponding to the operating condition is used as the coding value;

[0033] The XGBoost model can output a label according to the historical operating parameters, and this label is the coding sequence. In the embodiment of the present invention, if three time window slices are divided, that is, each slice has a duration of 5 minutes, if the normal operating condition, high-load condition, low-load condition, frequency conversion adjustment condition, start-stop condition, emergency condition, and commissioning condition are recorded as 0, 1, 2, 3, 4, 5, 6 in sequence, the output initial coding sequence is [0, 3, 0], that is, it represents that the most frequently occurring operating condition within the first 5 minutes is the normal operating condition, the most frequently occurring operating condition within the middle 5 minutes is the frequency conversion adjustment condition, and the most frequently occurring operating condition within the last 5 minutes is the normal operating condition;

[0034] To simplify the calculation amount in the embodiment of the present invention, two or more operating conditions may occur in each time window slice, and only the operating condition with the longest occurrence duration is selected.

[0035] In step S120, the adjustment of the initial coding sequence by the user according to the actual situation is received to obtain the final coding sequence. Specifically, it includes: the user can fine-tune the initial coding sequence [0, 3, 0]. For example, although it is predicted that frequency conversion is required in the second period, but the user plans to start the commissioning condition at this time, "3" can be adaptively replaced with "6" according to the actual situation, and the final coding sequence is [0, 6, 0].

[0036] In step S130, the fan volume flow rate in the historical operating parameters is input into the memory network model optimized by the sparrow search algorithm, and the output of the memory network model is corrected using the extreme learning machine model to obtain the value sequence of the change in the fan volume flow rate within the prediction period, which specifically includes:

[0037] Predict the change in the fan volume flow rate within the prediction period using the time series method, because the calculation of the fan power consumption directly depends on the fan volume flow rate;

[0038] Obtain the trained LSTM memory network model and the ELM extreme learning machine model. Among them, a set of hyperparameters is randomly generated as the initial sparrow population during the training of the LSTM memory network model. The hyperparameters include the number of hidden layer units, the number of layers, the learning rate, the batch size, and the optimizer selection, etc. The mean square error is used as the fitness value, and the search step size is dynamically adjusted through the adaptive weight to update the position of the sparrow until the fitness value converges and the iteration ends. The output corresponding to the training set of the optimized LSTM memory network model is used as the input of the untrained ELM extreme learning machine model;

[0039] The training of the ELM extreme learning machine model starts only after the training of the memory network model is completed. It is necessary to learn the output weight β to further correct the prediction error. Overall, the scheme uses LSTM as the main model and ELM as the auxiliary model to correct the prediction results. LSTM is good at capturing the long-term dependence relationship of time series, while ELM has the ability of fast learning and generalization.

[0040] During the process of power consumption prediction, only the verification part is involved and not the training. The verification process is as follows:

[0041] Input the fan volume flow rate in the historical operating parameters into the trained LSTM memory network model, and output the preliminary prediction result of the change in the fan volume flow rate within the prediction period. Use the trained ELM extreme learning machine model to correct the preliminary prediction result to obtain the value sequence. Preferably, the value sequence adopts a finer-grained segmentation effect, such as [Q1,..., Q15], that is, a value is generated every minute.

[0042] It should be noted that the predicted value sequence only participates in the non-frequency conversion working conditions, that is, the part where the power consumption can be directly predicted using the formula in the follow-up. In order to simplify the calculation process, the whole process within the prediction period is calculated; the reason for indirectly predicting the fan volume flow rate instead of directly predicting the power consumption is that the time series prediction results are not accurate enough in the case of noise data or working condition changes, and it is ultimately difficult to explain the specific source of the prediction results, which is not convenient for debugging and analysis.

[0043] In step S140, traverse the final coding sequence and predict the power consumption at each time step in each working condition stage: judge the working condition category of the current stage to be measured. If it is a non-variable frequency type, use the introduced temperature correction formula to predict the power consumption based on the value sequence; otherwise, use online learning to roll predict the power consumption and update the value sequence after the current stage to be measured, specifically including:

[0044] The judgment of the working condition category of the current stage to be measured is as follows:

[0045] If the working condition of the current stage to be measured belongs to one of the normal operation condition, high load condition, low load condition and variable frequency regulation condition, the category of the working condition of the current stage to be measured belongs to the non-variable frequency type and is relatively stable;

[0046] If the working condition of the current stage to be measured belongs to one of the start-stop condition, emergency condition, commissioning condition and other conditions except the non-variable frequency type, the category of the working condition of the current stage to be measured belongs to the variable frequency type. In this kind of working condition, the operating state of the system will suddenly change, resulting in the failure of the offline training model.

[0047] For the non-variable frequency type:

[0048] Based on the last updated value sequence, use formula 1 to predict the power consumption of the high-temperature fan at each moment within the time step span in the current stage to be measured, and use formula 2 to represent the expression of the temperature correction factor:

[0049] P = [(Q × ΔP)CF T / (η × η m ) Formula 1;

[0050] CF T = ρ / ρ T Formula 2;

[0051] P represents the predicted power consumption value, Q represents the corresponding fan volume flow rate in the value sequence, ΔP represents the fan pressure difference, CF T represents the temperature correction factor corresponding to working condition T, η represents the fan efficiency, η m represents the motor efficiency, ρ represents the high-temperature standard air density, ρ T represents the average air density of working condition T;

[0052] The fan pressure difference is obtained using CFD computational fluid dynamics software. The fan efficiency and motor efficiency are constants. The temperature correction factor is less than 1 when the temperature is high and vice versa. It should be noted that high and low temperatures are relative to the environment where the high-temperature fan is usually located;

[0053] In the field of predicting the power consumption of high-temperature fans, the motor efficiency is not considered. However, the power consumption of the entire fan system not only depends on the performance of the fan itself but also is affected by the efficiency of the motor in converting electrical energy into mechanical energy. If the motor efficiency is not considered, the losses generated by the motor during the energy conversion process will be ignored. In addition, the physical model formula method and time series prediction are effectively balanced, which is applicable to non-variable frequency working conditions.

[0054] For variable frequency type:

[0055] Obtain a trained incremental decision tree as an online learning model;

[0056] When using the online learning model to predict the power consumption at the current moment, splice the historical power consumption and the power consumption predicted before the current moment into a power consumption set with a preset continuous length in time, and input the power consumption set with the preset length into the online learning model to predict the power consumption; during the current stage to be measured, simultaneously predict the online sequence of the fan volume flow. Figure 2 This is a schematic diagram of the online rolling prediction of the embodiment of the present invention. As Figure 2 shown, it shows the process of online learning rolling prediction. Before tn+x is the time node when the prediction starts. Before this node is historical data, and after it is the prediction result. s represents the preset length. At each moment, the newly obtained prediction result is added to the end of the length set to predict the result of the next moment.

[0057] Calculate the difference between the online sequence and the value sequence at the last moment of the fan volume flow in the current prediction stage, and add the difference to the value sequence corresponding to all the moments that have not started to be predicted after the current stage to be measured. This round of update is completed.

[0058] In fact, the fan volume flow is not required during the variable frequency type stage, and the values corresponding to the corresponding stages have been predicted through time series in the value sequence. The reason for predicting another corresponding online sequence while using online learning to predict the power consumption is that the acquisition of the value sequence in step S130 does not consider the influence of working condition fluctuations, and the prediction error may be relatively large during the variable frequency working condition stage. The online sequence can consider this dynamic change to timely correct the influence on subsequent predictions, and adopts a simple method to balance directly between computational complexity and result accuracy.

[0059] In the embodiment of the present invention, the final coding sequence is [0, 6, 0], and the prediction process is as follows:

[0060] In the first 5 minutes of the prediction, Q1 to Q5 are substituted into Formula 1 for calculation, and the power consumption prediction results are E1 to E5 respectively; in the middle 5 minutes of the prediction, online learning is used to predict the online sequence of power consumption and fan volume flow, and the power consumption prediction results are E6 to E10. The fan volume flow at the last moment of the online sequence is L10, and the corresponding value of the value sequence is Q10, and L10-Q10=difference is calculated; Q11 to Q15 are updated to [Q11+difference, ..., Q15+difference]. In the last 5 minutes of the prediction, the updated Q is substituted into Formula 1 for calculation, and the power consumption prediction results are E11 to E15 respectively. The prediction is now completed.

[0061] It should be noted that if the final coding sequence is in the form of [0, 0, 6, 6, 0], the third slice and the fourth slice both belong to the debugging condition, and the same coding is merged into the same stage, that is, it is updated after the second debugging condition ends.

[0062] The above technical solutions of the embodiments of the present invention are illustrated with reference to the following drawings.

[0063] Figure 3 Schematic diagram of a complete prediction framework of an embodiment of the present invention. Figure 3 As shown in the figure, the complete process of high-temperature fan power consumption prediction is demonstrated, including using XGBoost to predict the coding sequence, using the memory network model LSTM combined with the extreme learning machine ELM to predict the fan volume flow, using the formula method for non-variable frequency working conditions, and using the online learning method for unstable variable frequency working conditions.

[0064] In summary, in response to the existing problems, the invention proposes a method for predicting the power consumption of a high-temperature fan. Considering the influence of different working conditions on the prediction of the power consumption of a high-temperature fan, for relatively stable non-variable frequency working conditions, the memory network model is used to predict the fan volume flow rate, and then the fan volume flow rate is brought into the formula for calculation, so as to fully grasp the essence of the change in power consumption, which lies in the change in the fan volume flow rate, and enhance the interpretability of the calculation; for variable frequency working conditions with more drastic changes, online learning is used to fully adapt to dynamic changes; in addition, the memory network model is optimized by the sparrow search model, and the extreme learning machine model is used to reduce the error of the memory network model prediction, thereby improving the accuracy of the fan volume flow prediction; temperature correction and motor efficiency are introduced into the formula, the former compensates for the important characteristics of the high-temperature fan, that is, the performance difference caused by temperature changes, and the latter takes into account the loss of the motor in the energy conversion process; the update of the value sequence can quickly correct the error of predicting the fan volume flow rate only by the memory network model, which is not accurate in the variable frequency working condition stage with large fluctuations, so as to avoid affecting the subsequent part that needs to be calculated based on the fan volume flow formula; the scheme as a whole has achieved good generalization ability and robustness.

[0065] Device Embodiment

[0066] According to an embodiment of the present invention, a power consumption prediction device for a high-temperature fan is provided. Figure 4 It is a schematic diagram of the power consumption prediction device for the high-temperature fan according to the embodiment of the present invention, as Figure 4 shown. The power consumption prediction device for the high-temperature fan according to the embodiment of the present invention specifically includes:

[0067] A working condition prediction module 40, which is used to define a prediction period, obtain the historical operation parameters of the high-temperature fan, and use a machine learning model to predict the initial coding sequence of the change of the working condition within the prediction period. Specifically, it is used for:

[0068] Obtain the historical operation parameters including the fan volume flow rate, power, and load rate;

[0069] Based on the historical operation parameters, use the trained XGBoost machine learning model to predict the initial coding sequence. Among them, the initial coding sequence divides the prediction period into time window slices that are integer multiples, and the label corresponding to the working condition in each time window slice is used as the coding value.

[0070] A user adjustment module 42, which is used to receive the adjustment of the initial coding sequence by the user according to the actual situation to obtain the final coding sequence;

[0071] A flow rate correction module 44, which is used to input the fan volume flow rate in the historical operation parameters into a memory network model optimized by the sparrow search method, and use an extreme learning machine model to correct the output of the memory network model to obtain the value sequence of the change of the fan volume flow rate within the prediction period. Specifically, it is used for:

[0072] Obtain the trained LSTM memory network model and ELM extreme learning machine model. Among them, a group of hyperparameters are randomly generated as the initial sparrow population during the training process of the LSTM memory network model. The mean square error is used as the fitness value, and the search step size is dynamically adjusted through the adaptive weight to update the position of the sparrow until the fitness value converges and the iteration ends. The output corresponding to the training set of the optimized LSTM memory network model is used as the input of the untrained ELM extreme learning machine model;

[0073] Input the fan volume flow rate in the historical operation parameters into the trained LSTM memory network model, output the preliminary prediction result of the change of the fan volume flow rate within the prediction period, and use the trained ELM extreme learning machine model to correct the preliminary prediction result to obtain the value sequence.

[0074] The power consumption prediction module 46 is used to traverse the final encoded sequence and predict the power consumption at each time step in each working condition stage: judge the working condition category of the current stage to be measured. If it is a non-variable frequency type, the power consumption is predicted based on the value sequence using the introduced temperature correction formula; otherwise, the online learning rolling prediction is used to predict the power consumption and update the value sequence after the current stage to be measured. Specifically, it is used for:

[0075] If the working condition of the current stage to be measured belongs to one of the normal operation condition, high load condition, low load condition and variable frequency regulation condition, the category of the working condition of the current stage to be measured belongs to the non-variable frequency type;

[0076] If the working condition of the current stage to be measured belongs to one of the start-stop condition, emergency condition, commissioning condition and other conditions except the non-variable frequency type, the category of the working condition of the current stage to be measured belongs to the variable frequency type.

[0077] For the non-variable frequency type:

[0078] Based on the last updated value sequence, use Formula 1 to predict the power consumption of the high-temperature fan at each moment within the time step span in the current stage to be measured, and use Formula 2 to represent the expression of the temperature correction factor:

[0079] P = [(Q × ΔP)CF T / (η × η m ) Formula 1;

[0080] CF T =ρ / ρ T Formula 2;

[0081] P represents the predicted power consumption value, Q represents the corresponding fan volume flow rate in the value sequence, ΔP represents the fan pressure difference, CF T represents the temperature correction factor corresponding to the working condition T, η represents the fan efficiency, η m represents the motor efficiency, ρ represents the high-temperature standard air density, ρ T represents the average air density of the working condition T.

[0082] For the variable frequency type:

[0083] Obtain the trained incremental decision tree as the online learning model;

[0084] When using the online learning model to predict the power consumption at the current moment, splice the historical power consumption and the power consumption predicted before the current moment into a power consumption set with a preset continuous time length, and input the power consumption set with the preset length into the online learning model to predict the power consumption; predict the online sequence of the fan volume flow rate simultaneously within the current stage to be measured.

[0085] Calculate the difference between the online sequence and the value sequence at the last moment of the fan volume flow rate in the current prediction stage, and add the difference to the value sequence corresponding to all the unstarted prediction moments after the current stage to be measured. This round of update is completed.

[0086] In summary, in view of the problems existing in the current situation, the high-temperature fan power consumption prediction device of the present invention takes into account the influence of different operating conditions on the high-temperature fan power consumption prediction. For relatively stable non-variable frequency operating conditions, a memory network model is used to predict the fan volume flow rate, and then the fan volume flow rate is substituted into the formula for calculation, fully grasping that the essence of the power consumption change lies in the change of the fan volume flow rate, enhancing the interpretability of the calculation; for variable frequency operating conditions with relatively drastic changes, online learning is used to fully adapt to the dynamic changes; in addition, the memory network model is optimized by the sparrow search model, and the extreme learning machine model is used to reduce the error of the memory network model prediction, improving the accuracy of the fan volume flow rate prediction; temperature correction and motor efficiency are introduced into the formula. The former compensates for an important characteristic of the high-temperature fan, that is, the performance difference caused by temperature change, and the latter takes into account the loss generated by the motor during the energy conversion process; the update of the value sequence can quickly correct the error that the fan volume flow rate is predicted only by the memory network model and is not accurate enough during the variable frequency operating condition stage with large fluctuations, avoiding affecting the subsequent part that needs to be calculated based on the fan volume flow rate formula; the overall solution has achieved good generalization ability and robustness.

[0087] Embodiment of Electronic Device

[0088] Figure 5 It is a schematic diagram of the electronic device according to an embodiment of the present invention. The electronic device 500 may include at least one processor 510 and a memory 520. The processor 510 may execute instructions stored in the memory 520. The processor 510 is communicatively connected to the memory 520 via a data bus. In addition to the memory 520, the processor 510 may also be communicatively connected to an input device 530, an output device 540, and a communication device 550 via the data bus.

[0089] The processor 510 may be any conventional processor, such as a commercially available CPU. The processor may also include, for example, a Graphic Process Unit (GPU), a Field Programmable Gate Array (FPGA), a System on Chip (SOC), an Application Specific Integrated Circuit (ASIC), or a combination thereof.

[0090] The memory 520 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0091] In an embodiment of the present disclosure, executable instructions are stored in the memory 520. The processor 510 can read the executable instructions from the memory 520 and execute the instructions to implement all or part of the steps of the high-temperature fan power consumption prediction method in any of the above exemplary embodiments.

[0092] Embodiment of computer-readable storage medium

[0093] In addition to the above methods and devices, an exemplary embodiment of the present disclosure may also be a computer program product or a computer-readable storage medium storing the computer program product. The computer program product includes computer program instructions that can be executed by a processor to implement all or part of the steps described in the high-temperature fan power consumption prediction method in any of the above exemplary embodiments.

[0094] The computer program product can be written in any combination of one or more programming languages for programming code to perform the operations of the embodiments of the present application. The programming languages include object-oriented programming languages such as Java, C++, etc., and also include conventional procedural programming languages such as the "C" language or similar programming languages and scripting languages (such as Python). The programming code can be executed entirely on the user computing device, partially on the user device, executed as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0095] The computer-readable storage medium can adopt any combination of one or more readable media. The readable media can be a readable signal medium or a readable storage medium. The readable storage medium can, for example, include but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or components, or any combination of the above. More specific examples of the readable storage medium include: static random access memory (SRAM) with one or more wire electrical connections, electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk, or any suitable combination of the above.

[0096] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting power consumption of a high-temperature fan, characterized in that: include: Define a prediction period, obtain historical operating parameters of the high-temperature fan, and use a machine learning model to predict an initial coding sequence of changes in operating conditions within the prediction period; receiving the adjustment of the initial coding sequence by the user according to actual conditions to obtain a final coding sequence; Inputting the fan volume flow rate in the historical operating parameters into the memory network model optimized by the sparrow search method, and correcting the output of the memory network model by the extreme learning machine model to obtain a value sequence of the fan volume flow rate changes within the prediction period; Traverse the final coding sequence and predict the power consumption according to the time step in each operating stage: determine the operating condition category of the current test stage. If it is a non-variable frequency type, use the temperature correction formula introduced based on the value sequence to predict the power consumption; otherwise, use online learning to roll the power consumption prediction and update the value sequence after the current test stage.

2. The method according to claim 1, characterized in that The initial coding sequence of obtaining the historical operating parameters of the high-temperature fan and using the machine learning model to predict the changes in the operating conditions within the prediction period specifically includes: Obtain historical operating parameters including fan volume flow, power and load rate; Based on the historical operating parameters, the trained XGBoost machine learning model is used to predict the initial coding sequence, wherein the initial coding sequence divides the prediction period into an integer multiple of time window slices, and the label corresponding to the operating condition in each time window slice is used as the coding value.

3. The method according to claim 1, characterized in that The step of inputting the fan volume flow rate in the historical operating parameters into the memory network model optimized by the sparrow search method and correcting the output of the memory network model by using the extreme learning machine model specifically includes: Obtain a trained LSTM memory network model and an ELM extreme learning machine model, wherein the LSTM memory network model randomly generates a set of hyperparameters as an initial sparrow population during the training process, uses the mean square error as the fitness value, dynamically adjusts the search step size to update the position of the sparrows through adaptive weights, and ends the iteration after the fitness value converges, and uses the output of the optimized LSTM memory network model corresponding to the training set as the input of the untrained ELM extreme learning machine model; The fan volume flow in the historical operating parameters is input into the trained LSTM memory network model, and the preliminary prediction result of the fan volume flow change within the prediction period is output. The trained ELM extreme learning machine model is used to correct the preliminary prediction result to obtain the value sequence.

4. The method according to claim 1, characterized in that The determination of the working condition category of the current test stage specifically includes: If the working condition of the current stage to be tested belongs to one of the normal operating condition, high load condition, low load condition and variable frequency adjustment condition, the category of the working condition of the current stage to be tested belongs to the non-variable frequency category; If the current operating condition in the test phase belongs to one of the starting and stopping conditions, emergency conditions, debugging conditions and other conditions except the non-frequency conversion type, the category of the current operating condition in the test phase belongs to the frequency conversion type.

5. The method according to claim 1, characterized in that The method of predicting power consumption based on the value sequence by introducing a temperature correction formula specifically includes: Based on the last updated value sequence, use Formula 1 to predict the power consumption of the high-temperature fan at each moment under the time step span in the current test stage, and use Formula 2 to express the expression of the temperature correction factor: P=[(Q×ΔP)CF T ] / (η×η m ) Formula 1; CF T =ρ / ρ T Formula 2; P represents the predicted power consumption value, Q represents the corresponding fan volume flow in the value sequence, ΔP represents the fan pressure difference, CF T represents the temperature correction factor corresponding to the operating condition T, η represents the fan efficiency, η m represents the motor efficiency, ρ represents the high temperature standard air density, ρ T Indicates the average air density of operating condition T.

6. The method according to claim 1, characterized in that The use of online learning to predict power consumption in a rolling manner specifically includes: Get the trained incremental decision tree as an online learning model; When using the online learning model to predict the current electricity consumption, the historical electricity consumption and the electricity consumption predicted before the current moment are spliced ​​into a time-continuous preset length electricity consumption set, and the preset length electricity consumption set is input into the online learning model to predict the electricity consumption; and the online sequence of the fan volume flow is simultaneously predicted in the current test stage.

7. The method according to claim 6, characterized in that The updating of the value sequence after the current test phase specifically includes: The difference between the volume flow rate of the fan at the last moment of the current prediction stage is calculated between the online sequence and the value sequence, and the difference is added to all value sequences corresponding to the moments not started for prediction after the current test stage, and this round of updating is completed.

8. A high-temperature fan power consumption prediction device, characterized in that: include: An operating condition prediction module, which is used to define a prediction period, obtain historical operating parameters of the high-temperature fan, and use a machine learning model to predict an initial coding sequence of operating condition changes within the prediction period; A user adjustment module, used to receive the user's adjustment to the initial coding sequence according to actual conditions, and obtain a final coding sequence; A flow correction module, used for inputting the fan volume flow in the historical operating parameters into the memory network model optimized by the sparrow search method, and correcting the output of the memory network model by using the extreme learning machine model to obtain a value sequence of the fan volume flow changing within the prediction period; The power consumption prediction module is used to traverse the final coding sequence and predict the power consumption according to the time step in each working condition stage: determine the working condition category of the current test stage, if it is a non-frequency conversion type, use the temperature correction formula introduced based on the value sequence to predict the power consumption; otherwise, use online learning to roll the power consumption prediction and update the value sequence after the current test stage.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the high-temperature fan power consumption prediction method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores an implementation program for information transmission, and when the program is executed by a processor, the steps of the high-temperature fan power consumption prediction method according to any one of claims 1 to 7 are implemented.

Images