Intelligent control method for building water supply and drainage

Abstract

The invention discloses an intelligent control method for building water supply and drainage, and belongs to the technical field of building water supply and drainage. Fault detection models corresponding to target water pumps in building water supply and drainage are deployed in advance by adopting a deep learning algorithm, and then the fault detection models can be used for detecting each target water pump; therefore, the abnormal water supply pump or drainage pump can be determined, and the overall operation power is adjusted to ensure that the building water supply and drainage system can carry out normal water supply and drainage, non-inductive treatment of water supply and drainage abnormity can be realized, and the problem of water supply and drainage abnormity caused by abnormity can be avoided.

Description

Technical Field

[0001] The invention belongs to the technical field of building water supply and drainage, and specifically relates to an intelligent control method for building water supply and drainage. Background Art

[0002] The building water supply and drainage system is used to remove sewage from residential buildings, public buildings and production buildings. The drainage system inside the building is generally composed of water receivers for sanitary appliances or production equipment, drainage pipes, clearing facilities, ventilation pipes, sewage and wastewater lifting equipment and local treatment structures. In the building water supply and drainage system, both the water supply pump and the drainage pump play an extremely important role. When there is an abnormality in the water supply pump or drainage pump, if it is not handled in time, it may not only affect the normal operation of the water supply and drainage system, but also cause direct damage to the water pump in more serious cases, further expanding the loss. Summary of the invention

[0003] The present invention provides an intelligent control method for building water supply and drainage, which is used to solve the problem that the prior art cannot detect water pump abnormalities in time, thus affecting the normal operation of the building water supply and drainage system.

[0004] An intelligent control method for building water supply and drainage, comprising:

[0005] A deep learning algorithm is used to pre-deploy a fault detection model corresponding to a target water pump in a building water supply and drainage system; wherein the target water pump includes a water supply pump and a drainage pump;

[0006] The operating parameters corresponding to each target water pump are collected in real time, and the operating parameters of each target water pump are identified and analyzed using a fault detection model to determine the fault state corresponding to each target water pump; wherein the fault state includes a normal state, an inoperable state under a fault, and an operable state under a fault;

[0007] When the fault states corresponding to each of the target water pumps are all in a normal state, the process returns to the identification and analysis step to continue testing the building water supply and drainage system;

[0008] When the fault state corresponding to any of the target water pumps is a non-operable state under fault, the operation of the corresponding target water pump is stopped, and the operating power of other water pumps of the same type is intelligently adjusted;

[0009] When the fault states corresponding to any of the target water pumps are all operable states under fault, the operating power of all water pumps of the same type is intelligently adjusted to complete the intelligent control of the building's water supply and drainage.

[0010] Furthermore, a deep learning algorithm is used to pre-deploy a fault detection model corresponding to the target water pump in the building water supply and drainage, including:

[0011] A CNN algorithm is used to construct a fault detection model corresponding to a target water pump in a building water supply and drainage system, and model parameters corresponding to the fault detection model are initialized multiple times to obtain multiple model parameter codes;

[0012] In the exploration phase, the model parameter encoding is locally explored through a dual-guidance strategy consisting of random guidance and population guidance, and the model parameter encoding after local exploration is obtained;

[0013] In the optimization stage, for the model parameter encoding after local exploration, a double-encirclement search strategy with escape capability is used to rapidly develop the model parameter encoding to obtain the model parameter encoding after rapid development;

[0014] In the global stage, for the model parameter encoding after rapid development, the adaptive T distribution search strategy is used to perform global development on the model parameter encoding to obtain the model parameter encoding after global development;

[0015] Determine whether the training end conditions are met. If so, deploy the fault detection model according to the model parameter encoding after global development, otherwise return to the exploration stage.

[0016] Furthermore, the model parameter encoding is locally explored through a dual-guidance strategy consisting of random guidance and population guidance, and the model parameter encoding after local exploration is obtained, including:

[0017] Generate the first decision factor between (0,1);

[0018] When the first decision factor is greater than the first decision probability, the model parameter code is randomly matched with another model parameter code to obtain a matching parameter code corresponding to each model parameter code, and the model parameter code is randomly guided according to the matching parameter code to obtain the model parameter code after local exploration;

[0019] When the first decision factor is less than or equal to the first decision probability, the center position code corresponding to all model parameter codes and the optimal parameter code among all model parameter codes are determined, and the model parameter code is population-guided according to all the optimal parameter codes and the center position code to obtain the model parameter code after local exploration;

[0020] Wherein, the first decision probability is set to 0.5.

[0021] Furthermore, for the model parameter coding after local exploration, a double-encirclement search strategy with escape capability is used to rapidly develop the model parameter coding, and the model parameter coding after rapid development is obtained, including:

[0022] For the model parameter encoding after local exploration, the current encoding escape ability factor is determined based on the current number of training times;

[0023] When the current coding escape capability factor is greater than the second decision probability, the model parameter coding is subjected to an optimal area surround search with escape capability according to the current coding escape capability factor and the optimal parameter coding, so as to obtain the model parameter coding after rapid development;

[0024] When the current code escape capability factor is less than or equal to the second decision probability, the model parameter code is subjected to an optimal position surround search according to the optimal parameter code to obtain the model parameter code after rapid development.

[0025] Furthermore, for the model parameter encoding after rapid development, an adaptive T distribution search strategy is used to perform global development on the model parameter encoding, and the model parameter encoding after global development is obtained, including:

[0026] For the model parameter encoding after rapid development, a T distribution with a degree of freedom parameter of t is used to generate random numbers to obtain the T distribution influence factor;

[0027] Based on the current number of training times, an exponential function is used to generate a nonlinear adaptive adjustment factor;

[0028] According to the T distribution influencing factor and the adaptive adjustment factor, a global greedy search is performed on the model parameter encoding to obtain the model parameter encoding after global development.

[0029] Further, judging whether the training end condition is met includes:

[0030] It is determined whether the current training number is greater than or equal to the preset maximum training number. If so, it is determined that the training end condition is met, otherwise it is determined that the training end condition is not met.

[0031] Furthermore, the fault detection model is deployed according to the model parameter encoding after global development, including:

[0032] According to the model parameter encoding after global development, the loss function value corresponding to each model parameter encoding is determined, and the optimal parameter encoding is re-determined according to the loss function value corresponding to each model parameter encoding;

[0033] The parameters included in the optimal parameter encoding are used as the final parameters of the fault detection model, and the fault detection model is deployed.

[0034] Furthermore, the operating parameters corresponding to each target water pump are collected in real time, including:

[0035] Based on the preset data sampling frequency, the operating current, operating voltage, speed and vibration frequency of each target water pump are collected to obtain the corresponding operating parameters of each target water pump.

[0036] Furthermore, when the fault state corresponding to any of the target water pumps is a non-operable state under fault, the operation of the corresponding target water pump is stopped, and the operating power of other water pumps of the same type is intelligently adjusted, including:

[0037] When the fault state corresponding to any of the target water pumps is a non-operable state under fault, the operation of the corresponding target water pump is stopped, and the operating power of the corresponding target water pump is determined to obtain the operating power to be compensated;

[0038] The operating power to be compensated is evenly distributed to other water pumps of the same type. During the distribution process, the power upper limit of each target water pump does not exceed its rated power. If it exceeds its rated power, the undistributed operating power to be compensated is distributed to other target water pumps that have not reached the rated power until all target water pumps reach the rated power.

[0039] When all target water pumps have reached the rated power, the unallocated operating power to be compensated will continue to be evenly distributed to other water pumps of the same type. During the distribution process, the power upper limit of each target water pump shall not exceed its peak power. If it exceeds its peak power, the unallocated operating power to be compensated will be distributed to other target water pumps that have not reached the peak power. When all target water pumps have reached the peak power, the adjustment will be stopped and an alarm will be generated.

[0040] Furthermore, when the fault state corresponding to any of the target water pumps is an operable state under fault, the operating power of all water pumps of the same type is intelligently adjusted to complete the intelligent control of building water supply and drainage, including:

[0041] When the fault states corresponding to any of the target water pumps are all in an operable state under fault, the operating power of the corresponding target water pump is reduced until the corresponding target water pump operates normally, and the reduced operating power is used as the operating power to be compensated;

[0042] The operating power to be compensated is evenly distributed to other water pumps of the same type. During the distribution process, the power upper limit of each target water pump does not exceed its rated power. If it exceeds its rated power, the undistributed operating power to be compensated is distributed to other target water pumps that have not reached the rated power until all target water pumps reach the rated power.

[0043] When all target water pumps have reached the rated power, the unallocated operating power to be compensated will continue to be evenly distributed to other water pumps of the same type. During the distribution process, the power upper limit of each target water pump shall not exceed its peak power. If it exceeds its peak power, the unallocated operating power to be compensated will be distributed to other target water pumps that have not reached the peak power. When all target water pumps have reached the peak power, the adjustment will be stopped and an alarm will be generated.

[0044] The present invention provides an intelligent control method for building water supply and drainage. A deep learning algorithm is used to pre-deploy a fault detection model corresponding to a target water pump in the building water supply and drainage. The fault detection model can then be used to detect each target water pump, thereby determining an abnormal water supply pump or drainage pump, and adjusting the overall operating power to ensure that the building water supply and drainage system can perform normal water supply and drainage. This method can realize non-sensing processing of water supply and drainage anomalies, and avoid water supply and drainage anomalies caused by anomalies. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0046] Figure 1 The present invention provides a flowchart of an intelligent control method for building water supply and drainage.

[0047] The above drawings have shown clear embodiments of the present invention, which will be described in more detail below. These drawings and text descriptions are not intended to limit the scope of the present invention in any way, but to illustrate the concept of the present invention for those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0048] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0049] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0050] like Figure 1 As shown, an embodiment of the present invention provides an intelligent control method for building water supply and drainage, including:

[0051] S11. Pre-deploy a fault detection model corresponding to a target water pump in a building water supply and drainage system using a deep learning algorithm; wherein the target water pump includes a water supply pump and a drainage pump;

[0052] Deep learning algorithms are widely used in data recognition and analysis. They can learn data under abnormal conditions to form a fault detection model that can detect faults, thereby achieving anomaly detection.

[0053] S12, collecting the operating parameters corresponding to each target water pump in real time, and using the fault detection model to identify and analyze the operating parameters of each target water pump to determine the fault state corresponding to each target water pump; wherein the fault state includes a normal state, an inoperable state under a fault, and an operable state under a fault;

[0054] The operating parameter corresponding to the target water pump can be a parameter such as current. As one of the important parameters for motor operation, current is usually accompanied by abnormal changes in current when different faults occur in the motor. For example, when a turn-to-turn short circuit or a broken rotor bar fault occurs in the motor, the vibration and noise of the motor will increase during operation, and the current will also change. Current data has the advantage of being easy to obtain and process. In many industrial applications, current data can be collected through devices such as motor drivers and sensors, and can be analyzed and processed by computers. At the same time, other parameters affected by the fault can also be used as operating parameters, so that accurate analysis of the fault state can be achieved.

[0055] In the embodiment of the present invention, the normal state indicates that the water pump has no faults and can work according to preset conditions; the non-operable state under fault means that the water pump has a fault and cannot operate normally, such as a short circuit fault, an open circuit fault, a rotor fault, a bearing fault, a chamber cleaning fault, etc.; the operable state under fault means that the water pump has a fault, but the motor can continue to operate normally by adjusting the operating parameters. For example, for faults such as poor power supply contact and abnormal motor vibration, the motor can continue to operate by reducing the operating power of the motor.

[0056] S13, when the fault status corresponding to each of the target water pumps is normal, return to the identification and analysis step to continue testing the building water supply and drainage system;

[0057] When the fault status corresponding to each of the target water pumps is normal, it indicates that the entire building water supply and drainage system is working normally. At this time, there is no need to adjust the operating parameters and continue testing.

[0058] S14. When the fault state corresponding to any of the target water pumps is a non-operable state under fault, the operation of the corresponding target water pump is stopped, and the operating power of other water pumps of the same type is intelligently adjusted;

[0059] For example, when the fault state of the water supply pump is such that it cannot operate due to a fault, the operating power of other water supply pumps will be intelligently adjusted. If it is a drainage pump, the operating power of other drainage pumps will be intelligently adjusted to ensure the normal operation of the entire water supply and drainage system and give the staff enough time to deal with the abnormality.

[0060] S15. When the fault status corresponding to any of the target water pumps is an operable state under fault, the operating power of all water pumps of the same type is intelligently adjusted to complete the intelligent control of the building's water supply and drainage.

[0061] The present invention provides an intelligent control method for building water supply and drainage. A deep learning algorithm is used to pre-deploy a fault detection model corresponding to a target water pump in the building water supply and drainage. The fault detection model can then be used to detect each target water pump, thereby determining an abnormal water supply pump or drainage pump, and adjusting the overall operating power to ensure that the building water supply and drainage system can perform normal water supply and drainage. This method can realize non-sensing processing of water supply and drainage anomalies, and avoid water supply and drainage anomalies caused by anomalies.

[0062] In an embodiment of the present invention, a deep learning algorithm is used to pre-deploy a fault detection model corresponding to a target water pump in a building water supply and drainage system, including:

[0063] The CNN (Convolutional Neural Networks) algorithm is used to construct a fault detection model corresponding to the target water pump in the building water supply and drainage, and the model parameters corresponding to the fault detection model are initialized multiple times to obtain multiple model parameter codes, that is, each model parameter code contains all the hyperparameters to be optimized (such as weights, thresholds, etc.) of the fault detection model.

[0064] In the exploration phase, the model parameter encoding is locally explored through a dual-guidance strategy consisting of random guidance and population guidance, and the model parameter encoding after local exploration is obtained;

[0065] In the optimization stage, for the model parameter encoding after local exploration, a double-encirclement search strategy with escape capability is used to rapidly develop the model parameter encoding to obtain the model parameter encoding after rapid development;

[0066] In the global stage, for the model parameter encoding after rapid development, the adaptive T distribution search strategy is used to perform global development on the model parameter encoding to obtain the model parameter encoding after global development;

[0067] Determine whether the training end conditions are met. If so, deploy the fault detection model according to the model parameter encoding after global development, otherwise return to the exploration stage.

[0068] In the prior art, basic gradient descent methods or intelligent optimization algorithms such as genetic algorithms are often used to optimize the parameters of deep learning algorithms, but they are often prone to fall into local optimal values, which ultimately leads to inaccurate fault detection. Therefore, an embodiment of the present invention provides an optimization algorithm with an exploration stage, an optimization stage and a global stage to provide better hyperparameter optimization effects, thereby improving the accuracy of fault detection and improving the efficiency and accuracy of intelligent control of building water supply and drainage.

[0069] In the embodiment of the present invention, the model parameter encoding is locally explored by a dual-guidance strategy consisting of random guidance and population guidance to obtain the model parameter encoding after local exploration, including:

[0070] Generate the first decision factor between (0,1);

[0071] When the first decision factor is greater than the first decision probability, the model parameter code is randomly matched with another model parameter code to obtain a matching parameter code corresponding to each model parameter code, and the model parameter code is randomly guided according to the matching parameter code to obtain the model parameter code after local exploration;

[0072] Optionally, the model parameter encoding is randomly guided according to the matching parameter encoding, and the model parameter encoding after local exploration is obtained as: X m (t+1)=X mrand (t)-r1*|X mrand (t)-2r2X m (t)|; where X m (t) represents the mth model parameter encoding in the tth training process, m = 1, 2, ..., M, M represents the total number of model parameter encodings, X mrand (t) represents the model parameter encoding X m (t) The corresponding matching parameter code, r1 represents the first random number between (0,1), r2 represents the second random number between (0,1), X m (t+1) represents the model parameter encoding X after local exploration m (t).

[0073] The embodiment of the present invention can use random guidance to enable the model parameter encoding to search a local area based on its own position and the positions of other model parameter encodings, thereby realizing neighborhood search.

[0074] When the first decision factor is less than or equal to the first decision probability, the center position code corresponding to all model parameter codes and the optimal parameter code among all model parameter codes are determined, and the model parameter code is population-guided according to all the optimal parameter codes and the center position code to obtain the model parameter code after local exploration;

[0075] Wherein, the first decision probability is set to 0.5.

[0076] Optionally, determining the center position code corresponding to all model parameter codes and the optimal parameter code among all model parameter codes includes:

[0077] According to the encoding of all model parameters, the average of each dimension is combined into a code to obtain the central position code;

[0078] Determine the loss function value corresponding to each model parameter encoding, and take the model parameter encoding with the smallest loss function value as the optimal parameter encoding.

[0079] According to all the optimal parameter codes and center position codes, the model parameter codes are population-guided, and the model parameter codes after local exploration are obtained as follows: m (t+1)=(X best (t)-2r3X m (t))-r4(LB+r5(UB-LB)); wherein r3 represents the third random number between (0,1), r4 represents the fourth random number between (0,1), r5 represents the fifth random number between (0,1), LB represents the parameter upper limit coding, and UB represents the parameter lower limit coding.

[0080] The embodiment of the present invention can refer to the position of the optimal parameter encoding and the position of the current population center to search the local area, thereby realizing the search of more unfamiliar areas and improving the possibility of finding the optimal solution.

[0081] In an embodiment of the present invention, for the model parameter coding after local exploration, a double surround search strategy with escape capability is adopted to rapidly develop the model parameter coding, and the model parameter coding after rapid development is obtained, including:

[0082] For the model parameter encoding after local exploration, the current encoding escape ability factor is determined based on the current number of training times;

[0083] Optionally, the current encoding escape capability factor is: Wherein, λ(t) represents the current coding escape capability factor in the t-th training process, λ0 represents the initial coding escape capability factor, and T represents the preset maximum number of training times.

[0084] When the current coding escape capability factor is greater than the second decision probability, the model parameter coding is subjected to an optimal area surround search with escape capability according to the current coding escape capability factor and the optimal parameter coding, so as to obtain the model parameter coding after rapid development;

[0085] Optionally, the model parameter encoding is subjected to an optimal area encirclement search with escape capability, and the model parameter encoding after rapid development is obtained as follows: X n (t+1)=(X best (t)-X n (t))-λ(t)|ξX best (t)-X n (t)|; where X n (t) represents the model parameter encoding after the nth local exploration in the tth training process, X n (t+1) represents the model parameter encoding X after rapid development n (t), X best (t) represents the optimal parameter encoding, ξ represents the escape strength, and ξ is a random number between (0,2).

[0086] When the current code escape capability factor is less than or equal to the second decision probability, performing an optimal position surround search on the model parameter code according to the optimal parameter code to obtain the model parameter code after rapid development;

[0087] Among them, the second decision probability is set to 0.5.

[0088] Optionally, the model parameter encoding is subjected to an optimal position surround search based on the optimal parameter encoding, and the model parameter encoding after rapid development is obtained as follows:

[0089] X n (t+1)=X best (t)+|X best (t)-X n (t)|*e bl *cos(2πl)

[0090]

[0091] Among them, e represents a natural constant, π represents the ratio of pi, cos represents the cosine function, l represents a random number between (-1,1), and b represents a spiral search path.

[0092] In the early stage of the algorithm, the optimal area is searched based on the optimal position information. While improving the convergence accuracy of the algorithm, the algorithm will not easily fall into the local optimum and has a certain ability to escape the local optimum. As the number of training times increases, the current encoding escape ability factor gradually decreases, and gradually turns to an encircling search for the optimal position, so that all model parameter encodings are searched around the optimal position, thereby improving the convergence accuracy and convergence speed of the algorithm.

[0093] In the embodiment of the present invention, for the model parameter coding after rapid development, an adaptive T distribution search strategy is used to perform global development on the model parameter coding to obtain the model parameter coding after global development, including:

[0094] For the model parameter encoding after rapid development, a T distribution with a degree of freedom parameter of t is used to generate random numbers to obtain the T distribution influence factor;

[0095] Based on the current number of training times, an exponential function is used to generate a nonlinear adaptive adjustment factor;

[0096] Optionally, the adaptive adjustment factor is set to: α = (cd)*r6*e (T-t) / T ; Wherein, c represents the first constant term and is set to 1; d represents the second constant term and is set to 0.5; r6 represents the sixth random number between (0,1).

[0097] According to the T distribution influencing factor and the adaptive adjustment factor, a global greedy search is performed on the model parameter encoding to obtain the model parameter encoding after global development, including:

[0098] According to the T distribution influencing factor and the adaptive adjustment factor, the model parameter encoding is searched as: j (t+1)=X j (t)+X j (t)*τ(t)*α; where X j (t) represents the model parameter encoding after the jth rapid development, X j (t+1) represents the model parameter encoding X after global development j (t), τ(t) represents the T distribution influence factor, and the T distribution influence factor can also be set to the Levy flight factor or the Gaussian variation factor.

[0099] When the loss function value corresponding to the model parameter encoding after the search decreases, the search is accepted, otherwise it is rejected.

[0100] The global stage provided by the embodiment of the present invention adjusts the global search by introducing T distribution and adaptive adjustment factors, so that the algorithm has sufficient ability to jump out of the local optimum. As the number of training times increases, the algorithm will gradually tend to the optimal solution. Therefore, it is necessary to gradually reduce the impact of the adaptive adjustment factor on the new solution to retain the elite solution of the population.

[0101] In the embodiment of the present invention, determining whether the training end condition is met includes:

[0102] It is determined whether the current training number is greater than or equal to the preset maximum training number. If so, it is determined that the training end condition is met, otherwise it is determined that the training end condition is not met.

[0103] In an embodiment of the present invention, the fault detection model is deployed according to the model parameter encoding after global development, including:

[0104] According to the model parameter encoding after global development, the loss function value corresponding to each model parameter encoding is determined, and the optimal parameter encoding is re-determined according to the loss function value corresponding to each model parameter encoding;

[0105] The parameters included in the optimal parameter encoding are used as the final parameters of the fault detection model, and the fault detection model is deployed.

[0106] In an embodiment of the present invention, the operating parameters corresponding to each target water pump are collected in real time, including:

[0107] Based on the preset data sampling frequency, the operating current, operating voltage, speed and vibration frequency of each target water pump are collected to obtain the corresponding operating parameters of each target water pump.

[0108] In an embodiment of the present invention, when the fault state corresponding to any of the target water pumps is a non-operable state under fault, the operation of the corresponding target water pump is stopped, and the operating power of other water pumps of the same type is intelligently adjusted, including:

[0109] When the fault state corresponding to any of the target water pumps is a non-operable state under fault, the operation of the corresponding target water pump is stopped, and the operating power of the corresponding target water pump is determined to obtain the operating power to be compensated;

[0110] The operating power to be compensated is evenly distributed to other water pumps of the same type. During the distribution process, the power upper limit of each target water pump does not exceed its rated power. If it exceeds its rated power, the undistributed operating power to be compensated is distributed to other target water pumps that have not reached the rated power until all target water pumps reach the rated power.

[0111] When all target water pumps have reached the rated power, the unallocated operating power to be compensated will continue to be evenly distributed to other water pumps of the same type. During the distribution process, the power upper limit of each target water pump shall not exceed its peak power. If it exceeds its peak power, the unallocated operating power to be compensated will be distributed to other target water pumps that have not reached the peak power. When all target water pumps have reached the peak power, the adjustment will be stopped and an alarm will be generated.

[0112] In an embodiment of the present invention, when the fault state corresponding to any of the target water pumps is an operable state under fault, the operating power of all water pumps of the same type is intelligently adjusted to complete the intelligent control of building water supply and drainage, including:

[0113] When the fault states corresponding to any of the target water pumps are all in an operable state under fault, the operating power of the corresponding target water pump is reduced until the corresponding target water pump operates normally, and the reduced operating power is used as the operating power to be compensated;

[0114] The operating power to be compensated is evenly distributed to other water pumps of the same type. During the distribution process, the power upper limit of each target water pump does not exceed its rated power. If it exceeds its rated power, the undistributed operating power to be compensated is distributed to other target water pumps that have not reached the rated power until all target water pumps reach the rated power.

[0115] When all target water pumps have reached the rated power, the unallocated operating power to be compensated will continue to be evenly distributed to other water pumps of the same type. During the distribution process, the power upper limit of each target water pump shall not exceed its peak power. If it exceeds its peak power, the unallocated operating power to be compensated will be distributed to other target water pumps that have not reached the peak power. When all target water pumps have reached the peak power, the adjustment will be stopped and an alarm will be generated.

[0116] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art that are not disclosed by the present invention. It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. An intelligent control method for building water supply and drainage, characterized in that: include: A deep learning algorithm is used to pre-deploy a fault detection model corresponding to a target water pump in a building water supply and drainage system; wherein the target water pump includes a water supply pump and a drainage pump; The operating parameters corresponding to each target water pump are collected in real time, and the operating parameters of each target water pump are identified and analyzed using a fault detection model to determine the fault state corresponding to each target water pump; wherein the fault state includes a normal state, an inoperable state under a fault, and an operable state under a fault; When the fault states corresponding to each of the target water pumps are all in a normal state, the process returns to the identification and analysis step to continue testing the building water supply and drainage system; When the fault state corresponding to any of the target water pumps is a non-operable state under fault, the operation of the corresponding target water pump is stopped, and the operating power of other water pumps of the same type is intelligently adjusted; When the fault states corresponding to any of the target water pumps are all operable states under fault, the operating power of all water pumps of the same type is intelligently adjusted to complete the intelligent control of the building's water supply and drainage.

2. The intelligent control method for building water supply and drainage according to claim 1 is characterized in that: The deep learning algorithm is used to pre-deploy the fault detection model corresponding to the target water pump in the building water supply and drainage, including: A CNN algorithm is used to construct a fault detection model corresponding to a target water pump in a building water supply and drainage system, and model parameters corresponding to the fault detection model are initialized multiple times to obtain multiple model parameter codes; In the exploration phase, the model parameter encoding is locally explored through a dual-guidance strategy consisting of random guidance and population guidance, and the model parameter encoding after local exploration is obtained; In the optimization stage, for the model parameter encoding after local exploration, a double-encirclement search strategy with escape capability is used to rapidly develop the model parameter encoding to obtain the model parameter encoding after rapid development; In the global stage, for the model parameter encoding after rapid development, the adaptive T distribution search strategy is used to perform global development on the model parameter encoding to obtain the model parameter encoding after global development; Determine whether the training end conditions are met. If so, deploy the fault detection model according to the model parameter encoding after global development, otherwise return to the exploration stage.

3. The intelligent control method for building water supply and drainage according to claim 2 is characterized in that: The model parameter encoding is locally explored through a dual-guidance strategy consisting of random guidance and population guidance, and the model parameter encoding after local exploration is obtained, including: Generate the first decision factor between (0,1); When the first decision factor is greater than the first decision probability, the model parameter code is randomly matched with another model parameter code to obtain a matching parameter code corresponding to each model parameter code, and the model parameter code is randomly guided according to the matching parameter code to obtain the model parameter code after local exploration; When the first decision factor is less than or equal to the first decision probability, the center position code corresponding to all model parameter codes and the optimal parameter code among all model parameter codes are determined, and the model parameter code is population-guided according to all the optimal parameter codes and the center position code to obtain the model parameter code after local exploration; Wherein, the first decision probability is set to 0.

5.

4. The intelligent control method for building water supply and drainage according to claim 3 is characterized in that: For the model parameter encoding after local exploration, a double-surrounding search strategy with escape capability is used to rapidly develop the model parameter encoding, and the model parameter encoding after rapid development is obtained, including: For the model parameter encoding after local exploration, the current encoding escape ability factor is determined based on the current number of training times; When the current code escape capability factor is greater than the second decision probability, the model parameter code is subjected to an optimal area surround search with escape capability according to the current code escape capability factor and the optimal parameter code, so as to obtain the model parameter code after rapid development; When the current code escape capability factor is less than or equal to the second decision probability, the model parameter code is subjected to an optimal position surround search according to the optimal parameter code to obtain the model parameter code after rapid development.

5. The intelligent control method for building water supply and drainage according to claim 4 is characterized in that: For the model parameter encoding after rapid development, an adaptive T distribution search strategy is used to perform global development on the model parameter encoding, and the model parameter encoding after global development is obtained, including: For the model parameter encoding after rapid development, a T distribution with a degree of freedom parameter of t is used to generate random numbers to obtain the T distribution influence factor; Based on the current number of training times, an exponential function is used to generate a nonlinear adaptive adjustment factor; According to the T distribution influencing factor and the adaptive adjustment factor, a global greedy search is performed on the model parameter encoding to obtain the model parameter encoding after global development.

6. The intelligent control method for building water supply and drainage according to claim 5 is characterized in that: Determine whether the training end conditions are met, including: It is determined whether the current training number is greater than or equal to the preset maximum training number. If so, it is determined that the training end condition is met, otherwise it is determined that the training end condition is not met.

7. The intelligent control method for building water supply and drainage according to claim 2 is characterized in that: Deploy the fault detection model according to the model parameter encoding after global development, including: According to the model parameter encoding after global development, the loss function value corresponding to each model parameter encoding is determined, and the optimal parameter encoding is re-determined according to the loss function value corresponding to each model parameter encoding; The parameters included in the optimal parameter encoding are used as the final parameters of the fault detection model, and the fault detection model is deployed.

8. The intelligent control method for building water supply and drainage according to claim 1 is characterized in that: Real-time collection of operating parameters corresponding to each target water pump, including: Based on the preset data sampling frequency, the operating current, operating voltage, speed and vibration frequency of each target water pump are collected to obtain the corresponding operating parameters of each target water pump.

9. The intelligent control method for building water supply and drainage according to claim 8, characterized in that: When the fault state corresponding to any of the target water pumps is a non-operable state under fault, the operation of the corresponding target water pump is stopped, and the operating power of other water pumps of the same type is intelligently adjusted, including: When the fault state corresponding to any of the target water pumps is a non-operable state under fault, the operation of the corresponding target water pump is stopped, and the operating power of the corresponding target water pump is determined to obtain the operating power to be compensated; The operating power to be compensated is evenly distributed to other water pumps of the same type. During the distribution process, the power upper limit of each target water pump does not exceed its rated power. If it exceeds its rated power, the undistributed operating power to be compensated is distributed to other target water pumps that have not reached the rated power until all target water pumps reach the rated power. When all target water pumps have reached the rated power, the unallocated operating power to be compensated will continue to be evenly distributed to other water pumps of the same type. During the distribution process, the power upper limit of each target water pump shall not exceed its peak power. If it exceeds its peak power, the unallocated operating power to be compensated will be distributed to other target water pumps that have not reached the peak power. When all target water pumps have reached the peak power, the adjustment will be stopped and an alarm will be generated.

10. The intelligent control method for building water supply and drainage according to claim 9, characterized in that: When the fault state corresponding to any of the target water pumps is an operable state under fault, the operating power of all water pumps of the same type is intelligently adjusted to complete the intelligent control of building water supply and drainage, including: When the fault states corresponding to any of the target water pumps are all in an operable state under fault, the operating power of the corresponding target water pump is reduced until the corresponding target water pump operates normally, and the reduced operating power is used as the operating power to be compensated; The operating power to be compensated is evenly distributed to other water pumps of the same type. During the distribution process, the power upper limit of each target water pump does not exceed its rated power. If it exceeds its rated power, the undistributed operating power to be compensated is distributed to other target water pumps that have not reached the rated power until all target water pumps reach the rated power. When all target water pumps have reached the rated power, the unallocated operating power to be compensated will continue to be evenly distributed to other water pumps of the same type. During the distribution process, the power upper limit of each target water pump shall not exceed its peak power. If it exceeds its peak power, the unallocated operating power to be compensated will be distributed to other target water pumps that have not reached the peak power. When all target water pumps have reached the peak power, the adjustment will be stopped and an alarm will be generated.

Images