CN-122018339-A - Regional water pump station intelligent gate management control method based on edge calculation

CN122018339ACN 122018339 ACN122018339 ACN 122018339ACN-122018339-A

Abstract

The invention discloses an intelligent gate management control method for regional water pump stations based on edge calculation, which belongs to the intelligent control neighborhood of the water pump stations and comprises the steps of calculating running state similarity coefficients and running environment similarity coefficients among regional water pump stations to obtain total running similarity coefficients, constructing an objective function for distributing edge nodes to regional edge gateways, constructing a time sequence prediction model Attention-LSTM for predicting the opening of the intelligent gate at the edge nodes, outputting an upper boundary and a lower boundary of an operation data prediction value through a Monte Carlo method to obtain an intelligent gate opening value range predicted in future time steps, and adjusting the opening of the intelligent gate within the optimal intelligent gate opening value range by the edge nodes based on the opening control objective function to realize the control of the intelligent gate. The intelligent gate opening control method and the intelligent gate opening control system can meet the accuracy of single intelligent gate opening control in the control process of the intelligent gate, and can also meet the rationality and relevance of cooperative control of the intelligent gate, and reduce the management control difficulty.

Inventors

WANG ZHUOYUE
DONG XIAOQING
MA YINGXIN
Lei xinyi
RAN ZHILIN
YAO MENG

Assignees

深圳信息职业技术大学

Dates

Publication Date: 20260512
Application Date: 20260413

Claims (7)

1. An intelligent gate management control method for regional water pump stations based on edge calculation is characterized by comprising the following steps: step S1, collecting historical operation data of water pump stations at edge nodes of the water pump stations, and calculating operation state similarity coefficients among regional water pump stations according to the fluctuation rate of the operation data; S2, calculating the similarity coefficient of the running environment among the regional water pump stations, and calculating the total running similarity coefficient among the regional water pump stations by combining the running state similarity coefficient; S3, constructing an objective function for distributing edge nodes to the regional edge gateway based on the total operation similarity coefficient between the water pump stations Using an objective function Distributing an edge node set to be managed to each regional edge gateway; S4, constructing a time sequence prediction model Attention-LSTM for predicting the opening degree of the intelligent gate at the edge node, preprocessing the collected historical operation data to obtain a time sequence characteristic data set, inputting the time sequence characteristic data set into the time sequence prediction model Attention-LSTM, and outputting a hidden state after Attention weighting; S5, outputting the hidden state through a full connection layer to obtain a predicted value of the operation data, and outputting an upper boundary and a lower boundary of the predicted value of the operation data through a Monte Carlo method to obtain an intelligent gate opening value range predicted in future time steps; And S6, inputting the predicted intelligent gate opening value range into the regional edge gateway by each edge node, taking out the optimal intelligent gate opening value range, issuing the optimal intelligent gate opening value range into each edge node, constructing an opening control objective function, and adjusting the opening of the intelligent gate within the optimal intelligent gate opening value range by the edge node based on the opening control objective function to realize the control of the intelligent gate.
2. The regional water pump station intelligent gate management control method based on edge calculation according to claim 1, wherein the step S1 comprises: step S11, setting a period T for dynamically distributing edge nodes to be managed to an area edge gateway, wherein each edge node sends the water pump station operation data acquired by an edge terminal in the history period T to a water pump station control center; Step S12, obtaining time sequence operation data of each water pump station Wherein M is the number of operation data collected in the history period T, n is the type of the operation data, i is the number of the water pump station, The method comprises the steps of collecting M-th operation data n of an ith water pump station in a history period T; step S13, calculating the fluctuation rate of the nth operation data of the water pump station in the history period T ; Step S14, obtaining the fluctuation rate data of all operation data of the water pump station in the history period T N is the number of kinds of operation data, The fluctuation rate of the Nth operation data of the water pump station in the history period T is used; s15, calculating the similarity coefficient of the running states of the water pump stations in the historical period T by using the fluctuation rate data; ; Wherein, the For the fluctuation rate of the nth operating data of the water pump station j in the history period T, A threshold of fluctuation rate for the nth operation data, As the weight of the nth operational data, The running state similarity coefficient between the water pump station j and the water pump station i.
3. The regional water pump station intelligent gate management control method based on edge calculation according to claim 2, wherein the step S2 comprises: s21, calculating the similarity coefficient of the running environment between the water pump stations according to the lift between the upstream and the downstream of the intelligent gate of the water pump station and the instantaneous water quality data flowing through the intelligent gate in a history period T; ; Wherein, the The lifts between the upstream and the downstream of the intelligent gates of the water pump station i and the water pump station j respectively, For ideal head between upstream and downstream, U is the type number of water quality data, Respectively the water pump station j and the water pump station i flow through the m-th instantaneous water quality data u of the intelligent gate in the history period T, Is an ideal value of water quality data u when the water pump station operates, Respectively the influence weights of the lift and water quality data on the operation of the water pump station, Is the similarity coefficient of the running environment between the water pump station j and the water pump station i, The number of the instantaneous water quality data collected in the history period T is as follows; s22, calculating the total operation similarity coefficient between the water pump stations according to the operation state similarity coefficient and the operation environment similarity coefficient; ; Wherein, the For the weight coefficients of the operating state and the operating environment, Is the total operation similarity coefficient between the water pump station j and the water pump station i.
4. The intelligent gate management control method for regional water pump stations based on edge calculation as claimed in claim 3, wherein in step S3, an objective function for assigning edge nodes to regional edge gateways is constructed according to the number W of regional edge gateways and management capability ; ; Wherein, the A set of edge nodes managed for a regional edge gateway, Representing a set of edge nodes The total operational similarity coefficients between the corresponding water pump stations of the inner edge nodes are summed, The maximum number of edge node management embodied for regional edge gateway management capabilities, To meet the edge node management quantity of the minimum management benefit of the regional edge gateway, To assign a managed number of edge nodes to regional edge gateways based on an objective function, For a set of edge nodes The maximum of the geographical distances between the intermediate edge nodes, The allowed threshold of geographic distance between edge nodes is managed for the regional edge gateway.
5. The regional water pump station intelligent gate management control method based on edge calculation according to claim 4, wherein the step S4 comprises: Step S41, constructing a time sequence prediction model Attention-LSTM for predicting the opening degree of the intelligent gate in the edge nodes, collecting historical operation data of the water pump station by utilizing edge terminals associated with the edge nodes, carrying out normalization processing on each historical operation data, and mapping the historical operation data to the data Within the interval, a time sequence characteristic data set is obtained ; ; Wherein, the Is the normalized value of the nth operation data, t is the acquisition time of the operation data, Respectively a first-order difference and a second-order difference, To normalize the operational data for acquisition time t-1, To acquire a first order difference of time t-1, Is the normalized value of the pressure difference of the pipe network, Respectively normalizing the inlet pressure and the outlet pressure of the water pump station pipe network, Is a normalized value of the upstream water level difference, The normalized values of the upstream water level and the downstream water level of the intelligent gate are respectively; Step S42, time sequence characteristic data set Dividing the time window into continuous input characteristic sequence data segments according to a set length B is the number of the input characteristic sequence data segment, and the continuous input characteristic sequence data segments are used for the data Inputting into LSTM layer of time sequence prediction model Attention-LSTM, extracting time sequence characteristics and outputting hidden state ; Step S43, hiding the output by attention mechanism Weighting and outputting the hidden state after attention weighting ; ; Wherein, the For the attention score, K is the latitude of the attention coefficient, K is the number of latitudes of the attention coefficient, In order for the attention coefficient to be a factor of attention, As the attention coefficient of the kth weft, In order for the attention to be weighted, Is an attention bias.
6. The regional water pump station intelligent gate management control method based on edge calculation according to claim 5, wherein the step S5 comprises: step S51, hiding state after weighted attention Obtaining a predicted value of the operation data through the output of the full connection layer, and outputting an uncertainty range of a 95% confidence interval through a Monte Carlo method to obtain an upper boundary and a lower boundary of the predicted value of the operation data; ; Wherein, the The weights and offsets of the full connection layer respectively, For future time steps A vector of predicted values for the operational data, For future time steps The predicted value of the operational data is calculated, To obtain a set of predicted values by setting M monte carlo samples, The representation is a function of the number of digits taken, Respectively a lower boundary and an upper boundary of the operation data predicted value; step S52, obtaining the upper boundary of the predicted value of the intelligent gate opening in the operation data And lower boundary Forming future time steps in water pump station Predicted intelligent gate opening value range 。
7. The regional water pump station intelligent gate management control method based on edge calculation according to claim 6, wherein the step S6 comprises: Step S61, each edge node uploads the predicted intelligent gate opening value ranges to an area edge gateway for managing the edge node, the area edge gateway obtains E predicted intelligent gate opening value ranges, and takes the overlapping areas of the E predicted intelligent gate opening value ranges, and the overlapping areas are used as future time steps An optimal intelligent gate opening value range G; step S62, the regional edge gateway transmits an optimal intelligent gate opening value range G to each edge node, and the edge nodes construct an opening control objective function based on the optimal loss and energy consumption of the intelligent gate ; ; Wherein, the The energy consumption coefficient and the energy consumption coefficient of the opening degree of the intelligent gate, Is the maximum opening degree adjustment amount allowed per unit time step, The opening of the intelligent gate is respectively two adjacent time steps; step S63 step in future time In, the edge node controls the objective function based on the opening degree And (3) taking the minimum value of the intelligent gate as a target, and adjusting the opening of the intelligent gate in the optimal value range G of the opening of the intelligent gate to realize the control of the intelligent gate.

Description

Regional water pump station intelligent gate management control method based on edge calculation Technical Field The invention relates to the field of intelligent control of water pump stations, in particular to an intelligent gate management control method of an area water pump station based on edge calculation. Background The water pump station is used as a core hub of the fluid conveying system, and the accurate control of the opening degree of the gate directly influences the water supply/drainage efficiency, the energy consumption cost and the pipe network safety. The existing gate control method mainly comprises three steps of manually adjusting based on manual experience, relying on subjective judgment of operators on water level and flow, having low control precision and lag response, being incapable of adapting to complex working condition changes, adopting local automatic control based on PLC (programmable logic controller), mostly adopting single parameter (such as water level or pressure) closed-loop adjustment, neglecting multi-parameter coupling influences such as flow, pipe network resistance, medium viscosity and the like, being easy to cause control oscillation and higher energy consumption, and uploading sensing data to a cloud for modeling and decision-making based on centralized intelligent control, wherein the problems of multi-parameter analysis can be realized, but data transmission delay (usually more than or equal to 100 ms), control failure when network congestion, poor adaptability of an edge working condition and a cloud model and the like exist. In recent years, the edge computing technology starts to be applied to the field of water pump station control by virtue of the advantages of local data processing and low delay response, but the existing scheme still has obvious defects that firstly, an edge node only bears a data forwarding or simple threshold judging function, predictive control and depth optimization are not realized, and the risk of fluctuation of working conditions cannot be avoided in advance, secondly, a cooperative control mechanism of a gate of a plurality of water pump stations in a region is lacked, and pressure imbalance and uneven flow distribution of a pipe network are easily caused by independent operation of each pump station. Disclosure of Invention Aiming at the defects in the prior art, the invention provides an intelligent gate management control method for an area water pump station based on edge calculation, which realizes cooperative management control of the intelligent gate of the area water pump station and optimizes intelligent gate opening control. In order to achieve the aim of the invention, the invention adopts the following technical scheme: The utility model provides a regional water pump station intelligent gate management control method based on edge calculation, it includes: step S1, collecting historical operation data of water pump stations at edge nodes of the water pump stations, and calculating operation state similarity coefficients among regional water pump stations according to the fluctuation rate of the operation data; S2, calculating the similarity coefficient of the running environment among the regional water pump stations, and calculating the total running similarity coefficient among the regional water pump stations by combining the running state similarity coefficient; S3, constructing an objective function for distributing edge nodes to the regional edge gateway based on the total operation similarity coefficient between the water pump stations Using an objective functionDistributing an edge node set to be managed to each regional edge gateway; S4, constructing a time sequence prediction model Attention-LSTM for predicting the opening degree of the intelligent gate at the edge node, preprocessing the collected historical operation data to obtain a time sequence characteristic data set, inputting the time sequence characteristic data set into the time sequence prediction model Attention-LSTM, and outputting a hidden state after Attention weighting; S5, outputting the hidden state through a full connection layer to obtain a predicted value of the operation data, and outputting an upper boundary and a lower boundary of the predicted value of the operation data through a Monte Carlo method to obtain an intelligent gate opening value range predicted in future time steps; And S6, inputting the predicted intelligent gate opening value range into the regional edge gateway by each edge node, taking out the optimal intelligent gate opening value range, issuing the optimal intelligent gate opening value range into each edge node, constructing an opening control objective function, and adjusting the opening of the intelligent gate within the optimal intelligent gate opening value range by the edge node based on the opening control objective function to realize the control of the intelligent gate. Further, step S1 includes: step S