Dynamic adjusting system for node flow of water supply network

Abstract

The invention discloses a dynamic regulation system for node flow of a water supply network, relates to the technical field of regulation of the water supply network, and solves the problem that health monitoring and leakage control of the water supply network are not realized by situation awareness. According to the method, the pressure difference and the temperature difference in the water supply pipe network and related data are collected, the flow coefficient of the water supply pipeline is confirmed based on the collected specific parameter data, numerical value confirmation is more comprehensive and accurate, the correlation accuracy of the corresponding numerical value is guaranteed, and through correlation-based time sequence data feature comparison, the correlation accuracy of the corresponding numerical value is improved. The states of downstream pipe burst, device abrasion and aging, downstream water consumption abnormity and the like are evaluated, and the application universality of the system is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of water supply network regulation, and in particular to a water supply network node flow dynamic regulation system. Background Art

[0002] Generally, the urban water supply network is nearly 10,000 kilometers long, serving an area of ​​several hundred square kilometers. Water pressure and flow regulation are important flow and energy regulation components of the urban water supply system, and are also important auxiliary facilities for production scheduling, planned water outages, and pipe burst repairs of municipal water supply networks. They play an important role in the hydraulic balance and leakage control of urban water supply networks, and are of great significance to the situational awareness of urban water supply networks, the realization of smart water functions, and energy-saving operations.

[0003] The International Water Association has summarized that the most advanced network pressure control requires the use of a zoning strategy to regulate the key valves at the boundaries of the high-pressure areas of the network, so that the pressure at the regional equalizing point is just met without redundancy; the application of this aspect in China has just started, and there are no mature equipment and systems for use in domestic water supply companies. The main bottlenecks are in two aspects: ① It is difficult to model the dynamic resistance characteristics of the regulating device at full stroke opening, resulting in the current domestic pressure reducing valves not having the function of "dynamically adjusting the boundary valve according to the change of water volume"; ② At present, domestic water supply companies have not yet solved the health monitoring and leakage control problems of the water supply network by relying on the situational awareness of intelligent regulating devices. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a water supply network node flow dynamic regulation system, which can sense network state parameters in the water supply network, self-evaluate the current fluid state, and self-regulate pressure and flow.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a water supply network node flow dynamic regulation system, which is composed of a state regulation unit, an energy regeneration unit, a multi-source sensing unit, an analysis and calculation unit, and a state evaluation unit;

[0006] Furthermore, the state regulating unit described in the present invention is composed of a cylinder, a piston, a crank, a connecting rod, a deflector cover, etc.

[0007] The energy regeneration unit described in the present invention is composed of an impeller, a transmission shaft, a generator and the like.

[0008] The multi-source sensing unit described in the present invention is composed of a temperature and pressure sensor, a differential pressure sensor, an angle sensor, etc.

[0009] The analysis and calculation unit of the present invention performs A / D conversion and analyzes and calculates the current fluid characteristic parameters according to a preset calculation formula and the flow characteristic data stored in the database.

[0010] The state evaluation unit described in the present invention extracts time domain features and frequency domain features of the fluid data based on the data provided by the analysis and calculation unit, performs feature-level and data-level fusion, builds a feature model for optimizing the least squares support vector machine parameters using a particle swarm algorithm, and performs self-state evaluation and self-regulation.

[0011] The state evaluation unit uses the large time series model Chronos to perform time series prediction on the provided traffic data; the Chronos model mainly uses the encoder-decoder structure of the T5 model, while the T5 model uses the encoder-decoder structure of the Transformer.

[0012] The present invention provides a water supply network node flow dynamic regulation system. Compared with the prior art, it has the following beneficial effects:

[0013] The present invention collects the pressure difference, temperature difference and related data inside the water supply network through the multi-source sensing unit, and confirms the flow coefficient of the water supply pipeline based on the collected specific parameter data. The numerical confirmation is more comprehensive and accurate, ensuring the correlation accuracy of the corresponding numerical values. By comparing the characteristics of the associated time series data, the downstream burst pipe, device wear and aging, and abnormal downstream water use are evaluated, ensuring the wide application of the system.

[0014] The Chronos model is used to perform correlation prediction on the flow of the water supply network. The Chronos model performs more stably on different data sets, can better adapt to diverse time series characteristics, is efficient and reliable in practical applications, and has the potential for wide application in large-scale time series prediction tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the overall plan of the present invention;

[0016] Figure 2 Schematic diagram of the Transformer model of the present invention;

[0017] Figure 3 Schematic diagram of the model architecture of Chronos of the present invention.

[0018] Figure numerals: 1. fairing; 2. piston; 3. connecting rod; 4. drive shaft; 5. crank; 6. cylinder; 7. support shaft; 8. impeller; 9. transmission shaft; 10. sealing cover; 11. generator; 20. angle sensor; 21. differential pressure sensor; 22. temperature and pressure sensor; 23. analysis and calculation unit; 24. state evaluation unit. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] See also Figure 1 , the present application provides a water supply network node flow dynamic regulation system, including a state regulation unit, an energy regeneration unit, a multi-source sensing unit, an analysis and calculation unit, and a state evaluation unit;

[0021] The state adjustment unit is mainly composed of a cylinder 6, a drive shaft 4, a crank 5, a connecting rod 3, a piston 2, and a shroud 1, wherein one end of the crank 5 is connected to the drive shaft 4, and the other end is connected to one end of the connecting rod 3, the other end of the connecting rod 3 is connected to the piston 2, and the end of the piston 2 is connected to the shroud 1; the rotation center axis of the piston 2 and the shroud 1 coincides with the rotation center axis of the cylinder 6, and the connection point of the connecting rod 3 and the piston 2 can move left and right on the above-mentioned rotation center axis, and has an eccentric distance from the rotation center formed by the crank 5 and the drive shaft 4;

[0022] The energy regeneration unit is mainly composed of a support shaft 7, an impeller 8, a transmission shaft 9 and a generator 11; one end of the support shaft 7 is connected to the cylinder 6, and the other end is connected to the impeller 8. The impeller 8 is installed in the cylinder 6, the lower end is supported by the support shaft 7, and the upper end is connected to the transmission shaft 9. The other end of the transmission shaft 9 passes through the sealing cover 10 and is connected to the generator 11. The electric energy generated by the generator 11 is used to drive the rotation of the shaft 4;

[0023] The multi-source sensing unit is composed of an angle sensor 20, a pressure difference sensor 21 and a temperature and pressure sensor 22. The angle sensor 20 is installed on the drive shaft 4, the temperature and pressure sensor 22 is installed on the inlet end of the cylinder 6, the high-pressure end of the pressure difference sensor 21 is connected to the inlet end of the cylinder 6, and the low-pressure end is connected to the inlet end of the cylinder 6; the detection signals of the pressure difference sensor 21 and the temperature and pressure sensor 22 are introduced into the analysis and calculation unit 23 for analysis and calculation; the analysis and calculation unit 23 and the state evaluation unit 24 are integrated in the embedded system, packaged and fixed on the cylinder 6, and the analysis result processed by the analysis and calculation unit 23 is transmitted to the state evaluation unit 24 for state evaluation analysis;

[0024] The specific methods for the water supply network node flow dynamic regulation system to determine the current flow include:

[0025] Based on the multi-source sensing unit, the flow rate, temperature and pressure data, and pressure difference data in the water supply pipeline are confirmed and transmitted to the analysis and calculation unit. The flow rate is determined as follows:

[0026] Q = f(P1, Px, T, Kd, ​​Cv) (Function 1); where Q is the flow rate, P1 is the inlet pressure, Px is the inlet and outlet pressure difference, T is the medium temperature, Kd is the opening of the device, and Cv is the flow coefficient of the device. This coefficient is a preset value and is prepared in advance by the relevant personnel;

[0027] An adapted electric device is used to drive the drive shaft 4 to rotate, the drive shaft 4 drives the crank 5 to rotate, the crank 5 drives the connecting rod 3 to rotate, the connecting rod 3 drives the piston 2 to move, and the deflector 1 moves with the piston 2. When the opening reaches a preset value, such as 55%, water is supplied to the inlet of the device. During the water supply process, the temperature and pressure sensor 22 in the multi-source sensing unit records the time domain data of the temperature and pressure of the inlet medium; the differential pressure sensor 21 records the time domain data of the differential pressure between the inlet and the outlet, and the angle sensor 20 records the current angle position of the crank 5;

[0028] The analysis and calculation unit performs correlation processing on the temperature and pressure data and the pressure difference data in the following manner:

[0029] Cv=1.156×Kv(Formula 1)

[0030]

[0031] Where: Kv is the device flow coefficient, ρ is the working density of the medium, and P0 is the standard density of the medium;

[0032] The current flow rate Q can be calculated by function (1), equation (1) and equation (2).

[0033] The analysis and calculation unit then calculates the Reynolds number of the current fluid, the average flow velocity of the medium, and the flow resistance coefficient of the device, and transmits the calculated analysis results to the state evaluation unit. The calculation method is as follows:

[0034]

[0035] Where: d is the inner diameter of the pipe, v is the average flow velocity of the medium, Re is the Reynolds number of the fluid flow, u is the viscosity of the medium, and K is the flow resistance coefficient of the device;

[0036] Through formula (3-5), the Reynolds number of the current fluid, the average flow velocity of the medium, and the flow resistance coefficient of the device can be calculated.

[0037] The state evaluation unit performs state evaluation analysis based on the currently confirmed analysis results, and uses the time series large model Chronos to perform time series prediction on the provided traffic data. Chronos mainly uses the encoder-decoder structure of the T5 model, while the T5 model uses the encoder-decoder structure of the Transformer.

[0038] Combination Figure 2 The Transformer model is a neural network architecture for processing sequence data. It captures the dependency of each element in the sequence with other elements based on the self-attention mechanism. The Transformer consists of an Encoder and a Decoder. Each part contains multiple layers, mainly composed of a multi-head self-attention mechanism and a feedforward neural network. The Encoder is responsible for encoding the input sequence and generating hidden representations, while the Decoder uses these representations combined with the target sequence to gradually generate outputs.

[0039] The core of Transformer is that it can process the entire sequence in parallel, rather than sequentially like traditional RNN, so it is more computationally efficient and is particularly suitable for processing long sequence tasks. In addition, the Masked self-attention mechanism of Decoder ensures that the output sequence is generated in order to avoid future information leakage.

[0040] In detail, T5 is very similar to the native Transformer structure. The difference between T5 and the original Transformer is that it uses a simplified version of LayerNorma li zat ion, that is, T5 removes the bias term in LayerNorm, making the normalization process simpler. In addition, the T5 model uses a simplified relative position encoding mechanism, in which each position encoding is a scalar that is directly added to the logits in the attention calculation. All layers share position encodings, but different attention heads in the same layer learn their own position encodings independently. The model presets a fixed number of position embeddings, corresponding to different key-query position differences, to enhance the modeling capabilities of relative position information.

[0041] like Figure 3 The figure shows the model architecture of Chronos. The Chronos model first scales and quantizes the time series data, converting continuous values ​​into discrete token sequences. Then, using the language model (T5 model)-based architecture, these token sequences are trained through the cross entropy loss function to predict future time series values. In the inference phase, tokens are sampled from the model through autoregression and mapped back to actual values ​​to generate prediction results.

[0042] Examples of practical applications:

[0043] Taking the minimum night traffic value of a certain community in 2023 and the traffic values ​​of Haida community in each period in 2023 as examples, the prediction results of the Chronos model and its comparison with the deep learning model are shown:

[0044] MSE / Mean Square Error RMSE / Root Mean Square Error <![CDATA[R 2 / Determination coefficient]]> LSTM 36.5133 6.0426 -3.5253 GRU 5.0189 2.2403 0.3944 Chronos 4.1153 2.0286 0.4817

[0045] Comparison table of numerical performance indicators of minimum night traffic in a certain community in 2023

[0046] MSE / Mean Square Error RMSE / Root Mean Square Error <![CDATA[R 2 / Determination coefficient]]> LSTM 23.5743 4.8553 0.5441 GRU 4.163 2.0403 0.9193 Chronos 4.004 2.0011 0.9225

[0047] Comparison table of traffic numerical performance indicators of a certain community in different periods of time in 2023

[0048] Based on the experimental results, the performance of the Chronos model in water supply network flow prediction is better than that of the LSTM and GRU deep learning models. In addition, the performance of the Chronos model on different data sets is more stable and can better adapt to the diverse time series characteristics, proving its efficiency and reliability in practical applications, and its potential for wide application in large-scale time series prediction tasks.

[0049] In addition, the system determines the fluid state based on the comparison of time series data, evaluates the downstream pipe burst, device wear and aging, abnormal downstream water use, etc. If an abnormal state occurs, the system controls the action of the electric device and adjusts the crank 5 angle according to the abnormal state handling mechanism.

[0050] Some of the data in the above formulas are dimensionless and numerically calculated. Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0051] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A water supply network node flow dynamic regulation system, characterized in that: It includes a state regulation unit, an energy regeneration unit, a multi-source sensing unit, an analysis and calculation unit, and a state evaluation unit; The multi-source sensing unit consists of an angle sensor, a pressure difference sensor and a temperature and pressure sensor. The angle sensor is installed on the driving shaft, the temperature and pressure sensor is installed at the inlet end of the cylinder, the high-pressure end of the pressure difference sensor is connected to the inlet end of the cylinder, and the low-pressure end is connected to the inlet end of the cylinder; the detection signals of the pressure difference sensor and the temperature and pressure sensor are introduced into the analysis and calculation unit for analysis and calculation; the analysis and calculation unit and the state evaluation unit are integrated in the embedded system, packaged and fixed on the cylinder, and the analysis results processed by the analysis and calculation unit are transmitted to the state evaluation unit for state evaluation analysis.

2. The water supply network node flow dynamic regulation system according to claim 1 is characterized in that: The state adjustment unit is mainly composed of a cylinder, a drive shaft, a crank, a connecting rod, a piston, and a fairing, wherein one end of the crank is connected to the drive shaft, the other end is connected to one end of the connecting rod, the other end of the connecting rod is connected to the piston, and the end of the piston is connected to the fairing; the rotation center axis of the piston and the fairing coincides with the rotation center axis of the cylinder, and the connection point between the connecting rod and the piston can move left and right on the above-mentioned rotation center axis, and has an eccentric distance from the rotation center formed by the crank and the drive shaft.

3. The water supply network node flow dynamic regulation system according to claim 1 is characterized in that: The energy regeneration unit is mainly composed of a support shaft, an impeller, a transmission shaft and a generator; one end of the support shaft is connected to the cylinder, and the other end is connected to the impeller. The impeller is installed in the cylinder, the lower end is supported by the support shaft, and the upper end is connected to the transmission shaft. The other end of the transmission shaft passes through a sealing cover and is connected to the generator. The electric energy generated by the generator is used to drive the rotation of the shaft.

4. The water supply network node flow dynamic regulation system according to claim 1 is characterized in that: The system determines the current flow rate in the following ways: Based on the multi-source sensing unit, the flow rate, temperature and pressure data, and pressure difference data in the water supply pipeline are confirmed and transmitted to the analysis and calculation unit. The flow rate is determined as follows: Q = f(P1, Px, T, Kd, ​​Cv) (Function 1); where Q is the flow rate, P1 is the inlet pressure, Px is the inlet and outlet pressure difference, T is the medium temperature, Kd is the opening of the device, and Cv is the flow coefficient of the device, which is a preset value; An adapted electric device is used to drive the drive shaft to rotate, the drive shaft drives the crank to rotate, the crank drives the connecting rod to rotate, the connecting rod drives the piston to move, and the deflector moves with the piston. When the opening reaches the preset value, water is supplied to the inlet of the device. During the water supply process, the temperature and pressure sensor in the multi-source sensing unit records the time domain data of the temperature and pressure of the inlet medium; the differential pressure sensor records the time domain data of the differential pressure between the inlet and the outlet, and the angle sensor records the current angular position of the crank; The analysis and calculation unit performs correlation processing on the temperature and pressure data and the pressure difference data in the following manner: Cv=1.156×Kv(Formula 1) Where: Kv is the flow coefficient of the device, ρ is the working density of the medium, and P0 is the standard density of the medium; The current flow rate Q is confirmed by function (1), equation (1) and equation (2).

5. The water supply network node flow dynamic regulation system according to claim 4 is characterized in that: The analysis and calculation unit calculates the Reynolds number of the current fluid, the average flow velocity of the medium, and the flow resistance coefficient of the device, and transmits the calculated analysis results to the state evaluation unit. The calculation method is as follows: Where: d is the inner diameter of the pipe, v is the average flow velocity of the medium, Re is the Reynolds number of the fluid flow, u is the viscosity of the medium, and K is the flow resistance coefficient of the device; By using formula (3-5), the Reynolds number of the current fluid, the average flow velocity of the medium, and the flow resistance coefficient of the device are calculated.

6. The water supply network node flow dynamic regulation system according to claim 5 is characterized in that: The state evaluation unit performs state evaluation analysis based on the currently confirmed analysis results, and uses the time series large model Chronos to perform time series prediction on the provided traffic data. Chronos mainly uses the encoder-decoder structure of the T5 model, while the T5 model uses the encoder-decoder structure of the Transformer.

7. The water supply network node flow dynamic regulation system according to claim 6 is characterized in that: The Transformer model captures the dependency of each element in the sequence with other elements based on the self-attention mechanism. The Transformer model consists of an Encoder and a Decoder. Each part contains multiple layers and is mainly composed of a multi-head self-attention mechanism and a feedforward neural network. The Encoder is responsible for encoding the input sequence and generating hidden representations, while the Decoder uses these representations in combination with the target sequence to gradually generate outputs.

8. The water supply network node flow dynamic regulation system according to claim 7 is characterized in that: The self-attention mechanism of the Decoder restricts the output sequence to be generated sequentially.

9. The water supply network node flow dynamic regulation system according to claim 7, characterized in that: The T5 model removes the bias term in LayerNorm, and adopts a simplified relative position encoding mechanism, in which each position is encoded as a scalar and directly added to the logits in the attention calculation; all layers share the position encoding, and the T5 model presets a fixed number of position embeddings, corresponding to different key-query position differences.

Images