Digital-intelligent integrated complex riverway environment monitoring system
Through the integrated digital and intelligent monitoring system, combined with fully connected neural networks and recurrent neural networks, data fusion is solved, the global problem of complex river environment monitoring is achieved, adaptive adjustment and resource utilization are achieved, and the overall and accurate monitoring is improved.
Patent Information
- Application Number
- CN202510349398.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-24
- Publication Date
- 2025-08-01
AI Technical Summary
The existing river monitoring system is difficult to achieve effective global monitoring of complex river environments, especially under multiple professional governance methods. The simple superposition of local system monitoring results leads to unsatisfactory overall monitoring results, and it is difficult to find strong negative results of superposition of multiple weak factors.
The digital and intelligent integrated monitoring system is adopted, including multiple regional monitoring subsystems, on-site communication servers, intelligent computing host groups and early warning control hosts. Data fusion processing is carried out through fully connected neural networks and recurrent neural networks, realizing monitoring from local to overall, and parallel computing is used to power the river water resources.
The global monitoring of complex river environments is realized, and the detection strategy can be adaptively adjusted, the natural environment can be avoided damage, and the natural resources can be used effectively, so as to improve the overall and accurate monitoring.
Smart Images
Figure CN120407153A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of computer systems, and particularly relates to a digital and intelligent integrated complex river environment monitoring method and system. Background Art
[0002] In recent years, the state has successively introduced a series of policies and regulations such as the Water Ten, the River and Lake Chief System, and the treatment of black and odorous water bodies for the construction of water ecological civilization, highlighting the priority role of river management in the construction of water ecological civilization. The prospects of river management projects are broad, but for the complex river environment of many upstream tributaries, due to the large number of specialties involved in the management projects and the extremely complex impacts that the management means may produce, it is difficult to ensure that the river management projects will not have negative impacts on the environment in the short term.
[0003] Therefore, the general approach to river management is to first establish a comprehensive monitoring system, and then split the management means into multiple stages, gradually manage and observe, so as to deal with it immediately when it is found that the management means have negative impacts on the environment, thereby improving the overall efficiency of river management and ensuring the overall effect.
[0004] However, the existing river monitoring systems generally have a narrow scope of application. For the complex river environment, it is difficult to achieve effective comprehensive monitoring. To meet different needs, installing multiple sets of river monitoring systems and then performing fusion processing on the monitoring data is the main means at present. However, due to the lack of effective fusion processing means, the existing technology mainly still monitors by simply superimposing the monitoring results of local systems. According to practical tests and simulation models, this simple superimposing method is difficult to ensure the global nature of monitoring, and its overall monitoring effect is very unsatisfactory. It is especially difficult to detect the situation where multiple weak factors are superimposed to form a strong negative result, resulting in frequent alarms from the downstream monitoring system while the upstream monitoring system shows no signs, and the overall monitoring is almost in a failure state. Summary of the Invention
[0005] To solve the above problems, the present invention provides a digital and intelligent integrated complex river environment monitoring system, which is characterized by including:
[0006] A monitoring subsystem, which is composed of multiple regional monitoring subsystems, is used to provide monitoring values for each region and perform subsystem calculations; when performing the subsystem calculations, the monitoring values are calculated according to the data of sensors in the river regions where they are located, and regional monitoring sequences are formed.
[0007] A field communication server, which is used to communicate with the monitoring subsystem and forward the data of the monitoring subsystem to the intelligent computing host group.
[0008] An intelligent computing host cluster is used to perform pre - calculation, core calculation, and timing calculation and send the calculation results to the early warning control host; among them, the core calculation: is used to parallel - input various regional pre - data into the core network, calculate the core feature data, and realize the extraction of combined features of the regional monitoring sequences; the pre - calculation inputs each regional monitoring sequence into the corresponding regional pre - network respectively, calculates the pre - data of each region, and realizes the pre - processing of each regional monitoring sequence; the timing calculation is used to input the core feature data into multiple recurrent networks respectively, calculate multiple early warning parameters, and the multiple early warning parameters are combined into an early warning parameter group to realize the extraction and combination of early warning features in multiple ways;
[0009] The early warning control host is used to perform early warning calculation; the early warning calculation inputs the early warning parameter group into the early warning function group, calculates multiple early warning probability values pn, and judges whether any early warning probability value pn exceeds the preset upper limit R. If it exceeds, it proceeds to the next step; if not, it returns to the subsystem calculation;
[0010] The early warning execution component is used to execute early warning alerts according to the instructions of the early warning control host; the early warning alert: is used to start a preset alert program to send out alert information.
[0011] Further, the regional pre - network and the core network are both fully - connected neural networks, and the recurrent network is a recurrent neural network.
[0012] Further, the regional pre - network, the core network, and the recurrent network are trained and calculated as an overall model; the regional pre - network outputs 15 - 20 items of data, the core network outputs 50 - 70 items of data, and the early warning parameters output by the recurrent network include 7 - 13 items of data.
[0013] Further, after the core calculation, a timing number judgment is also carried out. If the running timing number is less than 10, the output of the recurrent network in the timing calculation is truncated and returned to the subsystem calculation.
[0014] Further, the early warning function group is a linearized polynomial.
[0015] Further, the preset upper limit R is set by the user.
[0016] Further, the multiple regional monitoring subsystems include an upstream tributary monitoring system, a reservoir area monitoring system, a dam body monitoring system, and a downstream monitoring system; there is one set for each upstream tributary corresponding to the upstream tributary monitoring system.
[0017] Further, the on - site communication server, the intelligent computing host cluster, and the early warning execution host are deployed at the same location, communicate through a local network, and are 1 - 3 km away from the river channel.
[0018] Further, the monitoring subsystem sends data to the on-site communication server via wireless communication.
[0019] Further, the intelligent computing host cluster uses river water as cooling water and is powered by a water flow generator in the dam body, with a computing power of no less than 10240 TFLOPS FP16.
[0020] The present invention has the following beneficial effects:
[0021] 1) Global nature: Through the monitoring from local to overall and then to different aspects, and the reasonable fusion processing of data streams and neural network architecture design, the present application ensures the global nature of environmental detection.
[0022] 2) Adaptive adjustment: Through data accumulation and analysis, the present application can automatically adjust the detection strategy and parameters to adapt to the monitoring requirements of different river environments.
[0023] 3) Reasonable layout: Through the system layout scheme of the present application, it can adapt to various river environments and effectively avoid damage to the system caused by the natural environment.
[0024] 4) Effective utilization of environmental resources: By using river water as cooling water and being powered by a water flow generator in the dam body, the present application can effectively utilize natural environmental resources and save energy. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] Figure 1 is a connection schematic diagram provided according to an embodiment of the present invention.
[0026] Figure 2 is a schematic diagram of the calculation process provided according to an embodiment of the present invention;
[0027] Figure 3 is Figure 2 a schematic diagram of the data flow principle of DETAILED DESCRIPTION OF THE EMBODIMENTS
[0028] The present invention will be further described below with reference to the drawings and embodiments.
[0029] As Figure 1 shown in the connection schematic diagram, the present invention provides a digital and intelligent integrated complex river environment monitoring system, which is characterized in that it includes:
[0030] A monitoring subsystem, which consists of multiple regional monitoring subsystems, is used to provide monitoring values for each region and perform subsystem calculations; when performing the subsystem calculations, it calculates the monitoring values according to the data of the sensors in the river region where it is located, and forms monitoring sequences for each region.
[0031] Among them, the monitoring subsystem sends data to the on-site communication server through wireless communication. An optional wireless communication method is to use low-rate high-power wifi for ultra-long-distance communication, and another optional wireless communication method is to use LoRa modules for communication;
[0032] The multiple regional monitoring subsystems include the upstream tributary monitoring system, the reservoir area monitoring system, the dam body monitoring system, and the downstream monitoring system; there is one set for each upstream tributary corresponding to the upstream tributary monitoring system.
[0033] The on-site communication server is used to communicate with the monitoring subsystem and forward the data of the monitoring subsystem to the intelligent computing host group;
[0034] The intelligent computing host group is used to perform pre-computation, core computation, and time-series computation and send the computation results to the early warning control host;
[0035] Among them, core computation: is used to parallelly input the pre-data of each region into the core network, compute the core feature data, and realize the combined feature extraction of the regional monitoring sequence;
[0036] Regional pre-computation inputs each regional monitoring sequence into the corresponding regional pre-network respectively, computes the pre-data of each region, and realizes the preprocessing of each regional monitoring sequence;
[0037] Time-series computation is used to input the core feature data into multiple recurrent networks respectively, compute multiple early warning parameters, and merge the multiple early warning parameters into an early warning parameter group to realize the extraction and combination of early warning features in multiple ways.
[0038] Among them, the regional pre-network and the core network are both fully connected neural networks (FCNN, Full Connect Neural Network), and the recurrent network is a recurrent neural network (RNN, RerrentNeural Network); the main functions of the regional pre-network and the core network are both feature extraction. And since each item of the input data of the regional pre-network comes from sensors or data acquisition components connected and controlled by the monitoring subsystem, there is basically no data redundancy. Therefore, it is not suitable to use convolutional networks for processing, and using FCNN can effectively ensure the integrity of features; the recurrent network is used for final parameter calculation, and environmental monitoring undoubtedly has its time-series characteristics. Therefore, using RNN can effectively mine the time-series information and feature information in the data;
[0039] The regional pre-network, the core network, and the recurrent network are trained and computed as a whole model; the regional pre-network outputs 15 - 20 items of data, the core network outputs 50 - 70 items of data, and the early warning parameters output by the recurrent network include 7 - 13 items of data.
[0040] Furthermore, since the output data (i.e., the early warning parameters) of the recurrent network are processed by the early warning function group in subsequent processing, they should not be too few. Generally, it is advisable to be twice the number of early warning parameters.
[0041] After core calculation, a timing number judgment is also carried out. If the running timing number is less than 10, the output of the recurrent network in the timing calculation is truncated and returned to the subsystem calculation. The data processing of the pre-network and the core network does not consider the timing characteristics of the data, while the data processing of the recurrent network must ensure that the data has timing characteristics. Therefore, discarding the data results of the first 10 running timings has little impact on the monitoring and can ensure the effective operation of the pre-network, the core network, and the recurrent network as an overall model.
[0042] Among them, the recurrent network is set according to actual needs. For example, one recurrent network can be set for each of water level and flow monitoring, water quality monitoring, pollution monitoring, fish quantity monitoring, etc., for a total of four. Since the input data of these four recurrent networks are all the full output data of the core network, they can be expanded according to needs. The main considerations are the actual needs and the amount of data available for training.
[0043] The early warning control host is used to execute early warning calculations; in the early warning calculation, the early warning parameter group is input into the early warning function group to calculate multiple early warning probability values pn, and it is judged whether any early warning probability value pn exceeds the preset upper limit R. If it exceeds, it enters the next step; if it does not exceed, it returns to the subsystem calculation.
[0044] Among them, the early warning function group is a linearized polynomial; a typical form is as follows:
[0045]
[0046] In the formula, p1, p2,..., pn are early warning probability values, a1i, a2i,..., ani are weight parameters, ki are early warning parameters, m is the number of early warning parameters, and n is the number of early warning probability values.
[0047] The preset upper limit R is set by the user. The essence of the early warning parameter is similar to the probability value of a specific situation, in the form of {pn|n = 1, 2, 3, 4}. The calculation of the early warning function group plays a role in post-processing of weight distribution. Therefore, the preset upper limit R is generally set to 0.9.
[0048] The early warning execution component is used to execute early warning alerts according to the instructions of the early warning control host; the early warning alert: is used to start a preset warning program to send out warning information. The calculation flow chart is as Figure 2 shown, and the schematic diagram of the data flow principle is as Figure 3 shown.
[0049] The on-site communication server, the intelligent computing host cluster, and the early warning execution host are deployed at the same location, communicate through the local network, and are 1 - 3 km away from the river. On the one hand, this facilitates data communication. On the other hand, it can avoid being flooded during floods and can also utilize the river water flow for cooling.
[0050] The intelligent computing host cluster uses river water as the cooling water and is powered by the water flow generator in the dam body, with a computing power of not less than 10240 TFLOPS FP16.
[0051] A typical embodiment is as follows. Based on the above implementation manner, for a river where the tributaries on a certain three roads converge upstream, a set of upstream tributary monitoring systems is installed on each upstream tributary, which are respectively called the first upstream tributary monitoring system, the second upstream tributary monitoring system, and the third upstream tributary monitoring system. In the upstream tributary monitoring system, there are water level sensors, flow sensors, fish quantity monitoring components, fish quantity identification components, water quality monitoring components, etc., which are connected and controlled by a set of edge hosts (generally implemented by an industrial computer with an ARM - core MPU as the main control chip). A set of reservoir area monitoring system is installed in the upstream - converging reservoir. There is a dam - type hydropower station at the downstream end of the reservoir area, a dam body monitoring system is installed on the dam - type hydropower station, and a downstream monitoring system is installed on the downstream river channel of the dam - type hydropower station. Three sets of recurrent networks are set up, namely the first - type recurrent network, the second - type recurrent network, and the third - type recurrent network. The first - type recurrent network is used to calculate the early warning parameters for water level and flow monitoring, the second - type recurrent network is used to calculate the early warning parameters for pollution monitoring, and the third - type recurrent network is used to calculate the early warning parameters for fish quantity monitoring. The data of the first upstream tributary monitoring system, the second upstream tributary monitoring system, the third upstream tributary monitoring system, the reservoir area monitoring system, the dam body monitoring system, and the downstream monitoring system are respectively processed by the first upstream tributary pre - network, the second upstream tributary pre - network, the third upstream tributary pre - network, the reservoir area pre - network, the dam body pre - network, and the downstream pre - network. The first upstream tributary pre - network, the second upstream tributary pre - network, the third upstream tributary pre - network, the reservoir area pre - network, the dam body pre - network, the downstream pre - network, the core network, the first - type recurrent network, the second - type recurrent network, and the third - type recurrent network are used as a complete set of neural network models and are trained using historical data combined with simulation data. Regarding the acquisition of historical data and simulation data, it is the content of data collation in the prior art and is obtained by another technical system according to the needs of model training, so it will not be elaborated in this application.
[0052] Through the present invention, after local monitoring calculations of each subsystem are performed, the core network calculation can be used to integrate and calculate the global monitoring situation. Then, based on different monitoring needs, different recurrent networks are used for time - series analysis calculations, thereby effectively completing the monitoring from local to global and then to different aspects. This is more global than the way of simply superimposing the monitoring results of local systems in the prior art.
[0053] The above are only several specific embodiments of the present invention disclosed. However, the present invention is not limited thereto, and any changes that can be conceived by those skilled in the art should fall within the protection scope of the present invention.
Claims
1. An intelligent integration complex river environment monitoring system, characterized in that, Including: A monitoring subsystem, which consists of multiple regional monitoring subsystems, is used to provide monitoring values for each region and perform subsystem calculations; When performing the subsystem calculations, the monitoring values are calculated based on the data of sensors in the river channel regions, forming monitoring sequences for each region; A field communication server, which is used to communicate with the monitoring subsystem and forward the data of the monitoring subsystem to the intelligent computing host cluster; An intelligent computing host cluster, which is used to perform pre-computing, core computing, and time-series computing and send the computing results to the early warning control host; wherein, the core computing: is used to parallelly input the pre-region data from all channels into the core network, calculate the core feature data, and realize the extraction of combined features of the regional monitoring sequences; the pre-computing is used to separately input the monitoring sequences of each region into the corresponding regional pre-networks, calculate the pre-region data for each channel, and realize the preprocessing of the monitoring sequences of each region; the time-series computing is used to separately input the core feature data into multiple recurrent networks, calculate multiple early warning parameters, and merge the multiple early warning parameters into an early warning parameter group to realize the extraction and combination of early warning features in multiple ways; An early warning control host, which is used to perform early warning calculations; the early warning calculations input the early warning parameter group into an early warning function group, calculate multiple early warning probability values pn, and determine whether any early warning probability value pn exceeds a preset upper limit R. If it exceeds, proceed to the next step; if not, return to the subsystem calculations; An early warning execution component, which is used to execute early warning alerts according to the instructions of the early warning control host; the early warning alerts: are used to start a preset alert program to send alert messages.
2. The digital and intelligent integrated complex river environment monitoring system according to claim 1, wherein The regional pre-networks and the core network are both fully connected neural networks, and the recurrent network is a recurrent neural network.
3. The digital and intelligent integrated complex river environment monitoring system according to claim 1, wherein The regional pre-networks, the core network, and the recurrent network are trained and calculated as an integrated model; the regional pre-networks output 15 to 20 items of data, the core network outputs 50 to 70 items of data, and the early warning parameters output by the recurrent network include 7 to 13 items of data.
4. The digital intelligence integrated complex river environment monitoring system according to claim 1, characterized in that After the core computing, a time-series number judgment is also performed. If the running time-series number is less than 10, the output of the recurrent network in the time-series computing is truncated and returned to the subsystem calculations.
5. The digital intelligence integrated complex river environment monitoring system according to claim 1, wherein The early warning function group is a linearized polynomial.
6. The digital intelligence integrated complex river environment monitoring system according to claim 1, characterized in that The preset upper limit R is set by the user.
7. The digital intelligence integrated complex river environment monitoring system according to claim 1, characterized in that The multiple regional monitoring subsystems include an upstream tributary monitoring system, a reservoir area monitoring system, a dam body monitoring system, and a downstream monitoring system; one set of upstream tributary monitoring systems corresponds to each upstream tributary.
8. The digital intelligence integrated complex river environment monitoring system according to claim 1, characterized in that, The field communication server, the intelligent computing host cluster, and the early warning execution host are deployed at the same location, communicate through a local network, and are 1 to 3 km away from the river channel.
9. The digital and intelligent integrated complex river environment monitoring system according to claim 1, characterized in that, The monitoring subsystem sends data to the field communication server through wireless communication.
10. The digital intelligence integrated complex river environment monitoring system according to claim 1, wherein, The intelligent computing host cluster uses river water as cooling water and is powered by a water flow generator in the dam body, and its computing power is not less than 10240 TFLOPS FP16.