Rainfall monitoring data correction optimization method

Abstract

The invention relates to the technical field of rainfall data, in particular to a rainfall monitoring data correction and optimization method. Comprising the steps of setting a sampling rainfall station; capacitance rainfall monitoring data, voltage rainfall monitoring data and real rainfall data are recorded; constructing a rainfall station space capacitance deviation prediction model and a rainfall station space voltage deviation prediction model; the capacitance weight coefficient and the voltage weight coefficient of each rainfall station are optimally distributed; and the capacitance rainfall monitoring data and the voltage rainfall monitoring data recorded by each rainfall station in history are optimized. The advantages of a capacitance rainfall monitoring sensor and a voltage rainfall monitoring sensor are combined, a fusion scheme is independently designed for the unique geographical environment difference of each rainfall station, rainfall monitoring data can be corrected and optimized, and the accuracy of the data is remarkably improved.

Description

Technical Field

[0001] The invention relates to the technical field of rainfall data, and in particular to a rainfall monitoring data correction and optimization method. Background Art

[0002] Rainfall data is a very important basic data. It can be used to predict important natural disaster indicators such as droughts, floods, and river and reservoir water levels. Therefore, its accuracy is crucial.

[0003] Traditionally, rainfall monitoring is mainly done through tipping bucket mechanical measurement, but it is exposed to the natural environment for a long time and is prone to various physical failures, resulting in serious distortion of rainfall data. With the rapid development of electronic equipment, electronic sensors have become the mainstream means of data monitoring due to their low price and stability.

[0004] There are two commonly used quantity monitoring sensors. One is the capacitance monitoring sensor, which works on the principle that changes in rainfall will cause changes in the dielectric constant of the capacitor, which in turn leads to changes in the capacitance value. By measuring this change, a corresponding relationship between changes in rainfall and capacitance can be established, thereby detecting the amount of rainfall. The other is the voltage monitoring sensor, which uses the piezoelectric effect of piezoelectric vibrators to convert mechanical vibrations into electrical signals and calculates rainfall by measuring the voltage waveform generated by the impact of raindrops.

[0005] Its shortcomings are: 1. The voltage-type monitoring sensor is more accurate in monitoring data during heavy rainfall, but for rainfall with less impact, the impact force generated may not cause mechanical vibration, thus causing data distortion; 2. The capacitive monitoring sensor completes rainfall monitoring by the change of constants, but the humid environmental factors will also cause data distortion. Summary of the invention

[0006] The present invention discloses a method for correcting and optimizing rainfall monitoring data. The rainfall monitoring data is obtained through a capacitive rainfall monitoring sensor and a voltage rainfall monitoring sensor installed in a rainfall station. The specific method is as follows:

[0007] Sampling rainfall stations were randomly set up at equal intervals within the revised optimization area;

[0008] Record the capacitance rainfall monitoring data, voltage rainfall monitoring data and real rainfall data of the sampling rain gauge station during several rainfall events in the future time period;

[0009] According to the spatial coordinates of the rain gauge station and the actual rainfall data, the spatial capacitance deviation prediction model and the spatial voltage deviation prediction model of the rain gauge station are constructed;

[0010] For each rainfall station, the capacitance weight coefficient and voltage weight coefficient of each rainfall station are allocated by particle swarm optimization algorithm;

[0011] The capacitance rainfall monitoring data and voltage rainfall monitoring data recorded by each rainfall station in historical records are optimized using the capacitance weight coefficient and voltage weight coefficient corresponding to each rainfall station.

[0012] The advantage of this embodiment is that it combines the advantages of the capacitive rainfall monitoring sensor and the voltage rainfall monitoring sensor, and designs a fusion solution for each rain gauge station's unique geographical environment differences, which can correct and optimize the rainfall monitoring data and significantly improve the accuracy of the data. In addition, the geographical environment differences are connected and changed with the spatial coordinates of the rain gauge station, so the mathematical relationship between the geographical environment differences and the spatial coordinate changes can be fitted to complete the digital quantification of the geographical environment.

[0013] Furthermore, the rainfall station spatial capacitance deviation prediction model and the rainfall station spatial voltage deviation prediction model are both deep learning models, and the specific construction methods are as follows:

[0014] The real rainfall is divided into several real rainfall intervals, and for each rainfall station, the average capacitance rainfall monitoring deviation and the average voltage rainfall monitoring deviation in each real rainfall interval are calculated;

[0015] Based on the spatial coordinates of the rain gauge station, the interval where the real rainfall is located, and the average capacitance rainfall monitoring deviation or the average voltage rainfall monitoring deviation corresponding to the real rainfall interval, a training sample of the spatial capacitance deviation prediction model of the rain gauge station or the spatial voltage deviation prediction model of the rain gauge station is constructed;

[0016] The deep learning model is trained with training samples to obtain the spatial capacitance deviation prediction model and the spatial voltage deviation prediction model of the rainfall station.

[0017] The advantage of this embodiment is that different levels of rainfall have an important impact on the monitoring accuracy of the two sensors. Therefore, the rainfall range and spatial coordinates are used as the most important variables to determine the credibility of each of the two sensors, and a simple method for quantifying complex variables is found.

[0018] Furthermore, the real rainfall is divided into several real rainfall intervals, and the specific method is as follows:

[0019] Dividing the maximum true rainfall and the minimum true rainfall into a preset number of initial intervals at equal intervals;

[0020] Calculate the variance of the capacitance rainfall monitoring deviation and the variance of the voltage rainfall monitoring deviation of all rainfall stations in each initial interval;

[0021] Adjust the upper and lower limits of the initial interval to minimize the sum of the variance of the capacitance rainfall monitoring deviation and the variance of the voltage rainfall monitoring deviation;

[0022] When the number of iterations reaches the target, or the variance is less than the preset value, the true rainfall interval is obtained.

[0023] The advantage of this embodiment is that when dividing the real rainfall interval, it is necessary to divide the rainfall with the same deviation characteristics into the same interval. The scientific division of the real rainfall interval can improve the accuracy of the final data.

[0024] Furthermore, the deep learning model is a neural network model, and the specific construction method is as follows:

[0025] Determine the number of nodes in the input layer, the input layer is the coordinates of the rainfall station interval and the area where the actual rainfall is located;

[0026] Determine the number of hidden layer nodes;

[0027] Determine the number of nodes in the output layer, the output layer is capacitance deviation or voltage deviation;

[0028] Initialize weights and biases;

[0029] Set the learning rate and number of iterations;

[0030] Perform iterative training and update weights and biases through the back-propagation algorithm;

[0031] Verify the model accuracy, precision, and recall.

[0032] Furthermore, the particle swarm optimization algorithm is used to allocate the capacitance weight coefficient and voltage weight coefficient of each rainfall station. The specific method is as follows:

[0033] Get the spatial coordinates of the rainfall station;

[0034] The capacitance rainfall monitoring deviation and voltage rainfall monitoring deviation of the rainfall station in each real rainfall interval are predicted by using the rainfall station space capacitance deviation prediction model and the rainfall station space voltage deviation prediction model;

[0035] The fitness function is constructed, and the capacitance weight coefficient and voltage weight coefficient are solved by particle swarm optimization algorithm.

[0036] Furthermore, the particle swarm optimization algorithm is used to solve the capacitance weight coefficient and the voltage weight coefficient. The specific method is as follows:

[0037] Randomly generate particle positions and velocities;

[0038] Initialize the best individual position of each particle and the global best position;

[0039] According to the fitness function, calculate the fitness value of each particle;

[0040] According to the fitness value, update the individual best position and the global best position;

[0041] Update the velocity and position of each particle;

[0042] Stop iteration until convergence condition is reached;

[0043] Returns the global best position and fitness value as the optimal solution.

[0044] Furthermore, the fitness function has the following specific formula:

[0045]

[0046] Among them, f(x i ) is the fitness value of the i-th rainfall station, j is the j-th real rainfall interval, there are n real rainfall intervals in total, is the capacitance weight coefficient of the ith rainfall station, is the voltage weight coefficient of the ith rainfall station, is the average capacitance rainfall monitoring data of the i-th rain gauge station in the j-th real rainfall interval, is the average voltage rainfall monitoring data of the ith rainfall station in the jth real rainfall interval.

[0047] The advantage of this embodiment is that a constant fusion weight is set for each rainfall station, which can simplify the operation difficulty and can be easily implemented in an outdoor environment with poor conditions, and has strong applicability. The allocation of fusion weights takes into account the comprehensive deviation of each rainfall interval, takes into account each rainfall interval, and can improve the overall stability of the data.

[0048] Furthermore, the optimized and corrected rainfall monitoring data is the capacitance weight coefficient multiplied by the capacitance rainfall monitoring data, plus the voltage weight coefficient multiplied by the voltage rainfall monitoring data.

[0049] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings of the present invention are as follows.

[0051] Figure 1 It is a schematic diagram of the overall process of the present invention.

[0052] Figure 2 Schematic diagram of the process of dividing into real rainfall intervals.

[0053] Figure 3 Schematic diagram of the process of allocating capacitance weight coefficient and voltage weight coefficient to each rainfall station. DETAILED DESCRIPTION

[0054] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0055] A method for correcting and optimizing rainfall monitoring data, wherein the rainfall monitoring data is obtained through a capacitive rainfall monitoring sensor and a voltage rainfall monitoring sensor installed at a rainfall station, such as Figure 1 As shown, the specific steps are as follows:

[0056] S1. Set up sampling rainfall stations at random intervals within the modified optimization area.

[0057] In step S1, the spatial intervals between adjacent sampling rain gauges are as similar as possible, and rain gauges with special terrain and large terrain changes are sampled. The data samples obtained in this way can improve the accuracy of the deviation prediction model.

[0058] S2. Record the capacitance rainfall monitoring data, voltage rainfall monitoring data and actual rainfall data of the sampling rain station during several rainfall events in the future time period.

[0059] In step S2, the actual rainfall data is measured by a tipping bucket machine and recorded manually. During the recording period, the staff needs to regularly maintain the tipping bucket measuring machine to ensure its accuracy.

[0060] S3. According to the spatial coordinates of the rain gauge station and the actual rainfall data, a spatial capacitance deviation prediction model and a spatial voltage deviation prediction model of the rain gauge station are constructed.

[0061] S31. Divide the actual rainfall into several actual rainfall intervals, and for each rainfall station, calculate the average capacitance rainfall monitoring deviation and the average voltage rainfall monitoring deviation in each actual rainfall interval.

[0062] In step S31, the actual rainfall is divided into several actual rainfall intervals, such as Figure 2 As shown, the specific steps are as follows:

[0063] S311, dividing the maximum actual rainfall and the minimum actual rainfall into a preset number of initial intervals at equal intervals.

[0064] S312, calculating each initial interval, and calculating the variance of the capacitance rainfall monitoring deviation and the variance of the voltage rainfall monitoring deviation of all rain gauges in each initial interval.

[0065] S313. Adjust the upper and lower limits of the initial interval to minimize the sum of the variance of the capacitor rainfall monitoring deviation and the variance of the voltage rainfall monitoring deviation.

[0066] S314. When the number of iterations reaches the target, or the variance is less than a preset value, the true rainfall interval is obtained.

[0067] S32. Constructing training samples of a spatial capacitance deviation prediction model of a rainfall station or a spatial voltage deviation prediction model of a rainfall station based on the spatial coordinates of the rainfall station, the interval where the actual rainfall is located, and the average capacitance rainfall monitoring deviation or the average voltage rainfall monitoring deviation corresponding to the actual rainfall interval.

[0068] S33. Train the deep learning model with the training samples to obtain the rainfall station spatial capacitance deviation prediction model and the rainfall station spatial voltage deviation prediction model.

[0069] In step S33, the deep learning model is a BP neural network model, and the specific construction method is as follows:

[0070] Determine the number of nodes in the input layer. In this embodiment, the input layer is the coordinates of the rainfall station interval and the area where the actual rainfall is located;

[0071] Determine the number of hidden layer nodes;

[0072] Determine the number of output layer nodes. In this embodiment, the output layer is a capacitance deviation or a voltage deviation.

[0073] Initialize weights and biases;

[0074] Set the learning rate and number of iterations;

[0075] Perform iterative training and update weights and biases through the back-propagation algorithm;

[0076] Verify the model accuracy, precision, and recall.

[0077] S4. For each rainfall station, the capacitance weight coefficient and voltage weight coefficient of each rainfall station are allocated using the particle swarm optimization algorithm, such as Figure 3 As shown, the specific steps are as follows:

[0078] S41, randomly generate the position and velocity of particles;

[0079] S42, initializing the best individual position of each particle and the global best position;

[0080] S43, calculating the fitness value of each particle according to the fitness function;

[0081] The fitness function has the following specific formula:

[0082]

[0083] Among them, f(x i ) is the fitness value of the i-th rainfall station, j is the j-th real rainfall interval, there are n real rainfall intervals in total, is the capacitance weight coefficient of the ith rainfall station, is the voltage weight coefficient of the ith rainfall station, is the average capacitance rainfall monitoring data of the i-th rain gauge station in the j-th real rainfall interval, is the average voltage rainfall monitoring data of the ith rainfall station in the jth real rainfall interval.

[0084] S44, updating the individual best position and the global best position according to the fitness value;

[0085] S45, updating the speed and position of each particle;

[0086] S46, until the convergence condition is reached, stop iteration;

[0087] S47. Return the global best position and fitness value as the optimal solution.

[0088] S5. Optimize the capacitance rainfall monitoring data and voltage rainfall monitoring data recorded by each rainfall station in historical records using the capacitance weight coefficient and voltage weight coefficient corresponding to each rainfall station.

[0089] In step S5, the original rainfall monitoring data is 0.5 multiplied by the capacitance rainfall monitoring data, plus 0.5 multiplied by the voltage rainfall monitoring data, and the optimized and corrected rainfall monitoring data is the capacitance weight coefficient multiplied by the capacitance rainfall monitoring data, plus the voltage weight coefficient multiplied by the voltage rainfall monitoring data.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for optimizing rainfall monitoring data correction, characterized in that: The rainfall monitoring data is obtained through the capacitive rainfall monitoring sensor and the voltage rainfall monitoring sensor installed in the rainfall station. The specific method is as follows: Sampling rainfall stations were randomly set up at equal intervals within the revised optimization area; Record the capacitance rainfall monitoring data, voltage rainfall monitoring data and real rainfall data of the sampling rain gauge station during several rainfall events in the future time period; According to the spatial coordinates of the rain gauge station and the actual rainfall data, the spatial capacitance deviation prediction model and the spatial voltage deviation prediction model of the rain gauge station are constructed; For each rainfall station, the capacitance weight coefficient and voltage weight coefficient of each rainfall station are allocated by particle swarm optimization algorithm; The capacitance rainfall monitoring data and voltage rainfall monitoring data recorded by each rainfall station in historical records are optimized using the capacitance weight coefficient and voltage weight coefficient corresponding to each rainfall station.

2. The method for optimizing the correction of rainfall monitoring data according to claim 1, characterized in that: The rainfall station spatial capacitance deviation prediction model and the rainfall station spatial voltage deviation prediction model are both deep learning models. The specific construction methods are as follows: The real rainfall is divided into several real rainfall intervals, and for each rainfall station, the average capacitance rainfall monitoring deviation and the average voltage rainfall monitoring deviation in each real rainfall interval are calculated; Based on the spatial coordinates of the rain gauge station, the interval where the real rainfall is located, and the average capacitance rainfall monitoring deviation or the average voltage rainfall monitoring deviation corresponding to the real rainfall interval, a training sample of the spatial capacitance deviation prediction model of the rain gauge station or the spatial voltage deviation prediction model of the rain gauge station is constructed; The deep learning model is trained with training samples to obtain the spatial capacitance deviation prediction model and the spatial voltage deviation prediction model of the rainfall station.

3. The method for optimizing the correction of rainfall monitoring data according to claim 2, characterized in that: The actual rainfall is divided into several actual rainfall intervals. The specific method is as follows: Dividing the maximum true rainfall and the minimum true rainfall into a preset number of initial intervals at equal intervals; Calculate the variance of the capacitance rainfall monitoring deviation and the variance of the voltage rainfall monitoring deviation of all rainfall stations in each initial interval; Adjust the upper and lower limits of the initial interval to minimize the sum of the variance of the capacitance rainfall monitoring deviation and the variance of the voltage rainfall monitoring deviation; When the number of iterations reaches the target, or the variance is less than the preset value, the true rainfall interval is obtained.

4. The method for optimizing the correction of rainfall monitoring data according to claim 2, characterized in that: The deep learning model is a neural network model, and the specific construction method is as follows: Determine the number of nodes in the input layer, the input layer is the coordinates of the rainfall station interval and the area where the actual rainfall is located; Determine the number of hidden layer nodes; Determine the number of nodes in the output layer, the output layer is capacitance deviation or voltage deviation; Initialize weights and biases; Set the learning rate and number of iterations; Perform iterative training and update weights and biases through the back-propagation algorithm; Verify the model accuracy, precision, and recall.

5. The method for optimizing the correction of rainfall monitoring data according to claim 1, characterized in that: The particle swarm optimization algorithm is used to allocate the capacitance weight coefficient and voltage weight coefficient of each rainfall station. The specific method is as follows: Get the spatial coordinates of the rainfall station; The capacitance rainfall monitoring deviation and voltage rainfall monitoring deviation of the rainfall station in each real rainfall interval are predicted by using the rainfall station space capacitance deviation prediction model and the rainfall station space voltage deviation prediction model; The fitness function is constructed, and the capacitance weight coefficient and voltage weight coefficient are solved by particle swarm optimization algorithm.

6. The method for optimizing the correction of rainfall monitoring data according to claim 5, characterized in that: The particle swarm optimization algorithm is used to solve the capacitance weight coefficient and the voltage weight coefficient. The specific method is as follows: Randomly generate particle positions and velocities; Initialize the best individual position of each particle and the global best position; According to the fitness function, calculate the fitness value of each particle; According to the fitness value, update the individual best position and the global best position; Update the velocity and position of each particle; Stop iteration until convergence condition is reached; Returns the global best position and fitness value as the optimal solution.

7. The method for optimizing the correction of rainfall monitoring data according to claim 6, characterized in that: The fitness function has the following specific formula: Among them, f(x i ) is the fitness value of the i-th rainfall station, j is the j-th real rainfall interval, there are n real rainfall intervals in total, is the capacitance weight coefficient of the ith rainfall station, is the voltage weight coefficient of the ith rainfall station, is the average capacitance rainfall monitoring data of the i-th rain gauge in the j-th real rainfall interval, is the average voltage rainfall monitoring data of the ith rainfall station in the jth real rainfall interval.

8. The method for optimizing the correction of rainfall monitoring data according to claim 1, characterized in that: The optimized and corrected rainfall monitoring data is the capacitance weight coefficient multiplied by the capacitance rainfall monitoring data, plus the voltage weight coefficient multiplied by the voltage rainfall monitoring data.

Images