Mosquito killing equipment configuration method based on data driving and mosquito killing system

Abstract

The invention discloses a mosquito killing equipment configuration method based on data driving and a mosquito killing system, and relates to the technical field of mosquito control. The method comprises the following steps: obtaining an initial data set by obtaining the category and position information of buildings in an existing mosquito killing area, the category and position information of ancillary facilities, the position of mosquito killing equipment and mosquito catching amount information; processing the initial data set, and constructing an associated data set; building categories, building position information, ancillary facility categories and ancillary facility position information are used as input features, mosquito killing equipment positions are used as output features, and a neural network model is constructed; training the constructed neural network model; and inputting related data information of the current area of the to-be-configured mosquito killing equipment into the trained neural network model, and configuring the mosquito killing equipment in the current area according to an output result of the neural network model. According to the invention, the dependence on personal experience can be reduced, and unnecessary cost increase and waste are avoided while effective mosquito eradication is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of mosquito control, and in particular to a mosquito killing device configuration method and a mosquito killing system based on data driving. Background Art

[0002] At present, the location and number of mosquito control equipment mainly rely on the experience of the installers. This configuration method that relies purely on experience can easily lead to a series of problems. For example, when the number of mosquito control equipment is too small or the location is unreasonable, the mosquito control effect is poor; and when the number of mosquito control equipment is too large, it will cause excess mosquito control capacity, while increasing unnecessary equipment costs and energy waste. Summary of the invention

[0003] The object of the present invention is to provide a mosquito killing equipment configuration method based on data-driven, so as to reduce the dependence on personal experience and avoid unnecessary cost increase and waste while ensuring effective mosquito killing.

[0004] In order to achieve the above object, the present invention adopts the following technical solution: a mosquito killing device configuration method based on data drive, comprising the following steps:

[0005] Step 1: Obtain the building categories, location information of each building, categories of ancillary facilities, location information of each ancillary facility, installation location of mosquito control equipment and mosquito capture information of each mosquito control equipment in the existing mosquito control area to obtain an initial data set; using the mosquito capture in a unit time period as a reference standard, remove the mosquito control equipment-related data with a mosquito capture below a set threshold in the initial data set, and construct an associated data set of the remaining mosquito control equipment locations, building categories, building locations, ancillary facility categories, and ancillary facility locations;

[0006] Step 2: Using the building category, building location information, ancillary facility category, and ancillary facility location information as input features and the mosquito killing equipment location as output features, a neural network model is constructed;

[0007] Step 3: Use the associated data set constructed in step 1 to train the neural network model constructed in step 2;

[0008] Step 4: Input the building category, building location information, ancillary facility category, and ancillary facility location information of the current area where the mosquito repellent equipment is to be configured into the trained neural network model, and configure the mosquito repellent equipment for the current area according to the output results of the neural network model.

[0009] Further, the neural network model constructed in step two is a BP neural network model, which includes an input layer, a hidden layer, and an output layer. The input layer includes 4 neurons, corresponding to building category, building location information, accessory facility category, and accessory facility location information respectively. The hidden layer contains several neurons, and the output layer includes 1 neuron, corresponding to the location of the mosquito control device.

[0010] Further, the input and output calculation formulas of the hidden layer are as follows:

[0011] Input calculation:

[0012] Output calculation: h j = tanh(Z j )(2);

[0013] where, Z j is the input value of the j-th neuron in the hidden layer, W ij is the weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer, X i is the output value of the i-th neuron in the input layer, b j is the bias of the j-th neuron in the hidden layer, n is the number of neurons in the input layer, h j is the output value of the j-th neuron in the hidden layer, and tanh is the hyperbolic tangent function;

[0014] The input and output calculation formulas of the output layer are as follows:

[0015] Input calculation:

[0016] Output calculation:

[0017] where, Z k is the input value of the k-th neuron in the output layer, V jk is the weight from the j-th neuron in the hidden layer to the k-th neuron in the output layer, c k is the bias of the k-th neuron in the output layer, m is the number of neurons in the hidden layer, y k is the output value of the k-th neuron in the output layer, K is the number of neurons in the output layer, and Z m is the input value of the m-th neuron in the output layer.

[0018] Further, in step three, through the backpropagation of the training error, the weights and biases of the BP neural network model are corrected, and then the corrected weights and biases are substituted into the input and output calculation formulas of the hidden layer and the output layer to recalculate the output value of the output layer, so as to obtain the corrected output value. The specific steps are as follows:

[0019] (1) Calculate the error between the predicted value and the actual value of the output layer;

[0020] (2) Backpropagate the error to the hidden layer and calculate the error of each neuron;

[0021] (3) Update the weights and biases in the network according to the error and the learning rate;

[0022] (4) Use the corrected weights and biases to recalculate the input and output of the hidden layer;

[0023] (5) Use the corrected output of the hidden layer and the corresponding weights and biases to recalculate the input and output of the output layer to obtain the corrected output value.

[0024] Further, in step (1), the error between the predicted value and the actual value of the output layer is calculated by the following formula:

[0025] δ k = y k - t k (5);

[0026] where, δ k is the error of the k-th neuron in the output layer, y k is the output value of the k-th neuron in the output layer, that is, the predicted value, t k is the target value, that is, the actual value.

[0027] Further, in step (2), the error of each neuron in the hidden layer is calculated by the following formula:

[0028]

[0029] where, δ j is the error of the j-th neuron in the hidden layer, V jk is the weight from the j-th neuron in the hidden layer to the k-th neuron in the output layer, δ k is the error of the k-th neuron in the output layer, f′(Z j ) is the derivative of the activation function of the hidden layer, Z j is the input value of the j-th neuron in the hidden layer, and K is the number of neurons in the output layer.

[0030] Further, in step (3), the weights and biases in the network are updated by the following formula:

[0031] W′ ij = W ij - μ·δ j ·X i (7);

[0032] where, W′ ijis the updated weight of the hidden layer, W ij is the weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer, μ is the learning rate, δ j is the error of the j-th neuron in the hidden layer, X i is the output value of the i-th neuron in the input layer;

[0033] b′ j = b j - μ·δ j (8);

[0034] wherein, b′ j is the updated bias of the hidden layer, b j is the bias of the j-th neuron in the hidden layer, μ is the learning rate, δ j is the error of the j-th neuron in the hidden layer;

[0035] V′ jk = V jk - μ·δ k ·h j (9);

[0036] wherein, V′ jk is the updated weight of the output layer, V jk is the weight from the j-th neuron in the hidden layer to the k-th neuron in the output layer, μ is the learning rate, δ k is the error of the k-th neuron in the output layer, h j is the output value of the j-th neuron in the hidden layer;

[0037]

[0038] wherein, c′ k is the updated bias of the output layer, c k is the bias of the k-th neuron in the output layer, μ is the learning rate, δ k is the error of the k-th neuron in the output layer.

[0039] Furthermore, the auxiliary facilities include greening facilities, road facilities and auxiliary buildings. The greening facilities include pools, green belts, flower beds and lawns. The road facilities include sidewalks and driveways. The auxiliary buildings include pavilions and corridors.

[0040] Another object of the present invention is to provide a mosquito control system based on data driving, which configures mosquito control equipment according to the method described above.

[0041] Furthermore, the mosquito control system further includes a background control unit. The background control unit obtains meteorological data corresponding to the position of the current mosquito control area according to the position information of the current mosquito control area, and inputs the position information of the current mosquito control area, meteorological data, building categories within the mosquito control area, corresponding position information of the buildings, accessory facility categories, accessory facility position information, and mosquito control equipment installation position information into a mosquito quantity prediction model, and regulates the operating conditions of the mosquito control equipment in the corresponding mosquito control area according to the mosquito quantity prediction result output by the model; the mosquito quantity prediction model is obtained through the following method:

[0042] Step 1: Obtain the position information of all mosquito control areas within a set threshold range from the current mosquito control area, the building categories, building position information, accessory facility categories, accessory facility position information, mosquito control equipment installation positions within each mosquito control area, and the mosquito capture quantity information of each mosquito control equipment under different meteorological conditions to obtain a training data set;

[0043] Step 2: Construct a neural network model with the mosquito control area position, building category, building position, accessory facility category, accessory facility position, mosquito control equipment position, and meteorological data as input features and the mosquito capture quantity as the output feature, and train it with the training data set constructed in Step 1 to obtain a mosquito quantity prediction model.

[0044] Furthermore, the meteorological data includes temperature, humidity, and precipitation.

[0045] The present invention constructs an associated data set by obtaining relevant data of existing mosquito control areas, constructs a neural network model with building categories, position information, accessory facility categories, and position information as input features and mosquito control equipment positions as output features. After training with the associated data set, it can learn the complex mapping relationship between different environmental factors and the optimal positions of mosquito control equipment, so as to more accurately determine the positions of mosquito control equipment according to the actual situation of the area to be configured, avoid mosquito control blind spots caused by unreasonable positions, and effectively improve the mosquito control effect; at the same time, the model can also make the quantity of mosquito control equipment match the actual demand, avoid unnecessary equipment investment and energy consumption, and reduce equipment costs and operating costs. In addition, the present invention has strong popularization and adaptability. Whether it is a community, park in the city, or different scenarios such as rural areas and industrial parks, as long as the basic data such as building categories, position information, accessory facility categories, and position information of the area to be configured are obtained, the trained neural network model can be used for the configuration of mosquito control equipment. Moreover, as the mosquito control work continues and data accumulates, new data can be continuously added to the initial data set to further train and optimize the neural network model to make it more adaptable to changes in the actual situation and continuously improve the accuracy and effectiveness of mosquito control equipment configuration. Description of the Drawings

[0046] Figure 1 It is a flow schematic diagram of a method for configuring mosquito control equipment based on data-driven. Specific implementation manners

[0047] To facilitate those skilled in the art to better understand the improvements of the present invention over the prior art, the present invention will be further described below in conjunction with the drawings and embodiments.

[0048] Embodiment 1

[0049] The present invention provides a method for configuring mosquito control equipment, aiming to scientifically configure mosquito control equipment in a target area through a data-driven method, so as to reduce the dependence on personal experience, and while ensuring effective mosquito control, avoid unnecessary cost increase and waste.

[0050] The configuration method mainly includes the following steps (the general process can be seen in Figure 1 ):

[0051] 1. Data collection and preprocessing

[0052] 1.1 Obtain the initial data set:

[0053] Collect the building categories, the location information of each building, the accessory facility categories, the location information of each accessory facility, the installation location of mosquito control equipment, and the mosquito capture amount information of each mosquito control equipment in the existing mosquito control areas.

[0054] Among them, the accessory facilities include greening facilities, road facilities and accessory buildings, etc. For example, the greening facilities include pools, green belts, flower beds, lawns, etc., the road facilities include sidewalks, roadways, etc., and the accessory buildings include pavilions, corridors, etc.

[0055] 1.2 Data preprocessing:

[0056] Remove inefficient data: With the mosquito capture amount per unit time period as the reference standard, remove the data related to mosquito control equipment in the initial data set whose mosquito capture amount is lower than the set threshold.

[0057] Construct an associated data set: Construct an associated data set of the remaining mosquito control equipment locations, building categories, building locations, accessory facility categories, and accessory facility locations.

[0058] Encoding: Encode the category information in the associated data set and convert it into numerical data, and one-hot encoding or label encoding can be used.

[0059] 2. Determine the network structure

[0060] 2.1 Input layer: The number of neurons in the input layer should be consistent with the number of input features. In this embodiment, the building category, building location information, accessory facility category, and accessory facility location information are used as input features. Therefore, the input layer has 4 neurons.

[0061] 2.2 Hidden layer: The number of hidden layers and the number of neurons in each layer need to be adjusted according to the specific problem. For example, one hidden layer can be set, containing 10 neurons.

[0062] 2.3 Output layer: The number of neurons in the output layer should be consistent with the number of output variables. In this embodiment, the location of the mosquito control device is used as the output. Therefore, the output layer has 1 neuron.

[0063] 3. Determine the activation function

[0064] 3.1 Hidden layer activation function: Use the tanh or ReLU activation function to map the input to a smaller range, which helps the propagation of the gradient. In this embodiment, the tanh activation function is selected.

[0065] 3.2 Output layer activation function: Use the softmax activation function to map the output to the probability distribution of each category.

[0066] 4. Train the model using the associated dataset

[0067] 4.1 Loss function: Use the cross-entropy loss function to measure the difference between the predicted probability distribution and the actual probability distribution.

[0068] 4.2 Optimization algorithm: Select optimization algorithms such as gradient descent, momentum method, or Adam to update the weights and biases of the network.

[0069] 4.3 Training process:

[0070] 4.3.1 Initialization: Randomly initialize the weights and biases.

[0071] 4.3.2 Forward propagation:

[0072] 4.3.2.1 Input to the hidden layer:

[0073]

[0074] Among them, Z j is the input value of the j-th neuron in the hidden layer, W ij is the weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer, X i is the output value of the i-th neuron in the input layer, that is, one of the input features of the network, b j is the bias of the j-th neuron in the hidden layer, and n is the number of neurons in the input layer, that is, the number of input features.

[0075] 4.3.2.2 Output of the hidden layer (using the tanh (hyperbolic tangent) activation function):

[0076] h j = tanh(Z j );

[0077] where h j is the output value of the j-th neuron in the hidden layer, Z j is the input value of the j-th neuron in the hidden layer, and tanh is the hyperbolic tangent function that maps the input value Z j to a value between -1 and 1.

[0078] 4.3.2.3 Input to the output layer:

[0079]

[0080] where Z k is the input value of the k-th neuron in the output layer, V jk is the weight from the j-th neuron in the hidden layer to the k-th neuron in the output layer, h j is the output value of the j-th neuron in the hidden layer, i.e., the value after passing through the activation function of the hidden layer, c k is the bias of the k-th neuron in the output layer, and m is the number of neurons in the hidden layer.

[0081] 4.3.2.4 Output of the output layer (using the softmax activation function):

[0082]

[0083] where K is the number of neurons in the output layer, i.e., the number of classes, and y k is the output value of the k-th neuron in the output layer, representing the probability of the k-th class. Here, Z k is the input value of the k-th neuron in the output layer, and Z m is the input value of the m-th neuron in the output layer.

[0084] 4.3.3 Backpropagation:

[0085] 4.3.3.1 Error in the output layer:

[0086] δ k = y k - t k ;

[0087] where δ k is the error of the k-th neuron in the output layer, y k is the output value of the k-th neuron in the output layer, i.e., the predicted value of the network, and t kis the target value, i.e., the actual label or value.

[0088] 4.3.3.2 Error of the hidden layer:

[0089]

[0090] Among them, δ j is the error of the j-th neuron in the hidden layer, V jk is the weight from the j-th neuron in the hidden layer to the k-th neuron in the output layer, δ k is the error of the k-th neuron in the output layer, f′(Z j ) is the derivative of the activation function of the hidden layer, Z j is the input value of the j-th neuron in the hidden layer, and K is the number of neurons in the output layer.

[0091] 4.3.3.3 Update weights and biases:

[0092] Update of the hidden layer weights:

[0093] W′ ij = W ij - μ·δ j ·X i ;

[0094] Among them, W′ ij is the updated weight of the hidden layer, W′ ij is the weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer, μ is the learning rate, a hyperparameter that determines the step size of weight update, δ j is the error of the j-th neuron in the hidden layer, X i is the output value of the i-th neuron in the input layer, i.e., one of the input features of the network.

[0095] Update of the hidden layer bias:

[0096] b′ j = b j - μ·δ j ;

[0097] Among them, b′ j is the updated bias of the hidden layer, b j is the bias of the j-th neuron in the hidden layer, μ is the learning rate, a hyperparameter that determines the step size of bias update, δ j is the error of the j-th neuron in the hidden layer.

[0098] Update of the output layer weights:

[0099] V′ jk = V jk - μ·δ k ·hj ;

[0100] Among them, V' jk is the updated weight of the output layer, V jk is the weight from the j-th neuron in the hidden layer to the k-th neuron in the output layer, μ is the learning rate, a hyperparameter that determines the step size of weight update, δ k is the error of the k-th neuron in the output layer, h j is the output value of the j-th neuron in the hidden layer, that is, the value after passing through the activation function of the hidden layer.

[0101] Update of the output layer bias:

[0102]

[0103] Among them, c' k is the updated bias of the output layer, c k is the bias of the k-th neuron in the output layer, μ is the learning rate, a hyperparameter that determines the step size of bias update, δ k is the error of the k-th neuron in the output layer.

[0104] 5. Model Application

[0105] Input the building category, building location information, accessory facility category, and accessory facility location information of the target area where the mosquito control device is to be configured into the trained neural network model, and configure the mosquito control device for the target area according to the output result of the neural network model. Specifically, conduct on-site surveys or obtain the building category, building location information, accessory facility category, and accessory facility location information in the target area based on satellite maps, input this data into the trained BP neural network model, obtain the output result, and then configure the mosquito control device for the target area according to the output result.

[0106] Among them, the backpropagation process includes:

[0107] Calculate the error: First, calculate the error between the predicted value and the actual value in the output layer.

[0108] Error propagation: Propagate the error backward to the hidden layer and calculate the error of each neuron.

[0109] Weight and bias correction: Update the weights and biases in the network according to the error and the learning rate to reduce the prediction error.

[0110] Recalculate the output value:

[0111] Hidden layer input and output: Use the corrected weights and biases to recalculate the input and output of the hidden layer.

[0112] Output layer input and output: Use the corrected hidden layer output and the corresponding weights and biases to recalculate the input and output of the output layer.

[0113] Obtain the corrected output value: The output value of the final output layer will be closer to the actual value, thereby improving the prediction accuracy of the model.

[0114] Among them, the mosquito control equipment includes but is not limited to mosquito killing lamps. For example, a mosquito killing lamp with biological mosquito attracting and physical mosquito killing functions. Such mosquito killing lamps are usually equipped with a mosquito attracting lamp and a bionic inducer, as well as a vortex fan and a mosquito storage box. By attracting mosquitoes close to the mosquito killing lamp, the high-speed rotating vortex fan sucks the mosquitoes into the mosquito storage box and dries them to death.

[0115] To verify the feasibility of the above model, a simulation test was conducted in this embodiment, as follows.

[0116] Suppose there is an area and the following data is collected:

[0117]

[0118] I. Data preprocessing

[0119] Remove inefficient data:

[0120] Set the threshold of the mosquito capture amount to 50, and remove the records with the mosquito capture amount less than 50.

[0121] For example, remove the record with the mosquito capture amount of 50 and keep other records.

[0122] Encoding:

[0123] Perform one-hot encoding or label encoding on the building category, accessory facility category, and accessory facility subtype.

[0124] Standardize the building location information and accessory facility location information.

[0125] Perform label encoding on the mosquito control equipment location. For example, "entrance area" is encoded as 0, "corridor" is encoded as 1, and "public area" is encoded as 2.

[0126] Construct an associated dataset: Construct an associated dataset of the remaining mosquito control equipment location, building category, building location, accessory facility category, and accessory facility location.

[0127] Simulated dataset (after removing inefficient data):

[0128]

[0129] II. Construction of BP neural network model

[0130] Input layer: 6 neurons (building category, building location information, accessory facility category, accessory facility subtype, accessory facility location information).

[0131] Hidden layer: 1 hidden layer containing 10 neurons.

[0132] Output layer: 3 neurons (mosquito control equipment location).

[0133] III. Model Training

[0134] Use the constructed associated dataset to train the model, and update the weights and biases of the network through the backpropagation algorithm until the loss function converges or reaches the predetermined number of iterations.

[0135] IV. Prediction Simulation

[0136] Suppose there is the following new input data:

[0137]

[0138] V. Prediction Results

[0139] Input data: Input the above new input data into the trained BP neural network model.

[0140] Predicted output: The model outputs the predicted results of the installation location of the mosquito control equipment, for example:

[0141] For the first input (residential building, the accessory facility is a greening facility, and the subtype is a pool), the model predicts the installation location as "entrance area".

[0142] For the second input (commercial building, the accessory facility is a road facility, and the subtype is a driving lane), the model predicts the installation location as "public area".

[0143] VI. Result Analysis

[0144] Prediction accuracy: By comparing the model prediction results with the actual reasonable installation locations, it can be seen that the model has high accuracy.

[0145] Model feasibility: The prediction results of the model on multiple test samples are relatively accurate, indicating that the model has a certain feasibility and can reasonably predict the installation location of the mosquito control equipment according to the input features.

[0146] In addition, by outputting the installation location of the mosquito control device through a data-driven BP neural network model, various factors such as building category, building location information, accessory facility category, and accessory facility location information can be comprehensively considered. Compared with traditional empirical site selection, it is more scientific and accurate. By combining the installation location predicted by the model with the known mosquito control range and effect of the mosquito control device, effective mosquito control can be achieved. Specifically, the model can install the mosquito control device in areas where mosquitoes are active frequently and where breeding is suitable, for prevention and control at the source. Moreover, by understanding in advance the effective mosquito control range of the device, it can ensure that the coverage ranges of the devices at each installation point are connected to each other, avoid blind spots, maximize the overall efficiency of the mosquito control device, and improve the mosquito control efficiency.

[0147] It should be noted that the entrance area mentioned in the simulation test mainly refers to the import and export locations of the target area, that is, the main channels for people to enter or leave the area; the corridor mainly refers to the passage paths in the target area, that is, the main channels for people to move within the area; the public area mainly refers to the areas in the target area where people's activities are relatively concentrated, such as around leisure squares, viewing platforms, children's play areas, etc.

[0148] Generally speaking, in this embodiment, by obtaining the relevant data of the existing mosquito control areas to construct an associated data set, and using the building category, location information, accessory facility category, and location information as input features, and the mosquito control device location as the output feature, a neural network model is constructed. After being trained by the associated data set, it can learn the complex mapping relationship between different environmental factors and the optimal location of the mosquito control device, so as to more accurately determine the location of the mosquito control device according to the actual situation of the area to be configured, avoid mosquito control blind spots caused by unreasonable locations, and effectively improve the mosquito control effect; at the same time, the model can also make the number of mosquito control devices match the actual demand, avoid unnecessary equipment investment and energy consumption, and reduce equipment costs and operating costs. In addition, the method of this embodiment has strong popularization and adaptability. Whether it is a community, park in the city, or different scenarios such as rural areas and industrial parks, as long as the basic data such as the building category, location information, accessory facility category, and location information of the area to be configured are obtained, the trained neural network model can be used for the configuration of mosquito control devices. Moreover, as the mosquito control work continues and data accumulates, new data can be continuously added to the initial data set to further train and optimize the neural network model, making it more adaptable to changes in the actual situation and continuously improving the accuracy and effectiveness of mosquito control device configuration.

[0149] Embodiment 2

[0150] This embodiment provides a mosquito control system based on data driving. This system configures mosquito control devices according to the method in Embodiment 1. The system further includes a background control unit. The background control unit obtains meteorological data (including but not limited to temperature, humidity, precipitation, etc.) corresponding to the position of the current mosquito control area according to the position information of the current mosquito control area, and inputs the position information of the current mosquito control area, meteorological data, building categories in the mosquito control area, corresponding position information of buildings, accessory facility categories, accessory facility position information, and mosquito control device installation position information into a mosquito quantity prediction model, and regulates the operating conditions of mosquito control devices in the corresponding mosquito control area according to the mosquito quantity prediction result output by the model; the mosquito quantity prediction model is obtained through the following method:

[0151] Step 1: Obtain the position information of all mosquito control areas within a set threshold range from the current mosquito control area, building categories, building position information, accessory facility categories, accessory facility position information, mosquito control device installation positions, and mosquito capture quantity information of each mosquito control device under different meteorological conditions in each mosquito control area to obtain a training data set;

[0152] Step 2: Construct a neural network model with the mosquito control area position, building category, building position, accessory facility category, accessory facility position, mosquito control device position, and meteorological data as input features and the mosquito capture quantity as the output feature. After training it with the training data set constructed in Step 1, a mosquito quantity prediction model is obtained.

[0153] It should be noted that the neural network model (mosquito quantity prediction model) in this embodiment can be constructed and trained with reference to the neural network model (BP neural network model) in Embodiment 1. The principle is similar and will not be elaborated here.

[0154] In addition, the meteorological data of the current mosquito control area can not only adopt information such as temperature, humidity, and precipitation collected by local county or city meteorological stations, but also be surveyed on the spot. For example, according to monitoring requirements, suitable meteorological monitoring devices, including temperature sensors, humidity sensors, rain sensors, etc., are selected to monitor meteorological parameters in real time.

[0155] The mosquito control system in this embodiment can adjust the operating conditions of mosquito control devices according to the predicted mosquito quantity. For example, when the predicted mosquito quantity is lower than the set value, the mosquito control devices can be controlled to operate in a low-power state, and when the predicted mosquito quantity is higher than the set value, the corresponding mosquito control devices can be controlled to operate in a full-power state, so as to make the operation control of mosquito control devices more scientific and achieve the effect of energy saving.

[0156] The above is only a preferred embodiment of the present invention, and does not impose any form of limitation on the present invention. However, as long as the content does not deviate from the technical solution of the present invention, it still belongs to the patent scope of the technical solution of the present invention.

Claims

1. A mosquito killing equipment configuration method based on data drive, characterized in that: The following steps are involved: Step 1: Obtain the building categories, location information of each building, category of ancillary facilities, location information of each ancillary facility, installation location of mosquito control equipment and mosquito capture information of each mosquito control equipment in the existing mosquito control area to obtain an initial data set; using the mosquito capture in a unit time period as a reference standard, remove the mosquito control equipment related data with a mosquito capture below a set threshold in the initial data set, and construct an associated data set of the remaining mosquito control equipment locations, building categories, building locations, category of ancillary facilities, and location of ancillary facilities; Step 2: Using the building category, building location information, ancillary facility category, and ancillary facility location information as input features and the mosquito killing equipment location as output features, a neural network model is constructed; Step 3: Use the associated data set constructed in step 1 to train the neural network model constructed in step 2; Step 4: Input the building category, building location information, ancillary facility category, and ancillary facility location information of the current area where the mosquito repellent equipment is to be configured into the trained neural network model, and configure the mosquito repellent equipment for the current area according to the output results of the neural network model.

2. The method for configuring mosquito killing equipment based on data drive according to claim 1, characterized in that: The neural network model constructed in step 2 is a BP neural network model, which includes an input layer, a hidden layer and an output layer. The input layer includes 4 neurons, corresponding to the building category, building location information, ancillary facility category, and ancillary facility location information, respectively. The hidden layer contains several neurons, and the output layer includes 1 neuron, corresponding to the location of the mosquito killing equipment.

3. The data-driven mosquito killing equipment configuration method according to claim 2, characterized in that: The input and output calculation formulas of the hidden layer are as follows: Enter the calculation: Output calculation: h j =tanh(Z j )(2); Among them, Z j is the input value of the jth neuron in the hidden layer, W ij is the weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer, X i is the output value of the i-th neuron in the input layer, b j is the bias of the jth neuron in the hidden layer, n is the number of neurons in the input layer, and h j is the output value of the jth neuron in the hidden layer, and tanh is the hyperbolic tangent function; The input and output calculation formulas of the output layer are as follows: Enter the calculation: Output calculation: Among them, Z k is the input value of the kth neuron in the output layer, V jk is the weight from the jth neuron in the hidden layer to the kth neuron in the output layer, c k is the bias of the kth neuron in the output layer, m is the number of neurons in the hidden layer, and y k is the output value of the kth neuron in the output layer, K is the number of neurons in the output layer, and Z m is the input value of the mth neuron in the output layer.

4. The method for configuring mosquito killing equipment based on data drive according to claim 3, characterized in that: In step three, the weights and biases of the BP neural network model are corrected through the back propagation of the training error, and then the corrected weights and biases are substituted into the input and output calculation formulas of the hidden layer and the output layer, and the output value of the output layer is recalculated to obtain the corrected output value. The specific steps are as follows: (1) Calculate the error between the predicted value and the actual value of the output layer; (2) Back propagate the error to the hidden layer and calculate the error of each neuron; (3) Update the weights and biases in the network based on the error and learning rate; (4) Using the corrected weights and biases, recalculate the input and output of the hidden layer; (5) Using the corrected hidden layer output and the corresponding weights and biases, recalculate the input and output of the output layer to obtain the corrected output value.

5. The method for configuring mosquito killing equipment based on data drive according to claim 4, characterized in that: In step (1), the error between the predicted value and the actual value of the output layer is calculated by the following formula: δ k =y k -t k (5); Among them, δ k is the error of the kth neuron in the output layer, y k is the output value of the kth neuron in the output layer, i.e., the predicted value, t k is the target value, i.e. the actual value.

6. The method for configuring mosquito killing equipment based on data drive according to claim 4, characterized in that: In step (2), the error of each neuron in the hidden layer is calculated by the following formula: Among them, δ j is the error of the jth neuron in the hidden layer, V jk is the weight from the jth neuron in the hidden layer to the kth neuron in the output layer, δ k is the error of the kth neuron in the output layer, f′(Z j ) is the derivative of the hidden layer activation function, Z j is the input value of the jth neuron in the hidden layer, and K is the number of neurons in the output layer.

7. The method for configuring mosquito killing equipment based on data drive according to claim 4, characterized in that: In step (3), the weights and biases in the network are updated using the following formula: W′ ij =W ij -m·d j ·X i (7); Among them, W′ ij is the update weight of the hidden layer, W ij is the weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer, μ is the learning rate, δ j is the error of the jth neuron in the hidden layer, X i is the output value of the i-th neuron in the input layer; b′ j =b j -m·d j (8); Among them, b′ j is the update bias of the hidden layer, b j is the bias of the jth neuron in the hidden layer, μ is the learning rate, δ j is the error of the jth neuron in the hidden layer; V′ jk =V jk -m·d k ·h j (9); Among them, V′ jk is the updated weight of the output layer, V jk is the weight from the jth neuron in the hidden layer to the kth neuron in the output layer, μ is the learning rate, δ k is the error of the kth neuron in the output layer, h j is the output value of the jth neuron in the hidden layer; Among them, c′ k is the update bias of the output layer, c k is the bias of the kth neuron in the output layer, μ is the learning rate, and δ k is the error of the kth neuron in the output layer.

8. The method for configuring mosquito killing equipment based on data drive according to claim 1, characterized in that: The ancillary facilities include greening facilities, road facilities and ancillary buildings. The greening facilities include pools, green belts, flower beds and lawns. The road facilities include sidewalks and driveways. The ancillary buildings include pavilions and corridors.

9. Data-driven mosquito control system, characterized by: The mosquito killing device is configured according to the method described in any one of claims 1-8.

10. The data-driven mosquito killing system according to claim 9, characterized in that: It also includes a background control unit, which obtains meteorological data of a corresponding position according to the location information of the current mosquito killing area, and inputs the location information of the current mosquito killing area, meteorological data, building categories in the mosquito killing area, location information corresponding to the buildings, ancillary facility categories, ancillary facility location information, and mosquito killing equipment installation location information into a mosquito quantity prediction model, and adjusts the operating conditions of the mosquito killing equipment in the corresponding mosquito killing area according to the mosquito quantity prediction result output by the model; the mosquito quantity prediction model is obtained by the following method: Step 1, obtaining the location information of all mosquito-killing areas within a set threshold range from the current mosquito-killing area, the building category, building location information, ancillary facility category, ancillary facility location information, mosquito-killing equipment installation location, and mosquito capture information of each mosquito-killing equipment under different meteorological conditions, to obtain a training data set; Step 2: Using mosquito control area location, building category, building location, ancillary facility category, ancillary facility location, mosquito control equipment location and meteorological data as input features, and mosquito capture volume as output features, a neural network model is constructed. After training it using the training data set constructed in step 1, a mosquito population prediction model is obtained.

Images