Parameter optimization method and system based on variational quantum augmented bayesian network

Abstract

The application relates to a parameter optimization method and system based on a variational quantum enhanced Bayesian network, and relates to the field of quantum computing. The method comprises the following steps: obtaining an initial meteorological parameter value combination with the maximum lightning occurrence probability based on Bayesian network model traversal of historical meteorological parameter value combination data sets; mapping the initial meteorological parameter value combination into an initial bit string corresponding to an initial quantum state of a variational quantum circuit, running and measuring the variational quantum circuit to obtain a plurality of bit strings; parsing the bit strings into corresponding meteorological parameter value combinations; obtaining corresponding lightning occurrence probabilities based on the Bayesian network model; calculating a target function value based on the lightning occurrence probabilities, and adjusting the parameters of the variational quantum circuit until a convergence condition is met; and obtaining an optimal meteorological parameter value combination based on the converged variational quantum circuit. The application realizes optimization of meteorological parameter combinations with the maximum lightning occurrence probability as a constraint condition.

Description

Technical Field

[0001] This invention relates to the field of quantum computing, and in particular to a parameter optimization method and system based on variable quantum enhanced Bayesian networks. Background Technology

[0002] Quantum computing is a novel computing paradigm that utilizes the principles of quantum mechanics, possessing powerful computing capabilities and promising applications. Weather forecasting, the science of predicting future weather conditions using scientific methods and tools, is of great significance to people's daily lives and production activities. Lightning, in particular, is closely related to these activities. Lightning serves both as a scientific research subject, used to reveal the nature and effects of certain physical phenomena, and as a common testing tool in engineering, used to simulate extreme environments and test the resilience of equipment and systems. Furthermore, due to the enormous energy contained in lightning, precise control or simulation of its occurrence based on complex natural weather conditions is crucial in high-tech fields such as meteorology, aerospace safety, and environmental engineering. Typical applications range from artificial lightning and rainmaking operations to alleviating droughts or conducting scientific experiments, to constructing high-fidelity electromagnetic simulation environments for aircraft to test equipment safety, and even to using controlled high-energy discharges for industrial processing and materials handling, as well as revealing the physical nature of lightning through scientific research. In these scenarios, the core challenge lies in clarifying the nonlinear coupling relationship between lightning occurrence and multidimensional meteorological parameters—such as temperature, humidity, air pressure, wind speed, field strength, aerosol optical thickness, and cloud top / bottom height. However, traditional techniques often rely on historical experience or simple probabilistic models to adjust environmental meteorological parameters. Traditional methods struggle to capture the global dependencies between multiple variables, resulting in a degree of blindness in adjusting environmental meteorological parameters and consequently low efficiency. Summary of the Invention

[0003] To address the technical problems existing in the prior art, this invention proposes a parameter optimization method and system based on variable quantum augmented Bayesian networks. The optimization method includes the following steps:

[0004] Step S1: Obtain a dataset of historical meteorological parameter value combinations, which includes parameter values ​​for various meteorological conditions, as well as corresponding geographical location parameter values ​​and time parameter values.

[0005] Step S2: Based on the trained Bayesian network model, traverse the historical meteorological parameter value combination dataset and take the historical meteorological parameter value combination with the highest probability of lightning occurrence as the initial meteorological parameter value combination; wherein, the Bayesian network model is a model constructed and trained based on the probabilistic dependency between the meteorological parameter value combination and the probability of lightning occurrence.

[0006] Step S3: Set the initial values ​​of the variable quantum circuit parameters and map the initial meteorological parameter values ​​to the initial bit string corresponding to the initial quantum state of the variable quantum circuit.

[0007] Step S4: Run the variable quantum circuit to evolve the quantum state, and measure the final quantum state of the variable quantum circuit to obtain multiple bit strings corresponding to the final quantum state; wherein, the variable quantum circuit includes an initial state encoding module, a correlation module, a parameterized quantum gate module, and a measurement module. The initial state encoding module is used to encode the initial bit string into the initial quantum state of the variable quantum circuit; the correlation module includes multiple entanglement gates acting on the qubits representing any two meteorological parameters, used to apply entanglement between the multiple qubits representing meteorological parameters to establish a correlation between arbitrary meteorological parameters; the parameterized quantum gate module includes multiple layers of quantum circuit units connected end to end;

[0008] Step S5: Parse each bit string into a corresponding combination of meteorological parameter values;

[0009] Step S6: Input the combined meteorological parameter values ​​into the Bayesian network model to obtain the corresponding lightning occurrence probability;

[0010] Step S7: Calculate the objective function value based on the probability of lightning occurrence, wherein the objective function is a formula for calculating the probability of lightning occurrence;

[0011] Step S8: Adjust the variable quantum circuit parameters based on minimizing the objective function value;

[0012] Repeat steps S4 to S8 until the convergence condition of the variable quantum line is met, and then combine the optimal meteorological parameter values ​​based on the converged variable quantum line combination.

[0013] Optionally, the method further includes:

[0014] The training sample set for the Bayesian network model is constructed based on historical data. Each training sample includes a set of meteorological parameter values ​​and lightning occurrence parameter values. The meteorological parameter values ​​include parameter values ​​for various meteorological conditions, corresponding geographical location parameter values, and time parameter values.

[0015] A directed acyclic graph is constructed using meteorological parameters and lightning occurrence parameters as nodes, and the influence relationship between meteorological parameters and lightning occurrence as edges.

[0016] Construct a Bayesian network model based on the directed acyclic graph; and

[0017] The Bayesian network model is trained using the maximum likelihood estimation method based on the training sample set.

[0018] Optionally, the steps for constructing a training sample set for a Bayesian network model based on historical data include:

[0019] Obtain the raw lightning metadata dataset, where each piece of raw lightning metadata in the raw lightning metadata dataset is a three-dimensional grid data including geographical location, time, and lightning density;

[0020] The original lightning metadata of the three-dimensional grid is unfolded into one-dimensional lightning data, which includes geographical location, time and lightning density;

[0021] Obtain the raw meteorological dataset, wherein each piece of raw meteorological data in the raw meteorological dataset includes geographical location, time, and multiple meteorological condition parameter values;

[0022] Align lightning data and raw weather data according to geographic location and time;

[0023] The aligned lightning data and raw meteorological data are fused together to obtain primary training sample data; and

[0024] The primary training sample data is preprocessed, and lightning occurrence parameter data is constructed based on the lightning density data in the primary training sample to obtain the training sample data.

[0025] Optionally, the step of constructing the training sample set for the Bayesian network model further includes:

[0026] Based on the Monte Carlo method, a preset number of meteorological condition parameter values ​​are generated according to the data distribution characteristics of each meteorological condition; and

[0027] The meteorological condition parameter values ​​generated based on the Monte Carlo method are aligned with and fused with lightning data to obtain new training sample data.

[0028] Optionally, the step of mapping the initial combination of meteorological parameter values ​​to an initial bit string corresponding to the initial quantum state of the variable quantum circuit further includes:

[0029] Map each meteorological parameter value in the initial combination of historical meteorological parameter values ​​to a bit string of one or more binary bits;

[0030] The initial bit string is obtained by sequentially combining the bit strings corresponding to each meteorological parameter value;

[0031] The steps of mapping each meteorological parameter value to a bit string of one or more binary bits include:

[0032] Determine the parameter index for each meteorological parameter; and

[0033] Query the table that maps meteorological parameters to bit strings to obtain the corresponding bit strings.

[0034] Optionally, the method further includes the step of constructing a correspondence table between meteorological parameter indexes and bit strings:

[0035] When the value of the meteorological parameter is a continuous value, the first value range of the meteorological parameter is obtained, and the first value range is a continuous value interval.

[0036] The first numerical range is divided into N discrete numerical intervals, where N=2n; n is a natural number, and each discrete numerical interval is numbered to obtain the parameter index table of the meteorological parameters.

[0037] The parameter indices in the parameter index table are mapped sequentially into bit strings of length n in ascending order to obtain a correspondence table between parameter indices and bit strings;

[0038] When the value of the meteorological parameter is a discrete value, a second numerical range of the meteorological parameter is obtained, the second numerical range includes n discrete values, where n is a natural number;

[0039] Number the n discrete values ​​within the second numerical range to obtain a parameter index table for the meteorological parameters; and

[0040] Each parameter index in the parameter index table is mapped to a bit string of length n according to its position in the table; wherein, the position of each parameter index in the table is set to a first value, and the remaining positions are set to a second value. When the first value is 0, the second value is 1; when the first value is 1, the second value is 0.

[0041] Optionally, in step S5, the step of parsing the bit string into a combination of meteorological parameter values ​​includes:

[0042] The bit string is segmented to obtain bit substrings corresponding to each meteorological parameter;

[0043] Based on the bit substring of each meteorological parameter, query the table corresponding to the parameter index and bit string of each meteorological parameter to obtain the corresponding parameter index;

[0044] When the values ​​of meteorological parameters are continuous, the corresponding parameter range is obtained based on the parameter index;

[0045] A parameter value is determined within the parameter range based on a preset parameter value generation strategy;

[0046] When the meteorological parameter value is a discrete value, the corresponding discrete value is obtained based on the parameter index and used as the parameter value; and

[0047] The meteorological parameter values ​​are combined according to the order of the bit substrings in the bit string to obtain the meteorological parameter value combination corresponding to the bit string.

[0048] Optionally, the step of determining a parameter value within the parameter range based on a preset parameter value generation strategy includes:

[0049] Obtain the parameter value mapping function formula for the parameter interval of the corresponding meteorological parameter, wherein the parameter value mapping function formula includes an unknown variable;

[0050] Obtain the probability of the corresponding bit substring output by the current variable quantum circuit; and

[0051] The probability of the bit substring is assigned to the unknown variable in the mapping function formula to obtain the corresponding parameter value.

[0052] Optionally, the step of determining a parameter value within the parameter range based on a preset parameter value generation strategy includes:

[0053] Determine all sample parameter values ​​within the parameter range from historical data;

[0054] The probability of lightning occurrence is generated based on the Bayesian network model for each sample parameter value; and

[0055] The parameter value with the highest probability of lightning occurrence is determined as the parameter value.

[0056] Optionally, the method further includes:

[0057] The constraints for the meteorological parameter optimization problem are constructed, where the constraints are described as a lightning occurrence probability maximization problem using expression (1-1):

[0058] (1-1)

[0059] in, d represents an adjustable parameter of a variable quantum line, where d is a combination of meteorological parameter values. This represents the probability of lightning occurrence under the meteorological parameter value combination d.

[0060] The problem of maximizing the probability of lightning occurrence is mapped to a minimization problem, which is described by expression (1-2):

[0061] (1-2)

[0062] Construct the objective function of the minimization problem Its expression is shown in formula (1-3):

[0063] (1-3)

[0064] Where Y represents the set of all quantum states obtained in this measurement, and y represents the quantum state at each measurement. Indicates the variable component quantum circuit in parameters The probability distribution of the quantum state y obtained by time measurement, and d is the combination of meteorological parameter values ​​obtained by parsing the bit string x corresponding to the quantum state y. The probability value of lightning is obtained based on a Bayesian network model, and its expression is shown in formula (1-4):

[0065] (1-4)

[0066] Combining formula (1-4), formula (1-3) can be mapped to formula (1-5):

[0067] (1-5)

[0068] Wherein, j represents the measurement order, and M represents the total number of measurements; The combination of meteorological parameter values ​​is obtained by analyzing the bit string xj corresponding to the quantum state yj obtained from the j-th measurement.

[0069] Correspondingly, in step S7, the objective function value is calculated based on formula (1-5).

[0070] According to another aspect of the present invention, the present invention also provides a parameter optimization system based on a variable quantum augmented Bayesian network, comprising a dataset acquisition module, a Bayesian network module, an encoding module, a variable quantum circuit module, a bit string parsing module, and an optimization module; wherein, the dataset acquisition module acquires a dataset of historical meteorological parameter value combinations, the historical meteorological parameter value combinations including parameter values ​​of various meteorological conditions and corresponding geographical location parameter values ​​and time parameter values; the Bayesian network module, at the start of optimization, traverses the historical meteorological parameter value combination dataset based on the trained Bayesian network model, selecting the historical meteorological parameter value combination dataset with the highest probability of lightning occurrence. Historical meteorological parameter values ​​are used as the initial meteorological parameter value combination. During the optimization process, the corresponding lightning occurrence probability is generated based on the meteorological parameter value combination from the bit string parsing module. The Bayesian network model is constructed based on the probabilistic dependency between the meteorological parameter value combination and the lightning occurrence probability. The encoding module maps the initial meteorological parameter value combination to an initial bit string corresponding to the initial quantum state of the variable quantum circuit. The variable quantum circuit module runs the variable quantum circuit based on the set variable quantum circuit parameter values ​​to perform quantum state evolution and measures the final quantum state of the variable quantum circuit to obtain multiple... The final quantum state corresponds to a bit string, wherein the variable quantum circuit parameters are initial values ​​at the start of optimization, and during optimization, the variable quantum circuit parameters are parameter values ​​determined by the tuning module; wherein the variable quantum circuit includes an initial state encoding module, a correlation module, a parameterized quantum gate module, and a measurement module; the initial state encoding module is used to encode the initial bit string into the initial quantum state of the variable quantum circuit; the correlation module includes multiple entanglement gates acting on qubits representing any two meteorological parameters, used to apply entanglement between multiple qubits representing meteorological parameters to establish a correlation between arbitrary meteorological parameters; the parameterization... The quantum gate module includes multiple layers of interconnected quantum circuit units; the bit string parsing module parses the bit string obtained by measuring the terminal quantum states of the variable quantum circuit into a combination of meteorological parameter values; the optimization module sets the variable quantum circuit parameter values ​​as initial values ​​at the start of optimization, and during the optimization process, calculates the objective function value based on the lightning occurrence probability generated by the Bayesian network module, wherein the objective function is a formula for calculating the lightning occurrence probability, and adjusts the variable quantum circuit parameters by minimizing the objective function value until the convergence condition of the variable quantum circuit is met; wherein the optimal combination of meteorological parameter values ​​is obtained based on the converged variable quantum circuit.

[0071] According to another aspect of the present invention, the present invention also provides an electronic device, including a processor and a memory, wherein the memory stores computer instructions, and the processor executes the aforementioned parameter optimization method based on a variable quantum augmented Bayesian network when running the computer instructions.

[0072] According to another aspect of the present invention, the present invention also provides a computer-readable storage medium storing computer instructions that, when executed by a processor, perform the aforementioned parameter optimization method based on a variable quantum augmented Bayesian network.

[0073] The parameter optimization method and system based on variable quantum augmented Bayesian networks proposed in this invention upgrades scenario application from passive "experience simulation" to active "parameter optimization", avoiding the blind setting of parameter values ​​and improving the efficiency of solving scenario problems. Attached Figure Description

[0074] The preferred embodiments of the present invention will now be described in further detail with reference to the accompanying drawings, wherein:

[0075] Figure 1 This is a flowchart of a parameter optimization method based on a variable quantum augmented Bayesian network according to an embodiment of the present invention;

[0076] Figure 2 This is a flowchart of a method for constructing and training a variable quantum augmented Bayesian network according to an embodiment of the present invention;

[0077] Figure 3 This is a schematic diagram of a directed acyclic graph (DAG) structure according to an embodiment of the present invention;

[0078] Figure 4 This is a schematic diagram of a variable quantum circuit according to an embodiment of the present invention;

[0079] Figure 5 This is a flowchart of a method for constructing a parameter index table and a parameter index and bit string correspondence table for a continuous numerical meteorological parameter according to an embodiment of the present invention;

[0080] Figure 6 This is a flowchart of a method for constructing a parameter index table and a parameter index and bit string correspondence table for a discrete numerical meteorological parameter according to an embodiment of the present invention;

[0081] Figure 7 This is a flowchart of a method for parsing a measured bit string into a combination of meteorological parameter values ​​according to an embodiment of the present invention;

[0082] Figure 8 This is a block diagram of a parameter optimization system based on a variable quantum augmented Bayesian network according to an embodiment of the present invention.

[0083] Figure 9 This is a schematic diagram illustrating the statistical relationship between precipitation intensity and lightning density based on a dataset of historical meteorological parameter values.

[0084] Figure 10 This is a schematic diagram illustrating the statistical relationship between air pressure and lightning density based on a dataset of historical meteorological parameter values.

[0085] Figure 11 This is a schematic diagram illustrating the statistical relationship between cloud top height and lightning density based on a dataset of historical meteorological parameter values.

[0086] Figure 12 This is a structural principle block diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0087] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0088] In the following detailed description, reference can be made to the accompanying drawings, which form part of this application and illustrate specific embodiments of the present application. In the drawings, similar reference numerals describe substantially similar components in different figures. Specific embodiments of the present application are described in sufficient detail below to enable those skilled in the art to implement the technical solutions of the present application. It should be understood that other embodiments may also be utilized, or structural, logical, or electrical changes may be made to the embodiments of the present application.

[0089] This invention takes the meteorological system described by various meteorological conditions, geography, time, etc. as the research target, and uses formula (1-1) to describe the constraint condition of the combination of meteorological parameter values ​​that this invention seeks: maximizing the probability of lightning occurrence.

[0090] (1-1)

[0091] in, d represents an adjustable parameter in the system, where d is a combination of meteorological parameter values. Let d be the probability of lightning occurring when the meteorological parameter value combination is d. To find the meteorological parameter value combination that maximizes the probability of lightning occurrence, this invention provides a variable quantum augmented Bayesian network. This network uses the influencing factors of the meteorological system (referred to as meteorological parameters) as variables and optimizes the variable combinations and corresponding variable values.

[0092] See Figure 1 , Figure 1This is a flowchart of a parameter optimization method based on a variable quantum augmented Bayesian network according to an embodiment of the present invention. The optimization method includes the following steps:

[0093] Step S1: Obtain a dataset of historical meteorological parameter value combinations, which includes parameter values ​​for various meteorological conditions, as well as corresponding geographical location parameter values ​​and time parameter values.

[0094] Step S2: Based on the Bayesian network model, traverse the historical meteorological parameter value combination dataset to obtain the initial meteorological parameter value combination, wherein the initial meteorological parameter value combination is the historical meteorological parameter value combination with the highest probability of lightning occurrence; the Bayesian network model is a model constructed based on the probabilistic dependency relationship between the meteorological parameter value combination and the probability of lightning occurrence.

[0095] Step S3: Set the initial values ​​of the variable quantum circuit parameters and map the initial meteorological parameter values ​​to the initial bit string corresponding to the initial quantum state of the variable quantum circuit.

[0096] Step S4: Run the variable quantum circuit and perform final state measurement. Specifically, the final quantum state of the variable quantum circuit is measured to obtain multiple bit strings.

[0097] Step S5: Parse each bit string into a combination of meteorological parameter values.

[0098] Step S6: Input the combined meteorological parameter values ​​into the Bayesian network model to obtain the corresponding lightning occurrence probability.

[0099] Step S7: Calculate the objective function value based on the probability of lightning occurrence, wherein the objective function is a formula for calculating the probability of lightning occurrence.

[0100] Step S8: Determine whether the convergence condition of the variable quantum circuit is met. If the convergence condition of the variable quantum circuit is met, proceed to step S10; otherwise, proceed to step S9.

[0101] Step S9: Adjust the parameters of the variable quantum circuit and return to step S4.

[0102] Step S10: Determine the optimal combination of meteorological parameter values ​​based on the convergent variable quantum circuit.

[0103] The historical meteorological parameter value combination dataset in step S1 includes multiple data entries, each containing a set of meteorological parameter values. These meteorological parameters include geographical location parameters, time parameters, and various meteorological condition parameters. In one embodiment, the historical meteorological parameter value combination dataset comes from a satellite meteorological dataset. Examples of satellite meteorological datasets include global meteorological datasets provided by MODIS (Moderate-resolution image spectroradiometer) aboard Aqua (morning satellite) or Terra (afternoon satellite), and global meteorological datasets collected by AIRS (Atmospheric Infrared Sounder) aboard Aqua satellite. Each meteorological data entry in the global meteorological dataset includes geographical, temporal, and meteorological condition parameter values. Meteorological condition parameters in each piece of raw meteorological data include, for example, temperature, humidity, air pressure, wind speed, field strength, aerosol optical thickness, cloud top height, cloud base height, precipitation intensity, etc.

[0104] After obtaining the original satellite meteorological dataset, the data is preprocessed to obtain an initial dataset of historical meteorological parameter values. This includes replacing infinity values ​​in the original data with missing values, padding missing data with zeros, and standardizing data types.

[0105] To obtain sufficient data, this invention also generates a preset amount of meteorological condition data according to the distribution characteristics of each meteorological condition using the Monte Carlo method, and merges it into the initial historical meteorological parameter value combination dataset, thereby obtaining a historical meteorological parameter value combination dataset with a sufficient amount of data. The following are the data distribution characteristics and parameter values ​​used when generating meteorological condition parameters such as temperature, humidity, air pressure, wind speed, aerosol optical thickness, cloud top thickness, electric field intensity, and precipitation intensity.

[0106] Temperature: Normal distribution, mean 20℃, standard deviation 5℃.

[0107] Humidity: Beta distribution, parameters 2 and 5, with values ​​ranging from 0 to 100%.

[0108] Air pressure: Normal distribution, mean 1013 hPa, standard deviation 10 hPa.

[0109] Wind speed: Weibull distribution, parameter 2, amplification factor 5, range 0-10 m / s.

[0110] Aerosol optical thickness: Gamma distribution, parameter 2, scale parameter 0.1, dimensionless.

[0111] Cloud top thickness: uniformly distributed, ranging from 0 to 15 km.

[0112] Electric field strength: uniformly distributed, ranging from 100kV / m to 1MV / m.

[0113] Precipitation intensity: exponential distribution, scale parameter 1 mm / h-50 mm / h.

[0114] In the historical meteorological parameter value combination dataset, the meteorological conditions in each data point can include all of the aforementioned types or any of them.

[0115] This invention employs a variable quantum augmented Bayesian network, see details below. Figure 2 , Figure 2 This is a flowchart of a method for constructing and training a variable quantum augmented Bayesian network according to an embodiment of the present invention, specifically including the following steps:

[0116] Step S21: Construct a training sample set based on historical data. Each training sample includes a set of meteorological parameter values ​​and lightning occurrence parameter values. The meteorological parameters include various different meteorological condition parameters, as well as corresponding geographical location parameters and time parameters.

[0117] Step S22: Construct a directed acyclic graph. Specifically, a directed acyclic graph is constructed using each meteorological parameter and lightning occurrence parameter as nodes, and the influence relationship between the meteorological parameters and lightning occurrence as edges.

[0118] Step S23: Construct a Bayesian network model based on the directed acyclic graph.

[0119] Step S24: Train the Bayesian network model using the maximum likelihood estimation method based on the training sample set.

[0120] Step S25: Construct a variable quantum circuit, in which a meteorological parameter corresponds to one or more qubits.

[0121] In step S21, when constructing the training sample set, the original lightning metadata dataset is first obtained. In one embodiment, the original lightning metadata dataset in this invention comes from WWLLN (World Wideband Lightning Location Network), where each piece of original lightning metadata records the lightning density value for different geographical locations (including latitude and longitude) and different months. The main variables and structure of the original lightning metadata dataset are as follows:

[0122] lat, lon: latitude coordinates, longitude coordinates.

[0123] mon: Month index.

[0124] stroke_density: Lightning density value.

[0125] Since each piece of raw lightning metadata is a value on a lon × lat × mon three-dimensional grid, after obtaining the raw lightning metadata dataset, the raw lightning metadata of the three-dimensional grid is unfolded into one-dimensional lightning data. Specifically, the raw lightning metadata of the three-dimensional grid is decomposed into lightning data distributed according to time and geography. A lightning data table (lightning_data) is constructed by unfolding the three-dimensional grid in the lightning metadata set, which includes latitude coordinates, longitude coordinates, month, and lightning density values ​​(Stroke_density).

[0126] The lightning density value is then mapped to the parameter value of the lightning occurrence parameter. Specifically, when the lightning density parameter value is greater than 0, the corresponding lightning occurrence parameter value is 1, indicating that lightning has occurred; when the lightning density parameter value is equal to 0, the corresponding lightning occurrence parameter value is 0, indicating that no lightning has occurred.

[0127] Then, the lightning data and the data in the historical meteorological parameter value combination dataset obtained in step S1 are aligned and fused according to geographical location and time to obtain training sample data. In this embodiment, since the data in the historical meteorological parameter value combination dataset is aligned with the lightning data in terms of time and geographical distribution, the two types of data can be combined according to the same time and the same geographical distribution to complete the fusion of the two types of data and obtain the training sample set.

[0128] See Table 1 below, which contains the first 10 sample data from the training sample set in one embodiment.

[0129] Table 1

[0130]

[0131] In this table, lat and lon in the header represent latitude and longitude coordinates, respectively, in degrees. mon represents the month, s_d represents stroke density, temp represents temperature, humi represents humidity, press represents air pressure, w_s represents wind speed, a_o_t represents aerosol optical thickness, c_t_h represents cloud top height, c_b_h represents cloud base height, p_i represents precipitation intensity, f_s represents field strength, and l_o represents lightning occurrence. This example includes nine meteorological conditions; however, it should be noted that the data in this example is for illustrative purposes only, and the sample can certainly include other types of meteorological conditions, and the number is not limited to nine, but can be more than nine.

[0132] See Figure 3 , Figure 3 This is a schematic diagram of a Directed Acyclic Graph (DAG) structure according to an embodiment of the present invention. In this embodiment, in the DAG constructed in step S22, each meteorological condition parameter, geographical location parameter, and time parameter is a direct independent parent node of the lightning occurrence parameter node. In this embodiment, the 11 environmental parameter nodes are arranged in ascending order of node number n as 9 meteorological condition parameters: temperature (TP), air pressure (P), wind speed (WS), field strength (FS), aerosol optical thickness (AOT), cloud top height (CTH), cloud base height (CBH), precipitation intensity (PI), humidity (H), time (T), and geographical location (G).

[0133] In step S23, a corresponding Bayesian network model is constructed based on the DAG in step S22. The Bayesian network model calculates the corresponding conditional probabilities according to formula (3-1):

[0134] (3-1)

[0135] Where P(A|B) is the posterior probability, also known as the conditional probability, representing the probability of event A occurring given that event B has occurred. P(B|A) is the likelihood function, representing the probability of event B occurring given that event A has occurred. P(A) is the prior probability, representing the belief or probability estimate of event A given that event B has not occurred. P(B) is the marginal probability of event B, which is the sum of the probabilities of event B occurring in all possible scenarios, and is usually used as a normalization constant.

[0136] In this invention, event A represents lightning occurrence event L, and event B represents a combination of meteorological parameter values.

[0137] In step S24, the Bayesian network model is trained using maximum likelihood estimation (MLE) based on the training sample set to obtain the model parameters. The parameters of the Bayesian network model are the prior probability of lightning occurring, the prior probability of lightning not occurring, and the conditional probability distributions corresponding to different combinations of meteorological parameter nodes when lightning occurs and when lightning does not occur.

[0138] When training a Bayesian network model, the model calculates the probability of lightning occurrence under various combinations of meteorological parameter values ​​based on a given training sample set according to formula (3-1), and adjusts the parameters of the Bayesian network based on the maximum likelihood value, thereby learning the conditional probability distribution of the combination of meteorological parameter values ​​to the probability of lightning occurrence. The maximum likelihood estimation method (MLE) is a standard algorithm in this field and will not be elaborated upon here.

[0139] See Figure 4 , Figure 4 This is a schematic diagram of a variable quantum circuit according to an embodiment of the present invention. One or more qubits in the variable quantum circuit correspond to a meteorological parameter. The variable quantum circuit includes an initial state encoding module, a correlation module, a parameterized quantum gate module, and a measurement module. The initial state encoding module is used to encode an initial bit string into the initial quantum state of the variable quantum circuit. The initial state encoding module can be any quantum circuit that encodes binary numbers into quantum states in the present technology. The correlation module includes multiple entanglement gates acting on qubits representing any two meteorological parameters. These entanglement gates, such as CNOT gates, are used to apply entanglement between the multiple qubits representing the meteorological parameters, thereby establishing a correlation between arbitrary meteorological parameters and simulating the relationship between environmental conditions and lightning occurrence.

[0140] The parameterized quantum gate module comprises multiple layers of interconnected quantum circuit units. For example, each layer of quantum circuit units includes a quantum gate Rz(α), Rx(α), or Ry(α) acting on each qubit, or any combination of these types of quantum gates. The rotation angle α of the quantum gate is an adjustable parameter in the quantum circuit.

[0141] In step S2, when traversing the historical meteorological parameter value combination dataset based on the trained Bayesian network model, a lightning occurrence probability is obtained for each data point. In one embodiment, the historical meteorological parameter value combination with the highest lightning occurrence probability is determined as the initial meteorological parameter value combination for optimization. However, it is also possible to use several historical meteorological parameter value combinations with the highest lightning occurrence probability as initial meteorological parameter value combinations, obtain an optimal meteorological parameter value combination based on each initial meteorological parameter value combination, and then determine one of the multiple optimal meteorological parameter value combinations as the final output. For example, each optimal meteorological parameter value combination is input into the Bayesian network model to obtain the corresponding lightning occurrence probability, and the meteorological parameter value combination with the highest lightning occurrence probability is used as the final output.

[0142] In step S3, when mapping the initial historical meteorological parameter value combination obtained from the Bayesian network model to the initial bit string corresponding to the initial quantum state of the variable quantum circuit, each meteorological parameter value in the initial historical meteorological parameter value combination is first mapped to a bit string of one or more binary bits, and then these bit strings are sequentially combined to obtain the initial bit string. In this invention, a parameter index table and a correspondence table between the parameter index table and the bit string are constructed for each meteorological parameter. When the parameter value of the meteorological parameter is a continuous value, such as temperature, air pressure, etc., the method for constructing the parameter index table and the correspondence table between the parameter index table and the bit string is described in [reference needed]. Figure 5 , Figure 5 This is a flowchart illustrating a method for constructing a parameter index table and a parameter index-bit string correspondence table for a continuous numerical meteorological parameter according to an embodiment of the present invention. The method includes the following steps:

[0143] Step S301: Obtain the numerical range of the meteorological parameters. Taking the temperature parameter as an example, determine the numerical range of the temperature parameter value as [Tmin, Tmax].

[0144] Step S302: Divide the numerical range into N discrete parameter intervals, satisfying N=2 n n is a natural number.

[0145] Step S303: Number each discrete parameter interval to obtain a parameter index table for the meteorological parameters. For example, the i-th parameter interval is represented as: [Tmin+i•ΔT, Tmin+(i+1)•ΔT]. Where ΔT=(Tmin+Tmax) / N. The parameter index table is, for example, [0,1,2,……,(N-1)].

[0146] Step S304: Construct a mapping table between parameter indices and bit strings. Specifically, the parameter indices in the parameter index table are mapped sequentially to bit strings of length n in ascending order. For example, for a parameter divided into 8 parameter intervals, when the length of the bit string n=3, the correspondence between the parameter index and the bit string is shown in Table 2 below:

[0147] Table 2

[0148]

[0149] Therefore, when mapping one meteorological parameter value in the initial historical meteorological parameter value combination to a bit string of one or more binary bits, if the meteorological parameter value is a continuous value, the corresponding discrete parameter range and the corresponding parameter index are first determined based on its current parameter value, and then the corresponding bit string is obtained by querying the correspondence table between the parameter index and the bit string based on the parameter index.

[0150] The meteorological parameter values ​​in this invention may also be discrete numerical types, such as time parameters. The time parameters include 12 discrete values, representing 12 values ​​from January to December. See also... Figure 6 , Figure 6 This is a flowchart illustrating a method for constructing a parameter index table and a parameter index-bit string correspondence table for a discrete numerical meteorological parameter according to an embodiment of the present invention. The method includes the following steps:

[0151] Step 311: Obtain the numerical range of the meteorological parameter, which includes n discrete values, where n is a natural number. For the time parameter, n=12.

[0152] Step 312 involves numbering multiple discrete values ​​within the numerical range to obtain a parameter index table for the meteorological parameters. For example, for time parameters, the parameter index table might be {1, 1, ..., 12}. Each parameter index corresponds to a numerical value.

[0153] Step 313: Map each parameter index in the parameter index table to a bit string of length n according to its position in the table. For example, for a parameter index, set the position of each parameter index in the table to 1 or 0, and set the other positions to 0 or 1 respectively, thus obtaining the corresponding bit string. For example, as shown in Table 3, for the time parameter, the parameter index table has a length of 12, and each month corresponds to a parameter index. The first bit of the bit string for January is 1, and the following 11 bits are 0.

[0154] Table 3

[0155]

[0156] Therefore, when mapping a meteorological parameter value from the initial combination of historical meteorological parameter values ​​to a bit string of one or more binary bits, if the meteorological parameter value is a discrete value, the parameter index corresponding to the target meteorological parameter value is determined based on the parameter index table. For example, when the time parameter value is June, the corresponding parameter index is 6. Then, the corresponding bit string is obtained by querying the correspondence table between the parameter index table and the bit string table. The bit string for the time parameter value of June is: 00000010 0000.

[0157] According to the aforementioned Figure 5 and Figure 6 The method shown converts each meteorological parameter value into a bit string, and then combines these bit strings sequentially to form a bit string corresponding to a combination of meteorological parameter values. Thus, through... Figure 4 The initial state encoding module encodes the bit strings corresponding to the initial meteorological parameter values ​​into the variable quantum circuit.

[0158] In step S4, when measuring the final quantum state of the variable quantum circuit, each measurement yields a bit string x corresponding to the final quantum state. Therefore, after M measurements, M bit strings x are obtained, denoted as X. i ={x i 1,x i 2,……,x i M}. Where i is a natural number, represents the number of optimization attempts, and M represents the number of measurements, for example: M=1000.

[0159] In step S5, when parsing each bit string into the corresponding combination of meteorological parameter values, since there are multiple repeated bit strings in the m bit strings x, for example, for a quantum circuit with 4 quantum bits, when the number of measurements is sufficient, 2 can be obtained. 4 There are 2 bit strings, {0000}, {0001}, ..., {1111}, where each bit string has a different number of bits. Therefore, when parsing the bit strings, only 2 bits need to be parsed. 4 A single bit string is sufficient.

[0160] See Figure 7 , Figure 7 This is a flowchart of a method for parsing a measured bit string into a combination of meteorological parameter values ​​according to an embodiment of the present invention. The method for parsing a measured bit string into a combination of meteorological parameter values ​​includes the following steps:

[0161] Step S51: The bit string is split to obtain bit substrings corresponding to each meteorological parameter.

[0162] Step S52: Determine a meteorological parameter as the target meteorological parameter, and use the corresponding bit substring as the target bit substring.

[0163] Step S53: Query the correspondence table between the parameter index of the target meteorological parameter and the bit string to obtain the parameter index of the corresponding target bit substring.

[0164] Step S54: Determine whether the target meteorological parameter value is a continuous numerical type. If the target meteorological parameter value is a continuous numerical type, proceed to step S551. If the target meteorological parameter value is a discrete numerical type, proceed to step S552.

[0165] Step S551: Based on the parameter index, query the numerical parameter interval table of the target meteorological parameter to obtain the corresponding numerical parameter interval.

[0166] Step S561: Determine a parameter value within the numerical parameter range based on a preset parameter value generation strategy.

[0167] Step S552: Query the parameter index table of the target meteorological parameters to obtain the discrete values ​​corresponding to the parameter indexes.

[0168] Step S57: Determine if there are any unparsed bit substrings of meteorological parameters. If so, return to step S52; otherwise, combine all parsed parameter values ​​into a single meteorological parameter value combination. This meteorological parameter value combination can be, for example, a structured array such as D={temp:T1; humi:H1;……; f_s:FS1}, where each meteorological parameter and its value are separated from other meteorological parameters and their values ​​by semicolons. The part before the colon is the meteorological parameter name, and the part after the colon is the meteorological parameter value. Alternatively, the meteorological parameter value combination can be a group of parameter values, such as D={T1, H1,……,FS1}, which is a sequentially arranged combination of meteorological parameter values, where each data point represents the value of a meteorological parameter.

[0169] In step S561, there can be multiple strategies for generating parameter values. For example, each meteorological parameter has a corresponding parameter value mapping function formula, which includes an unknown variable. For example, for the meteorological parameter temperature, its parameter value mapping function formula is shown in equation (4-1):

[0170] T=T low +α•(T high -T low (4-1)

[0171] Wherein: T high and T low These represent the two endpoints of the parameter range, for example, the parameter range can be represented as: [Thigh ,T low ), such as: [T high , T low = [20, 30). α∈[0,1), is an unknown variable. In this embodiment, the probability of the corresponding bit substring is calculated from the currently obtained M bit strings. For example, when the number of measurements M=1000, 1000 bit strings are obtained. The first two bits of each bit string correspond to the temperature parameter. The first two bits of the currently parsed bit string are 00. Then, the number of bit strings with the first two bits being 00 is counted in the 1000 bit strings. The counted number of bit strings is divided by 1000 to obtain the probability of the bit substring being 00. The probability is the value of α. Thus, the specific temperature parameter value is obtained according to formula (4-1).

[0172] For example, all sample parameter values ​​within the parameter range are determined from historical data. For example, the parameter range for temperature parameter values ​​is [20, 30). All data within this range are found from the historical meteorological parameter value combination dataset as samples. These samples are then sequentially input into the Bayesian network model to obtain the lightning occurrence probability for each sample. The parameter value of the sample with the highest lightning occurrence probability is determined as the parameter value corresponding to the bit substring.

[0173] After parsing the combination of meteorological parameter values ​​corresponding to each measurement bit string, in step S6, each combination of meteorological parameter values ​​is input into the Bayesian network model to obtain the corresponding lightning occurrence probability. Then, in step S7, the objective function value is calculated.

[0174] As mentioned above, the present invention uses formula (1-1) to describe the constraint conditions of the optimization problem as a problem of maximizing the probability of lightning occurrence:

[0175] (1-1)

[0176] in, d is an adjustable parameter, and d is a combination of meteorological parameter values; Let d be the probability of lightning occurring when the meteorological parameter values ​​are combined as d.

[0177] To utilize quantum circuitry for optimization, the problem of maximizing the probability of lightning occurrence is mapped to a minimization problem, as described by the following equation (1-2):

[0178] (1-2)

[0179] Then construct the objective function for minimizing the problem. Its expression is shown in formula (1-3):

[0180] (1-3)

[0181] Where Y represents the set of all quantum states obtained in this measurement, and y is one of the quantum states. Indicates the variable component quantum circuit in parameters The probability distribution of quantum states is measured in real time, and d is the combination of meteorological parameter values ​​obtained by analyzing the bit string x corresponding to quantum state y. The probability value of lightning is obtained based on a Bayesian network model, and its expression is shown in formula (1-4):

[0182] (1-4)

[0183] Combining formula (1-4), formula (1-3) can be mapped to formula (1-5):

[0184] (1-5)

[0185] Where j represents the measurement order, M represents the total number of measurements; d j Let y be the quantum state y based on the j-th measurement. j The corresponding bit string x j The combination of meteorological parameter values ​​obtained from analysis.

[0186] Correspondingly, in step S7, after measuring the final quantum state of the variable quantum circuit M times in this round of optimization, M bit strings x are obtained, denoted as X. i ={x i 1,x i 2,……, x i j ,……, x i M} Where i is a natural number representing the optimization order, j is the measurement index at the i-th optimization, M is the number of measurements, and each bit string x i j The analysis yields a target meteorological parameter value combination d. i j Therefore, in the i-th optimization process, the objective function value of the i-th optimization can be calculated using formula (1-5). .

[0187] In step S8, the convergence condition of the variable quantum circuit is determined when the objective function value reaches its minimum. The specific steps for determining when the objective function value reaches its minimum are as follows:

[0188] The objective function value is obtained in each calculation. At that time, calculate the difference between the objective function value obtained in the previous calculation and the value obtained in the previous calculation. For details, please refer to formula (5-1):

[0189] (5-1)

[0190] Determine the difference in the objective function value Is it less than 0? If A value less than 0 indicates that the current objective function value is less than the previous objective function value, meaning that the previously adjusted or set quantum circuit parameters met the requirements. If... A value greater than 0 indicates that the calculated objective function value after adjusting or setting the parameters has not changed in the direction of minimization, and also indicates that the adjusted or set parameters are inappropriate. This means that the difference in objective function values ​​is the key parameter in this invention. Instructions for adjusting parameters.

[0191] Then compare the difference in the objective function values. The difference between the value of the objective function and the preset value ε is used to determine the value of the objective function. Is it less than the preset value ε? If the difference in the objective function value is less than the preset value ε? If the difference is not less than the preset value ε, then continue adjusting the quantum circuit parameters for the next round of optimization. If the value is less than the preset value ε, it indicates that the optimization in the current round has met the requirements. Typically, to obtain stable optimization quantum parameters, the initial difference between the two values ​​must satisfy the objective function. Even when the value is less than the preset value ε, continue measuring or slightly adjusting the quantum circuit parameters until the difference between the objective function value and the preset value is satisfied after a certain number of consecutive optimization iterations (integers greater than 2) is reached. If the condition is less than the preset value ε, then the quantum circuit is considered to have converged and the condition for stopping optimization is met.

[0192] The preset value ε is a pre-defined threshold value used to limit the allowable range of changes in the objective function value in adjacent iterations. Its value is determined based on the range of the objective function value and the statistical error of quantum measurement. In one embodiment, the preset value ε is not less than the statistical fluctuation amplitude introduced by quantum measurement.

[0193] When the quantum circuit converges, in step S10, the method for determining the optimal combination of meteorological parameter values ​​based on the converged variable quantum circuit is as follows: Fix the variable quantum circuit parameters and perform multiple final measurements on the quantum circuit, with the number of measurements being much greater than the number of measurements during optimization to ensure that the probabilities of various quantum states can be measured. Then, the occurrence probabilities of various quantum states (corresponding to bit strings) are statistically analyzed, and the bit string corresponding to the quantum state with the highest probability is taken as the optimal bit string. The optimal combination of meteorological parameter values ​​is then obtained by parsing the optimal bit string, and finally, the corresponding lightning occurrence probability is obtained based on a Bayesian network model. Another method is as follows: Fix the variable quantum circuit parameters, perform multiple final measurements on the quantum circuit, with the number of measurements being much greater than the number of measurements during optimization to ensure that various quantum states can be measured, then parse the bit strings corresponding to the quantum states into combinations of meteorological parameter values, and finally, the corresponding lightning occurrence probability is obtained based on a Bayesian network model. The combination of meteorological parameter values ​​with the highest lightning occurrence probability is determined as the optimal combination of meteorological parameter values.

[0194] This invention, based on Bayesian networks, finds the meteorological parameter combination with the highest probability of lightning occurrence by traversing various meteorological parameter combinations. Then, using this as an initial combination, it finds another meteorological parameter combination with the highest probability of lightning occurrence based on a variable quantum circuit. The former optimizes the meteorological parameter combination, while the latter optimizes the specific meteorological parameter values ​​through combination optimization using a variable quantum circuit. During optimization, the objective function value is calculated using the conditional probabilities obtained from the Bayesian network, thus reflecting the global dependencies between multiple meteorological variables and imposing global constraints on the parameter search space. Furthermore, the parallelism of quantum computing significantly accelerates the optimization process, enabling rapid convergence to the optimal solution and improving the optimization speed.

[0195] In another aspect, the present invention also provides a parameter optimization system based on a variable quantum augmented Bayesian network, see [link to relevant documentation]. Figure 8 , Figure 8 This is a block diagram illustrating the principle of a parameter optimization system based on a variable quantum augmented Bayesian network according to an embodiment of the present invention. In this embodiment, the optimization system includes a dataset acquisition module 100, a Bayesian network module 101, an encoding module 102, a variable quantum circuit module 103, a bit string parsing module 104, and a tuning module 105. The dataset acquisition module 100 acquires a dataset of historical meteorological parameter value combinations, which includes parameter values ​​for various meteorological conditions, as well as corresponding geographical location parameter values ​​and time parameter values. The historical meteorological parameter value combination dataset includes multiple data points, each representing a historical meteorological parameter value combination.

[0196] At the start of optimization, the Bayesian network module 101 traverses the historical meteorological parameter value combination dataset based on the trained Bayesian network model, and selects the historical meteorological parameter value combination with the highest probability of lightning occurrence as the initial meteorological parameter value combination. In other words, it finds a historical meteorological parameter value combination with the highest probability of lightning occurrence and uses it as the initial meteorological parameter value combination. During the optimization process, it generates the corresponding lightning occurrence probability based on the meteorological parameter value combination transmitted by the bit string parsing module 104 and sends it to the tuning module 105. The Bayesian network model is a model constructed based on the probabilistic dependency between the meteorological parameter value combination and the lightning occurrence probability.

[0197] The encoding module 102 maps the initial meteorological parameter values ​​to initial bit strings corresponding to the initial quantum states of the variable quantum circuit. The variable quantum circuit module 103 runs the variable quantum circuit based on the set variable quantum circuit parameter values ​​to evolve the quantum states, and measures the final quantum state of the variable quantum circuit to obtain multiple bit strings corresponding to the final quantum state, which are then sent to the bit string parsing module 104. At the start of optimization, the variable quantum circuit parameter values ​​are initial values; during optimization, the variable quantum circuit parameter values ​​are the parameter values ​​determined by the tuning module 105.

[0198] The bit string parsing module 104 parses the bit string obtained from the variable quantum line measurement into a combination of meteorological parameter values ​​and sends it to the Bayesian network module 101.

[0199] During the optimization process, the tuning module 105 calculates the objective function value based on the lightning occurrence probability generated by the Bayesian network module 101. The objective function is a formula for calculating the lightning occurrence probability. Based on the objective function value, the variable quantum circuit parameters are determined until the variable quantum circuit converges. After convergence, the variable quantum circuit module 103 fixes the variable quantum circuit parameters and performs final multiple measurements on the quantum circuit. Then, it calculates the occurrence probability of various quantum states (corresponding to bit strings) and selects the bit string corresponding to the quantum state with the highest probability as the optimal bit string. The bit string parsing module 104 parses the optimal bit string to obtain the optimal combination of meteorological parameter values. The Bayesian network module 101 obtains the corresponding lightning occurrence probability based on the optimal combination of meteorological parameter values. Alternatively, after the variable quantum circuit converges, the variable quantum circuit module 103 fixes the variable quantum circuit parameters and performs multiple final measurements on the quantum circuit to obtain bit strings corresponding to various quantum states. Then, the bit string parsing module 104 parses the bit strings into combinations of meteorological parameter values ​​and sends them to the Bayesian network module 101. The Bayesian network module 101 obtains the corresponding lightning occurrence probability and determines the combination of meteorological parameter values ​​with the highest lightning occurrence probability as the optimal combination of meteorological parameter values.

[0200] The method and system provided by this invention, based on a dataset of historical meteorological parameter combinations, can not only determine which parameter combinations have a higher probability of causing lightning, but also further identify parameter values ​​with a higher probability of causing lightning from their respective numerical ranges. For example, see... Figure 9 , Figure 9 This is a statistical diagram illustrating the relationship between precipitation intensity and lightning density based on a dataset of historical meteorological parameter combinations. The horizontal axis represents precipitation intensity, and the vertical axis represents lightning density; each blue dot represents a data point. The red dashed line represents the precipitation intensity with the highest probability of lightning occurrence obtained using a simple Bayesian algorithm, while the green dashed line represents the precipitation intensity with the highest probability of lightning occurrence obtained using the parameter optimization method based on a variable quantum augmented Bayesian network provided in this invention.

[0201] See Figure 10 , Figure 10 This is a statistical diagram illustrating the relationship between air pressure and lightning density based on a dataset of historical meteorological parameter values. The horizontal axis represents air pressure, and the vertical axis represents lightning density; each blue dot represents a data point. The red dashed line represents the air pressure at which the probability of lightning occurrence is maximized using a simple Bayesian algorithm, while the green dashed line represents the air pressure at which the probability of lightning occurrence is maximized using the parameter optimization method based on a variable quantum enhanced Bayesian network provided in this invention.

[0202] See Figure 11 , Figure 11 This is a statistical diagram illustrating the relationship between cloud top height (cloud_top_hight) and lightning density based on a dataset of historical meteorological parameter values. The horizontal axis represents cloud top height, and the vertical axis represents lightning density; each blue dot represents a data point. The red dashed line represents the cloud top height with the highest probability of lightning occurrence obtained using a simple Bayesian algorithm, while the green dashed line represents the cloud top height with the highest probability of lightning occurrence obtained using the parameter optimization method based on a variable quantum augmented Bayesian network provided in this invention.

[0203] from Figures 9 to 11 As can be seen, the parameter optimization method of the variable quantum enhanced Bayesian network provided by this invention, under the constraint of maximizing the probability of lightning occurrence, introduces a variable quantum circuit as a search and optimization mechanism, which can more effectively discover high-risk meteorological condition combinations and correct parameter values ​​determined by simple algorithms.

[0204] The method and system provided by this invention can solve the problem of selecting and optimizing meteorological parameters in various application scenarios. For example, when constructing an aviation simulation environment or testing the tolerance of equipment, the parallel search advantage of quantum computing provided by this invention is used to accurately back-calculate the optimal combination of meteorological parameters leading to extreme lightning strike environments (such as obtaining a specific cloud height and aerosol concentration ratio) by maximizing the probability of lightning occurrence as a constraint. This provides a scientific and rigorous test benchmark for verifying the safety of the system in extreme environments, and improves the protection design by simulating extreme working conditions that are difficult for simple algorithms to detect. Similarly, in scenarios such as artificial lightning induction that require active weather control, this system no longer relies on manual trial and error, but instead converges quickly to the optimal solution through quantum circuit evolution, correcting the deviation parameter values ​​determined by simple algorithms, and providing operators with precise guidance for setting meteorological conditions. This significantly improves the success rate and efficiency of artificial lightning induction and rainfall operations, achieving a leap from "blind adjustment" to "precise optimization". In terms of lightning prediction, this invention provides a new prediction approach. By determining the combination of meteorological parameters when the probability of lightning occurrence is highest, this invention can provide simulated data for comparison with actual meteorological forecasts, thereby improving the accuracy of meteorological forecasts and reducing the damage of lightning to people's production and lives.

[0205] In another aspect, embodiments of the present invention also provide an electronic device, see [link to previous document]. Figure 12 , Figure 12 This is a structural principle block diagram of an electronic device according to an embodiment of the present invention, such as... Figure 12 As shown, the electronic device includes a processor 601 and a memory 602. The memory stores computer instructions. When the processor 601 executes the computer instructions, it performs the parameter optimization method based on variable quantum augmented Bayesian network provided by the present invention.

[0206] Specifically, processor 601 may include a central processing unit (CPU) or a graphics processing unit (GPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement embodiments of the present invention. Memory 602 includes data or instructions. For example, memory 602 may be at least one of the following: a hard disk drive (HDD), read-only memory (ROM), random access memory (RAM), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, Universal Serial Bus (USB) drive, or other physical / tangible storage device. Alternatively, memory 602 may include removable or non-removable (or fixed) media. Furthermore, memory 602 may be internal or external to the integrated gateway disaster recovery device. Memory 602 may be non-volatile solid-state memory. In other words, memory 602 typically includes a tangible (non-transitory) computer-readable storage medium (such as a memory device) encoded with executable instructions, wherein the stored executable instructions, when executed by processor 601 (such as by one or more processors), can implement the parameter optimization method based on variable quantum augmented Bayesian networks in the embodiments of the present invention.

[0207] In one example Figure 12 The illustrated electronic device may also include a communication interface 603 and a bus 610. The processor 601, memory 602, and communication interface 603 are connected via the bus 610 and communicate with each other. The communication interface 603 is mainly used to enable communication between modules, devices, units, and / or equipment within the electronic device.

[0208] Bus 610 may be hardware, software, or both. For example, the bus may include at least one of the following: Accelerated Graphics Port (AGP) or other graphics bus, Enhanced Industry Standard Architecture (EISA) bus, Front Side Bus (FSB), HyperTransport (HT) interconnect, Industry Standard Architecture (ISA) bus, Infinite Bandwidth Interconnect, Low Pin Count (LPC) bus, memory bus, Microchannel Architecture (MCA) bus, Peripheral Component Interconnect (PCI) bus, PCI-Express (PCI-X) bus, Serial Advanced Technology Attachment (SATA) bus, Video Electronics Standards Association Local (VLB) bus, or other suitable bus. Bus 610 may include one or more buses. Although specific buses are described or shown in embodiments of the invention, any suitable bus or interconnection may be considered in embodiments of the invention.

[0209] In another aspect, embodiments of the present invention also provide a computer-readable storage medium storing computer program instructions. When executed by a processor, these computer program instructions implement the aforementioned parameter optimization method based on a variable quantum augmented Bayesian network. The computer-readable storage medium includes classical computer-readable storage media, such as the aforementioned storage device 602, i.e., read-only memory (ROM), random access memory (RAM), disk storage media, optical storage media, flash memory, electrical, optical, or other physical / tangible storage devices. It may also include storage media for storing quantum information that are readable by a quantum computer, such as quantum random access memory (QRAM). QRAM can be considered a quantum version of RAM in a classical computer. Through QRAM, quantum superposition states containing information can be created. Compared to RAM, which requires reading each element individually, superimposed data can be read at superimposed addresses. QRAM can be implemented using physical methods such as optics, semiconductor quantum dots, superconducting circuits, and ion traps.

[0210] The flowcharts and / or block diagrams of the methods and systems of embodiments of the present invention have been described above by way of example, and related aspects have been described. It should be understood that each block or combination thereof in the flowcharts and / or block diagrams can be implemented by computer program instructions, by dedicated hardware performing a specified function or action, or by a combination of dedicated hardware and computer instructions. When implemented in hardware, it can be, for example, an electronic circuit, an application-specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc.; when implemented in software, it is a program or code segment used to perform the required task. The program or code segment can be stored in memory or transmitted over a transmission medium or communication link via data signals carried in a carrier wave. The code segment can be downloaded via a computer network such as the Internet, an intranet, etc.

[0211] The above embodiments are for illustrative purposes only and are not intended to limit the invention. Those skilled in the art can make various changes and modifications without departing from the scope of the invention. Therefore, all equivalent technical solutions should also fall within the scope of the invention.

Claims

1. A parameter optimization method based on variable quantum augmented Bayesian networks, characterized in that, include: Step S1: Obtain a dataset of historical meteorological parameter value combinations, which includes parameter values ​​for various meteorological conditions, as well as corresponding geographical location parameter values ​​and time parameter values. Step S2: Based on the trained Bayesian network model, traverse the historical meteorological parameter value combination dataset and take the historical meteorological parameter value combination with the highest probability of lightning occurrence as the initial meteorological parameter value combination; wherein, the Bayesian network model is a model constructed and trained based on the probabilistic dependency between the meteorological parameter value combination and the probability of lightning occurrence. Step S3: Set the initial values ​​of the variable quantum circuit parameters and map the initial meteorological parameter values ​​to the initial bit string corresponding to the initial quantum state of the variable quantum circuit. Step S4: Run the variable quantum circuit to evolve the quantum state, and measure the final quantum state of the variable quantum circuit to obtain multiple bit strings corresponding to the final quantum state; wherein, the variable quantum circuit includes an initial state encoding module, a correlation module, a parameterized quantum gate module, and a measurement module. The initial state encoding module is used to encode the initial bit string into the initial quantum state of the variable quantum circuit; the correlation module includes multiple entanglement gates acting on the qubits representing any two meteorological parameters, used to apply entanglement between the multiple qubits representing meteorological parameters to establish a correlation between arbitrary meteorological parameters; the parameterized quantum gate module includes multiple layers of quantum circuit units connected end to end; Step S5: Parse each bit string into a corresponding combination of meteorological parameter values; Step S6: Input the combined meteorological parameter values ​​into the Bayesian network model to obtain the corresponding lightning occurrence probability; Step S7: Calculate the objective function value based on the probability of lightning occurrence, wherein the objective function is a formula for calculating the probability of lightning occurrence; Step S8: Adjust the variable quantum circuit parameters based on minimizing the objective function value; Repeat steps S4 to S8 until the convergence condition of the variable quantum circuit is met, and obtain the optimal combination of meteorological parameter values ​​based on the converged variable quantum circuit.

2. The parameter optimization method based on variable quantum augmented Bayesian networks according to claim 1, characterized in that, Further includes: The training sample set for the Bayesian network model is constructed based on historical data. Each training sample includes a set of meteorological parameter values ​​and lightning occurrence parameter values. The meteorological parameter values ​​include parameter values ​​for various meteorological conditions, corresponding geographical location parameter values, and time parameter values. A directed acyclic graph is constructed using meteorological parameters and lightning occurrence parameters as nodes, and the influence relationship between meteorological parameters and lightning occurrence as edges. Construct a Bayesian network model based on the directed acyclic graph; and The Bayesian network model is trained using the maximum likelihood estimation method based on the training sample set.

3. The parameter optimization method based on variable quantum augmented Bayesian networks according to claim 2, characterized in that, The steps for constructing a training sample set for a Bayesian network model based on historical data include: Obtain the raw lightning metadata dataset, where each piece of raw lightning metadata in the raw lightning metadata dataset is a three-dimensional grid data including geographical location, time, and lightning density; The original lightning metadata of the three-dimensional grid is unfolded into one-dimensional lightning data, which includes geographical location, time and lightning density; Obtain the raw meteorological dataset, wherein each piece of raw meteorological data in the raw meteorological dataset includes geographical location, time, and multiple meteorological condition parameter values; Align lightning data and raw weather data according to geographic location and time; The aligned lightning data and raw meteorological data are fused together to obtain primary training sample data; and The primary training sample data is preprocessed, and lightning occurrence parameter data is constructed based on the lightning density data in the primary training sample to obtain the training sample data.

4. The parameter optimization method based on variable quantum augmented Bayesian networks according to claim 3, characterized in that, The steps for constructing the training sample set for the Bayesian network model further include: Based on the Monte Carlo method, a preset number of meteorological condition parameter values ​​are generated according to the data distribution characteristics of each meteorological condition; and The meteorological condition parameter values ​​generated based on the Monte Carlo method are aligned with and fused with lightning data to obtain new training sample data.

5. The parameter optimization method based on variable quantum augmented Bayesian networks according to claim 1, characterized in that, The step of mapping the combination of initial meteorological parameter values ​​to an initial bit string corresponding to the initial quantum state of the variable quantum circuit further includes: Map each meteorological parameter value in the initial combination of historical meteorological parameter values ​​to a bit string of one or more binary bits; The initial bit string is obtained by sequentially combining the bit strings corresponding to each meteorological parameter value; The steps of mapping each meteorological parameter value to a bit string of one or more binary bits include: Determine the parameter index for each meteorological parameter; and Query the table that maps meteorological parameters to bit strings to obtain the corresponding bit strings.

6. The parameter optimization method based on variable quantum augmented Bayesian networks according to claim 5, characterized in that, This further includes the step of constructing a mapping table between meteorological parameter indexes and bit strings: When the value of the meteorological parameter is a continuous value, the first value range of the meteorological parameter is obtained, and the first value range is a continuous value interval. The first numerical range is divided into N discrete numerical intervals, where N=2. n n is a natural number, and each discrete numerical interval is numbered to obtain the parameter index table of the meteorological parameters; The parameter indices in the parameter index table are mapped sequentially into bit strings of length n in ascending order to obtain a correspondence table between parameter indices and bit strings; When the value of the meteorological parameter is a discrete value, a second numerical range of the meteorological parameter is obtained, the second numerical range includes n discrete values, where n is a natural number; Number the n discrete values ​​within the second numerical range to obtain a parameter index table for the meteorological parameters; and Each parameter index in the parameter index table is mapped to a bit string of length n according to its position in the table; wherein, the position of each parameter index in the table is set to a first value, and the remaining positions are set to a second value. When the first value is 0, the second value is 1; when the first value is 1, the second value is 0.

7. The parameter optimization method based on variable quantum augmented Bayesian networks according to claim 5 or 6, characterized in that, In step S5, the step of parsing the bit string into a combination of meteorological parameter values ​​includes: The bit string is segmented to obtain bit substrings corresponding to each meteorological parameter; Based on the bit substring of each meteorological parameter, query the table corresponding to the parameter index and bit string of each meteorological parameter to obtain the corresponding parameter index; When the values ​​of meteorological parameters are continuous, the corresponding parameter range is obtained based on the parameter index; A parameter value is determined within the parameter range based on a preset parameter value generation strategy; When the meteorological parameter value is a discrete value, the corresponding discrete value is obtained based on the parameter index and used as the parameter value; and The meteorological parameter values ​​are combined according to the order of the bit substrings in the bit string to obtain the meteorological parameter value combination corresponding to the bit string.

8. The parameter optimization method based on variable quantum augmented Bayesian networks according to claim 7, characterized in that, The steps for determining a parameter value within the parameter range based on a preset parameter value generation strategy include: Obtain the parameter value mapping function formula for the parameter interval of the corresponding meteorological parameter, wherein the parameter value mapping function formula includes an unknown variable; Obtain the probability of the corresponding bit substring output by the current variable quantum circuit; and The probability of the bit substring is assigned to the unknown variable in the mapping function formula to obtain the corresponding parameter value.

9. The parameter optimization method based on variable quantum augmented Bayesian networks according to claim 7, characterized in that, The steps for determining a parameter value within the parameter range based on a preset parameter value generation strategy include: Determine all sample parameter values ​​within the parameter range from historical data; The probability of lightning occurrence is generated based on the Bayesian network model for each sample parameter value; and The parameter value with the highest probability of lightning occurrence is determined as the parameter value.

10. The parameter optimization method based on variable quantum augmented Bayesian networks according to claim 1, characterized in that, Further includes: The constraints for the meteorological parameter optimization problem are constructed, where the constraints are described as a lightning occurrence probability maximization problem using expression (1-1): （1-1） in, d represents an adjustable parameter of a variable quantum line, where d is a combination of meteorological parameter values. This represents the probability of lightning occurrence under the meteorological parameter value combination d. The problem of maximizing the probability of lightning occurrence is mapped to a minimization problem, which is described by expression (1-2): （1-2） Construct the objective function of the minimization problem Its expression is shown in formula (1-3): （1-3） Where Y represents the set of all quantum states obtained in this measurement, and y represents the quantum state at each measurement. Indicates the variable component quantum circuit in parameters The probability distribution of the quantum state y obtained by time measurement, and d is the combination of meteorological parameter values ​​obtained by parsing the bit string x corresponding to the quantum state y. The probability value of lightning is obtained based on a Bayesian network model, and its expression is shown in formula (1-4): （1-4） Combining formula (1-4), formula (1-3) can be mapped to formula (1-5): （1-5） Wherein, j represents the measurement order, and M represents the total number of measurements; Let y be the quantum state obtained based on the j-th measurement. j The corresponding bit string x j The combination of meteorological parameter values ​​obtained from the analysis; Correspondingly, in step S7, the objective function value is calculated based on formula (1-5).

11. A parameter optimization system based on a variable quantum augmented Bayesian network, characterized in that, It includes a dataset acquisition module, a Bayesian network module, an encoding module, a variable quantum circuit module, a bit string parsing module, and an optimization module; The dataset acquisition module acquires a dataset of historical meteorological parameter value combinations, which includes parameter values ​​for various meteorological conditions, as well as corresponding geographical location parameter values ​​and time parameter values. At the start of optimization, the Bayesian network module iterates through the historical meteorological parameter value combination dataset based on the trained Bayesian network model, and uses the historical meteorological parameter value combination with the highest probability of lightning occurrence as the initial meteorological parameter value combination. During the optimization process, the corresponding lightning occurrence probability is generated based on the meteorological parameter value combination from the bit string parsing module. The Bayesian network model is a model constructed based on the probabilistic dependency between the meteorological parameter value combination and the lightning occurrence probability. The encoding module maps the initial meteorological parameter values ​​into an initial bit string corresponding to the initial quantum state of the variable quantum circuit. The variable quantum circuit module operates the variable quantum circuit based on the set variable quantum circuit parameter values ​​to evolve the quantum state, and measures the final quantum state of the variable quantum circuit to obtain multiple bit strings corresponding to the final quantum state. At the start of optimization, the variable quantum circuit parameter values ​​are initial values; during optimization, the variable quantum circuit parameter values ​​are the parameter values ​​determined by the tuning module. The variable quantum circuit includes an initial state encoding module, a correlation module, a parameterized quantum gate module, and a measurement module. The initial state encoding module encodes the initial bit string into the initial quantum state of the variable quantum circuit. The correlation module includes multiple entanglement gates acting on qubits representing any two meteorological parameters, used to apply entanglement between the multiple qubits representing meteorological parameters to establish a correlation between arbitrary meteorological parameters. The parameterized quantum gate module includes multiple layers of interconnected quantum circuit units. The bit string parsing module parses the bit string obtained from measuring the terminal quantum state of the variable quantum line into a combination of meteorological parameter values; The optimization module sets the variable quantum line parameters to initial values ​​at the start of optimization. During the optimization process, it calculates the objective function value based on the lightning occurrence probability generated by the Bayesian network module. The objective function is a formula for calculating the lightning occurrence probability. The variable quantum line parameters are adjusted by minimizing the objective function value until the convergence condition of the variable quantum line is met. Among them, the optimal combination of meteorological parameter values ​​is obtained based on the convergent variable quantum circuit.

12. An electronic device comprising a processor and a memory, wherein the memory stores computer instructions, characterized in that, When the processor executes the computer instructions, it performs the parameter optimization method based on the variable quantum augmented Bayesian network as described in any one of claims 1-10.

13. A computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are executed by the processor to perform the parameter optimization method based on the variable quantum augmented Bayesian network as described in any one of claims 1-10.

Images